XFX GeForce 6800 Xtreme – an eBay special (with LONG story)

[momaka]

So I bought this gem at the end of last summer on eBay (or thereabout). Why I haven't posted it earlier? Well, because I am slow and tend to leave pictures on my hard drive until they ferment a good amount before I post them . Anyways that moment is here and now.

Ladies and gentlemen, brace yourselves, as I have another ghetto mod… except this time, I wasn't the mastermind behind it.

https://www.badcaps.net/forum/attach...1&d=1444698583
https://www.badcaps.net/forum/attach...1&d=1444698583

So let's see what we have here:
Old socket 370 cooler to cool the heatsink – check
Zip ties – check
Bulged caps - check
Dust, dirt, filth, and some small dead roach bodies – check
Awful smell of cigarettes that I am glad cannot be captured as part of the picture – check

Now you may be wondering, why would anyone sane buy something like this? Well, maybe that's because I am not . That, and the fact that the auction started at just a penny and no one bid on it at all.

So yes, I got this video card for just $0.01 – a great deal if you don't factor in the shipping (which was something along the lines of $8-12, I don't remember exactly anymore.) But what really interested me was the seller's ghetto mod. Moreover, he described and pictured this video card exactly as it was (minus the bulged caps, but I saw them before buying). He warned of the heavy cigarette smell, explained how and why the mod was done (original video card fan failed), posted GPU-Z screenshots of the video card running, and even offered DOA warranty. In short, he just said he didn't want to throw it away, since he had it running for a long time and so would rather someone find a home for it. Given that no one bid and that the video card had a lot of stuff to be fixed (but was otherwise functional), I decided to grab it just as a fun project.

When the shipping box arrived at my door, I immediately noticed the strong smell of cigarettes. So the first thing I did – take it outside for a wash. First, I removed the socket 370 heatsink and fan. Next, I put the garden hose on full tilt and gave the video card a good blast. Immediately, the card looked 100x better. Of course, there was still quite some room for improvement. Using an old tooth brush and a generous amount of dish detergent, I scrubbed every surface at least twice. After another rinse with the hose, the card looked great. Unfortunately, that was only as far as the visuals went. The strong smell of smoke was still there even after the wash. So, the next thing I did is I took off the GPU heatsink and gave the board a quick wipe with a bit of alcohol.

Now that did the trick for the smell . The result of all of that cleaning:
https://www.badcaps.net/forum/attach...1&d=1444698583

You can also see on this picture that the original fan was indeed damaged. The seller said his son tried to clean it because it was seized, but the fan was so stuck that eventually the whole fan blade assembly broke. Fortunately, he thought this might be repairable, so he saved all of the parts. Thus, the auction also included all of the broken fan parts in a small bag.

Upon inspecting them, I found out why the fan broke: the magnet sensor/FET was cooked. This likely overheated the coils in the stator and caused the plastic to become fragile. Another notable thing is the rust on the magnet of the rotor. It is more than just surface rust – it is deep. Perhaps this caused the fan to stall in the first place?
- I think so, because this is not the first time I have seen this. An MSI GeForce 7600 GS (given to me by a classmate) also had this issue. On that one, even the coils were burned to a crisp! Other cards on which I have indirectly seen this bad fan are PNY GeForce 8500 (or maybe it was a 9500) as well as a few other select ATI cards, often described as having "bad", "stuck", or "rusted" fans on eBay. Looking at the two fans I have from the GeForce 7600 GS and this 6800, it appears the fans are made by the same manufacturer. There is no label to identify the fan from the 6800, but the 7600 fan has "ARX CeraDyna FAN" written on the sticker. UL number E145724. Its PCB, however, has UL number E210118. Most notable, perhaps, is that both fans have plastic shafts on the rotor assembly, and the PCBs are nearly identical. So I think this is definitely one and the same company that makes them. Here is a picture of the two fans, with the XFX 6800 fan on the left and the MSI 7600 fan on the right:
https://www.badcaps.net/forum/attach...1&d=1444698583

Finally, the capacitors: those Evercon ME really needed to go – even the group of three at the top that aren't bulged. For those of you who are new to the forums, avoid GSC, Evercon, Sacon, and Elcon like the plague! They are the most terrible electrolytic capacitors in the history of electronics (or maybe strongly rivaled by Rulycon… but let's not even go there. )

Starting with the four capacitors at the top right corner by the power connector (that's CE4/CE27, CE6/CE25, CE5/CE26, and CE30/CE31), those filter the GPU V_core. Originally populated by three 6.3 V, 1500 uF Evercon ME 10 mm caps. I don't think they were bad yet - at least they didn't appear so, but I didn't have an ESR meter handy to check them. I changed them for two 4V, 820 uF Fujitsu FPCAP RE polymers, accompanied by one 6.3V, 2200 uF Nichicon HZ. This combo may have slightly lower total capacitance, but the ESR and ripple current specs are much more superior that the original setup. For anyone recapping this video card, feel free to experiment here. There are four cap spots in total, and the board has holes for both 8 mm and 10 mm diameter thru-hole caps (or if you are brave, SMT ). Given the not-so-great specs of Evercon ME, it is possible to use even something like Panasonic FR and FM, Nichicon HW, Rubycon ZLG, or Chemicon KZM.

The next set of caps that really needed my attention – the two bulged 6.3 V, 1500 uF Evercon ME 10 mm caps filtering the RAM Vdd rail. I've looked at other XFX GeForce 6800 Xtreme video cards on eBay, and they always seem to have these two caps bulged. These are in spots CE23/CE28 and CE24/CE29. Much like the GPU V_core caps, the board gives options here again for both 8 mm and 10 mm thru-hole caps as well as 8/10 mm SMT. I had a 6.3V, 1500 uF Rubycon MBZ 10 mm cap that I wanted to test (it appeared good on ESR meter before, even though it bulged once due to high heat on reflow machine), as well as a 10V, 470 uF Nichicon HC 10x12 cap from a Radeon 9500/9700 video card. On paper, both of these caps have reasonably close ESR and ripple current to Evercon ME's. So I tried them out, and they worked fine on my card.

Finally, there are the 12 V input filters for the VRMs above (i.e. GPU V_core and RAM Vdd). It appears that the GPU V_core VRM draws power from the 12 V rail on the 4-pin power connector, whereas the RAM Vdd VRM draws its power from the 12 V rail on the PCI-E connector.

The GPU VRM input filter caps are in spots CE2/CE10 (which filters the 12 V rail directly as it enters the power connector) and CE1/CE11 (which filters the 12 V rail after a 1 uH inductor before it is chopped by a MOSFET). Both of these caps are 16 V, 470 uF, 8 mm diameter SMT style. I don't know their brand, though (if "5 D" on the first line means anything to you). The board again gives the same three options as for the GPU V_core and RAM Vdd. I tried removing one of these caps, but I couldn't. Since they weren't bad, I left them alone.

As for the RAM Vdd VRM input filter, it has a single 16 V, 470 uF L.H. Nova LE 8 mm diameter cap sitting after a 1 uH inductor. Cap spot is CE16/CE17. Since L.H. Nova is another crappy brand, I replaced it with a 16 V, 1500 uF Nichicon HN 10x20 mm cap – a bit overkill, but I figured it's good to have the extra filtering so that there is less ripple sent back through the PCI-E's power pins.

And that is all of the caps for the major rails. There is also a 10 V, 560 uF Sanyo SVP 10x12 mm SMT polymer cap on my board, in spot CE3. I suspect it is filtering the secondary GPU rail (around 1-1.2 V IIRC) or a 5 V rail. Either way, it is a polymer, so I didn't touch it.

Here is what my recap looked after I was done with it and ready to temporarily test the video card (thus, pay no attention to the ghetto attached fan .)
https://www.badcaps.net/forum/attach...1&d=1444698583
https://www.badcaps.net/forum/attach...1&d=1444698583

Everything worked great on my tests (which consisted of me just playing a bunch of games for an hour or so). Therefore, there was one last thing left to do: attach a fan in a less ghetto manner. Since I have two dead GeForce 7900 video cards, I pulled the squirrel cage fan from one and put it on the 6800 (because apparently the 7900 heatsinks are not compatible with the 6800). The fan fit perfectly. It just needed something on top so that the air would get forced through the heatsinks. For this, I cut out a large piece of transparent PET/PETE plastic from a fruit container. Some bending and cutting, and this is what it looked like in the end:



I don't want to praise myself too much, but personally I think it looks even better than the original XFX heatsink. With two green LEDs, I think it will look even showcase worthy! (Okay, I give myself too much credit now. ).

... And I hit post text limit... lol?
TO BE CONTINUED...

[momaka]

Re: XFX GeForce 6800 Xtreme – an eBay special (with LONG story)

...Continiung from above...

The only problem: that fan from the 7900 is quite loud. I guess the original fan on this 6800 was slower, so the default settings supply higher voltage at idle. And during 3D operation, the fan is at full blast, which makes this little 12 V, 0.18 A Delta fan really scream. What's worse is that the GPU core temperatures are still pretty high – up to 60C in games! (Which is getting very close to the limit where these GeForce chips start to fail from the bumpgate problem.) But I suspect that is just due to the design of this heatsink (with the GPU die right under the fan rather than under the fins). Despite liking the looks so far, I might mess around with the fan and heatsink a bit more in the future to improve the temperatures (especially if the card will be used for 3D stuff.)

Which leads to my last point… gaming. Is this video card really worth anything these days in terms of games? -Absolutely not! It struggles at 20-30 FPS average in Half-Life 2 Episode 2 (at 1024x768 resolution and most settings maxed except for AA, which is OFF). And last part of test chamber 18 in Portal (with which I have tested on most of my other video cards) also makes it struggle a bit at 20-ish FPS. The other levels run much better, though. Slightly older games like Collin McRae Rally 2004 run butter-smooth at 80 FPS average on 1024x768 resolution. So I guess the 6 series GeForce is just too old. According to benchmarks, the GeForce is about on par with the Radeon 9800, which is pretty ancient now. But that shouldn't surprise you, as the 6800 has 8 ROPs. 8 pixel shaders, and 4 vertex shaders – exactly the same as the Radeon 9800.

Here is a GPU-Z of my GeForce 6800 Xtreme:


Nonetheless, I have to say that this was still another fun video card restoration/"upcycling" project for me. The video card doesn't see much use now, since it isn't very powerful. Every once in a while, though, I do pop it into my test PCs. And it definitely works fine as a functional and cool-looking "wall art" piece
So who knows, maybe someday this will be the last running XFX GeForce 6800 .

[mockingbird]

Re: XFX GeForce 6800 Xtreme – an eBay special (with LONG story)

Hmmm A1 silicon. That's good I think. Probably doesn't suffer from the design defects of the later revisions.

Good score for what you paid. The card should actually do quite well if you disable antialiasing and anisotropic filtering. They used to take a heavy hit back in those days with them enabled. You can also vastly increase the framerate if you run the games in 16-bit color -- trust me, it's not as bad as you think. I used to do it all the time and it was well worth the quality hit not having to spend hundreds on a new card.

The 16-bit trick won't work on Radeon cards though. Because of the architecture, the performance on them is the same regardless of the color depth.

[ChaosLegionnaire]

Re: XFX GeForce 6800 Xtreme – an eBay special (with LONG story)

excellent repair/recap, cleaning and modding, man! i luv to refurbish retro hardware like this. how did u manage to clean it until it was just like new?! thats astonishing!

i especially know the value of the gf 6 series cuz its the only dx9c series of cards supported on good ol win98. the gf 7 series are driver supported on xp or later only. same with the ati x100 (still dx9b) series and x1000 series being driver supported on xp or later. the gf 6800 series were also one of the best cards back in the day to play doom 3.

one probable cause of the poor performance in recent games u are experiencing might be due to the driver version u are using. from the gpu-z screenshot, i saw u are using forceware 93 or something. that is quite an old driver back from 2006 or thereabouts. i suggest u update to a newer driver like the last driver supporting the gf 6 series like forceware 295.73. nvidia deprecated driver support for the gf 6 and 7 series in the forceware 300 or newer drivers. so the forceware 295 drivers or below should be the last and most recent drivers supporting your card.

i have the 6800 ultra and 6600 gt (both purchased second hand) running the third party customised xtreme-g 295.73 driver and both cards work great.

judging by the paper specs of the 6800 xt on wikipedia, it has identical core config and fillrates as the radeon 9700 pro. the only two exceptions i can see is with the mem bandwidth and core clock speed. yours seems to be running with a 25mhz core factory overclock (6800 xt core stock speed is 325 mhz) and extra higher speed or factory overclocked memory at 500mhz (1ghz effective ddr with 32gb/s of resultant mem bandwidth) when the stock mem speeds for the 6800 xt should be 350mhz (700mhz effective ddr with only 22.4gb/s of resultant mem bandwidth). so the end result is a card that is a bit faster than the 9700 pro. im guessing the performance is in the ballpark of the 9800 pro or 9800 xt like u said.

with the higher mem speed, your card should be bottlenecked less by anti-aliasing and anisotropic filtering. u can try overclocking your card to take advantage of the extra memory bandwidth and also thereby boosting the pixel fillrate if u are concerned about the card being too slow. however, that could accelerate the bumpgate problem and not recommended since its running close to the thermal limit of the bumpgate issue.

which brings me to a suggestion on improving the cooling of the card. that small fan of a similar size to the radeon 9700/9800 stock fan may not cut it for cooling the card since the 9700/9800 stock hsf is inadequate for those cards either and ppl often replace them. i noticed that it looks like the mounting screw holes for the card form a perfect square? if that is so then perhaps u could find a 60/70/80mm fan with fan mounting holes that align with the mounting holes of the heatsink perfectly? then drill holes in the heatsink or remove the silver retaining nut in the heatsink and use long screws to bolt the fan and heatsink together on the card? one disadvantage tho, is that the hsf of the card will occupy and take up the adjacent pci slot or two. another disadvantage is that it may warp the card and accelerate the bumpgate problem or u could ghetto a metal backplate for the card or use a metal brace to reinforce the top side of the card.

i did sumthing like that for my passive cooled fanless radeon hd 5450 (conventional pci version for htpc use on old machines as an mpeg4 accelerator card). i found that a 50mm delta hhb fan i had, aligned with the heatsink mounting holes for the card perfectly. so i removed the push pins for the heatsink and bolted the fan and heatsink together with 30mm long screws. that way if the card spoilt, i could just put the push pins back and rma the card and the manufacturer would be none the wiser.

hope u have as much fun modding and refurbishing good ol' stuff like i did!

[momaka]

Re: XFX GeForce 6800 Xtreme – an eBay special (with LONG story)

Hmmm A1 silicon. That's good I think. Probably doesn't suffer from the design defects of the later revisions.

It's possible. After all, it didn't fail with the seller's ghetto mod, which I doubt worked too well. And it didn't fail with the stuck fan either. Or I got lucky .

Good score for what you paid. The card should actually do quite well if you disable antialiasing and anisotropic filtering. They used to take a heavy hit back in those days with them enabled.

Yes, anti-aliasing (AA) can take a big hit on some older cards. That said, I don't remember noticing too much of a performance hit when it is set to 2x on this video card. Maybe 5-10 FPS at the most. But I usually keep it off, since I primarily use CRT monitors, which even at 1024x768 look fairly smooth thanks to their non-square pixels.

As for anisotropic filtering (AF), I never found there to be any significant gain in performance when it is disabled - not in the Source engine anyways. Even on really old video cards like my 64 MB Asus Radeon 9200 SE, the difference is 1-2 FPS. So I leave it on 8x usually. Now, if I was to switch to bi-linear filtering, IIRC, that did make some difference. But the visuals really don't look as good with AF set to bi-linear mode.

The biggest performance hit for this video card (and my eVGA 6200/6600) are really from the resolution. 1024x768 is the limit for games running the Source engine. Anything beyond that runs terrible. And even a low 800x600 resolution doesn't make too much of an improvement. Then again, my test PC has the no-Steam version of the Source games I own, so it could be that those were optimized for newer hardware.

The 16-bit trick won't work on Radeon cards though. Because of the architecture, the performance on them is the same regardless of the color depth.

^ That is some good info. Thanks for sharing .
Yeah, since most of the video cards in my collection are ATI Radeon, I never noticed much gain in 16-bit, so I always left it at 32-bit.

excellent repair/recap, cleaning and modding, man! i luv to refurbish retro hardware like this. how did u manage to clean it until it was just like new?! thats astonishing!

Thanks!
I find dish detergent to be great at de-greasing and dissolving grime. I used to use clothes detergent before, and it is good too - but dish detergent beats it. Also, I always save my used toothbrushes exactly for this reason . They can scrub PCBs very well without damaging the components.

i especially know the value of the gf 6 series cuz its the only dx9c series of cards supported on good ol win98.

Cool!
The thing is, this is a PCI-E video card. On PCs with PCI-E slots, I usually don't put anything older than XP. But it's good to know in case I do set up a Windows 98 system. I have a few very old games that simply run too finicky on 2000/XP. Fortunately, those games will run okay even on anemic video cards like S3 Savage 4. Talk about oldschool there .

one probable cause of the poor performance in recent games u are experiencing might be due to the driver version u are using. from the gpu-z screenshot, i saw u are using forceware 93 or something. that is quite an old driver back from 2006 or thereabouts. i suggest u update to a newer driver like the last driver supporting the gf 6 series like forceware 295.73.

Ah okay, I'll give that a try.
Perhaps that is indeed the reason I don't get very good framerates.

I picked the older driver on purpose, because back in the day when I was still rocking my 9200 SE, the drivers upgrades rarely did anything good in terms of performance. Often, the newer ATI drivers just ended up eating more memory, which was an issue with my old PC (it was a true "built-on-a-budget" PC )

the gf 6800 series were also one of the best cards back in the day to play doom 3.

That's one game I never got into, for some reason. Same with GTA series. Mafia was good alternative to that, though (IMO).

yours seems to be running with a 25mhz core factory overclock (6800 xt core stock speed is 325 mhz) and extra higher speed or factory overclocked memory at 500mhz (1ghz effective ddr with 32gb/s of resultant mem bandwidth) when the stock mem speeds for the 6800 xt should be 350mhz (700mhz effective ddr with only 22.4gb/s of resultant mem bandwidth).

You know, I forgot to check what the memory chips were on my 6800. Normally I do that so I can see what the chips can handle (and what is their normal operating speed). But I don't think XFX overclocked the RAM on this video card. Most likely, they just went with high-end chips that really are capable of running 1 GHz effective DDR.

so the end result is a card that is a bit faster than the 9700 pro. im guessing the performance is in the ballpark of the 9800 pro or 9800 xt like u said.

Well, I only have a Radeon 9700 (non-Pro) to compare it to. But yes, they were running more or less the same. Maybe the 6800 had only a slight edge over the 9700.

My test system consisted of an AMD A64 3200+ @ 2.5 GHz and 1 or 1.5 GB RAM @ 200 MHz. Motherboard was AsRock 939Dual-SATAII mobo, which has both PCI-E and AGP slot (though the AGP seems to be running to the SB, so it might be running at PCI bandwidth, I don't know).

which brings me to a suggestion on improving the cooling of the card. that small fan of a similar size to the radeon 9700/9800 stock fan may not cut it for cooling the card since the 9700/9800 stock hsf is inadequate for those cards either and ppl often replace them.

Yes, the heatsinks on the Radeon 9700 and 9800 were terribly undersized. I have three working 9700's right now (I bought 8 of them on eBay a long time ago, all faulty/as-is). They have been modded with one of those improved Xbox 360 GPU heatsinks. With a 120 mm fan blowing in the case close to the video card, the heatsink gets only slightly warm. I'd guesstimate the die is running around 60C, since my 6200/6600 I mentioned above has the same modded heatsink, and it runs around 53C max in games with the same 120 mm fan setup.

Hehe, ghetto modding heatsinks is nothing new to me now. I think I might try the modified heatsink I made for my 7900. It consists of a Xbox 360 CPU Al heatsink. I just have to see if the 6800 has those odd bolting holes that the 7900 has.

hope u have as much fun modding and refurbishing good ol' stuff like i did!

Absolutely!

I always learn something new from doing this stuff *and* since most of it is old hardware, it's not like I wasted a lot of cash if I did something wrong. Of course, I still try to be as careful as possible.

[goodpsusearch]

Re: XFX GeForce 6800 Xtreme – an eBay special (with LONG story)

The end product is just so beautiful to look at...

[Wester547]

Re: XFX GeForce 6800 Xtreme – an eBay special (with LONG story)

Very nice repair, as always. GSC capacitors, never failing to disappoint.

Yes, anti-aliasing (AA) can take a big hit on some older cards. That said, I don't remember noticing too much of a performance hit when it is set to 2x on this video card. Maybe 5-10 FPS at the most. But I usually keep it off, since I primarily use CRT monitors, which even at 1024x768 look fairly smooth thanks to their non-square pixels.

Graphics cards that lack raw power - IE memory bandwidth, fill rate, a narrow memory bus - will suffer when high FSAA levels are enabled (as well as in games that rely more on traditional texture operations for rendering, rather than taking advantage of more modern shader architecture). FWIW, the R300/Radeon 9700 Pro was the first graphics card to truly make FSAA usable because of the 256-bit memory interface (the increased number of TMUs and ROPs also helped). What an incredibly forward looking card for its time - four years ahead of its time (2002), actually. As of nine years ago, it was still capable of running the then modern games at decent settings. Can't say the same now adays, because everything adheres to that dreadful vehicle called planned obsolescence.

And yeah, CRTs' built-in FSAA features are indeed a plus. If that fails me, I just use my glasses (being a tad near sighted) to make the jaggies go away. Real-life anti-aliasing. Awesome.

^ That is some good info. Thanks for sharing .
Yeah, since most of the video cards in my collection are ATI Radeon, I never noticed much gain in 16-bit, so I always left it at 32-bit.

Yes, this would explain why my Radeon 9500 Pro was never faster, sometimes even slower, in 16-bit mode. Ah well. It was about 5-30x faster than the GeForce 2 MX that preceded it at the time that I "upgraded" (12.5 years ago), so how could I complain of it?

[Joe Black]

Re: XFX GeForce 6800 Xtreme – an eBay special (with LONG story)

A splendid job!

[Wester547]

Re: XFX GeForce 6800 Xtreme – an eBay special (with LONG story)

I picked the older driver on purpose, because back in the day when I was still rocking my 9200 SE, the drivers upgrades rarely did anything good in terms of performance. Often, the newer ATI drivers just ended up eating more memory, which was an issue with my old PC (it was a true "built-on-a-budget" PC )

On my older cards, I keep away from the latest drivers. All they've ever done is break more than they fixed - another part of the "planned obsolescence" model.

That's one game I never got into, for some reason. Same with GTA series. Mafia was good alternative to that, though (IMO).

I always loved the DOOM games, even though they are more cult classics than masterpieces.

Yes, the heatsinks on the Radeon 9700 and 9800 were terribly undersized.

Well, they didn't output too much heat for the high end cards of the time, compared to the heat output of the "high end" cards to come thereafter, with much bigger heatsinks (IE the GeForce 8800 series).

Both of these caps are 16 V, 470 uF, 8 mm diameter SMT style. I don't know their brand, though (if "5 D" on the first line means anything to you.

Those are very likely Sunco AX SMT series - the font, as well as the vent stamp, is a give away. "5D" is probably the date code - December of 2005. Look at the datasheet attached.

[lti]

Re: XFX GeForce 6800 Xtreme – an eBay special (with LONG story)

Upon inspecting them, I found out why the fan broke: the magnet sensor/FET was cooked. This likely overheated the coils in the stator and caused the plastic to become fragile. Another notable thing is the rust on the magnet of the rotor. It is more than just surface rust – it is deep. Perhaps this caused the fan to stall in the first place?
- I think so, because this is not the first time I have seen this. An MSI GeForce 7600 GS (given to me by a classmate) also had this issue. On that one, even the coils were burned to a crisp!

It's also possible that the motor driver wasn't able to detect that the fan had seized and kept power applied to that pair of coils until they burned up.

Most notable, perhaps, is that both fans have plastic shafts on the rotor assembly

ARX CeraDyna fans have ceramic shafts.

http://www.arx-group.com/ceradyna.html

[ChaosLegionnaire]

Re: XFX GeForce 6800 Xtreme – an eBay special (with LONG story)

The thing is, this is a PCI-E video card. On PCs with PCI-E slots, I usually don't put anything older than XP. But it's good to know in case I do set up a Windows 98 system. I have a few very old games that simply run too finicky on 2000/XP. Fortunately, those games will run okay even on anemic video cards like S3 Savage 4. Talk about oldschool there .

eeeks, i totally forgot about pci-e support (or lack thereof) on win9x. i was thinking of my gf 6 agp cards and forgot that this was pci-e... i think u can forget what i said about using that card on win9x/me cuz there are no mainboard drivers or support for pci-e on win9x/me and your card will be stuck at conventional pci speeds which will bottleneck it. win9x/me only has mainboard drivers for agp at the most.

also your board manufacturer says that the board requires and supports win2k or later. so even if it succeeds in booting up the win9x/me gui at all, u will have to run it without mainboard drivers (and thus no agp acceleration either) as the drivers only support win2k or later. so it becomes irrelevant if u use an agp card on win9x/me or not.

Motherboard was AsRock 939Dual-SATAII mobo, which has both PCI-E and AGP slot (though the AGP seems to be running to the SB, so it might be running at PCI bandwidth, I don't know).

i think that may be on another dual agp/pci-e board. asrock made quite a few of those dual graphics boards for various cpu sockets. your board's specs and manual dont mention of any speed decrease or crippling on the pci-e x16 slot or agp 8x slot. im gunna assume somehow the northbridge design is one of those flexible "either-or (non-simultaneous) parallel connection" designs that can support either an agp 8x slot connected to it or a pci-e x16 slot connected to it on mainboards but not both in use at the same time. thus it can run at full 8x or x16 speed. the only way to know for sure is to run a 3d benchmark like 3dmark, aquamark etc. and see if your fps scores are similar to those with similar cards and configs.

EDIT: googled for reviews of your board. both techpowerup and anandtech talk about your board. here are the reviews of your board:

https://www.techpowerup.com/reviews/...939Dual-SATA2/
https://www.techpowerup.com/reviews/...-SATA2/14.html
http://www.anandtech.com/show/1782
http://www.anandtech.com/show/1782/9

so it appears from the benchmarks, your board's agp and pci-e x16 slot are not crippled in anyway. even tho the agp slot goes through the southbridge, it doesnt use a pci to agp bridge. thus the agp slot runs at full 8x speed without any performance penalty. my guess was right. in fact, even better. both the agp and pci-e x16 slot can run video cards simultaneously!

one of those crippled dual graphics slot designs was probably this asrock 775Dual-880Pro. i remember looking at and being interested in this board back in the day because i was interested in lga775 cpu support on win98 as well as the ddr + ddr2 support but i was quite disappointed at the crippled pci-e x16 slot running at x4 which defeats the purpose of upgrading to the more powerful and newer pci-e video cards. running at x4 prolly means it goes thru the southbridge. i also found out that win98 doesnt support pci-e. so i deemed the board a white elephant for me and scrapped the idea and decided to wait further for the core 2 architecture to come out and then do a full system overhaul.

On my older cards, I keep away from the latest drivers. All they've ever done is break more than they fixed - another part of the "planned obsolescence" model.

yes i concede that is a valid point. nvidia especially is infamous for removing driver support for older games on newer drivers. so i guess the only solution is to adopt a "dual driver" model for your system. install the newer drivers when playing recent games and then uninstall the newer drivers and (re)install the older drivers when playing older games. tho this can be a hassle as it requires a couple of reboots to uninstall and reinstall the drivers. not to mention frequent driver swapping can fock up your windows installation.

another way of avoiding this constant driver swapping debacle is to dual boot the system. one with older driver and the other with newer driver. now that only requires one reboot. the older games can also be installed and associated in the registry to the os with the older driver version and the newer games to the newer driver os. keeps things organised and neat.

[momaka]

Re: XFX GeForce 6800 Xtreme – an eBay special (with LONG story)

Thank you all for the positive comments so far .

FWIW, the R300/Radeon 9700 Pro was the first graphics card to truly make FSAA usable because of the 256-bit memory interface (the increased number of TMUs and ROPs also helped). What an incredibly forward looking card for its time - four years ahead of its time (2002), actually. As of nine years ago, it was still capable of running the then modern games at decent settings.

I know, it is crazy if you think about it. When the R300 came out, no game could make it sweat at all. Heck, not even flex a muscle. It wasn't until 2004 or so when Valve came with Half-Life 2 (and perhaps Unreal Tournament 2004 and Far Cry) that the R300/360 finally started to see some games use its power. I remember back in high school, a classmate of mine saved his lunch money for almost 2 years to buy a Radeon 9800 Pro. When Crysis came out in 2007, he was really happy because he could run it without having to upgrade his PC.

Well, they didn't output too much heat for the high end cards of the time, compared to the heat output of the "high end" cards to come thereafter, with much bigger heatsinks (IE the GeForce 8800 series).

You are right.
The R300 and R360 used about 30 and 50 Watts under idle and full load respectively. But that's still quite a bit for those heatsinks they had. Moreover, both the Radeon 9700 and 9800 used the same yellow or pink "chewing gum" stuff as the thermal compound, which is quite terrible. And to top it off, the GPU had a metal shim around it to keep the heatsink spaced evenly above the GPU die. The problem is, the heatsink never really touched the GPU die. So the thermal compound layer was quite thick. I presume all of this is what led to the demise of many of these video cards. So far, I have revived 4 out 8 that I bought. One started artifacting again, though - but this was with a Radeon 9800-style heatsink. So I can only assume the original heatsinks just don't quite cut it.
I think these do, though
https://www.badcaps.net/forum/showpo...&postcount=239

Those are very likely Sunco AX SMT series - the font, as well as the vent stamp, is a give away. "5D" is probably the date code - December of 2005. Look at the datasheet attached.

How do you find these things?! SMD electrolytic caps have always baffled me.

I guess since these are Sanyo, they get to stay, then .

ARX CeraDyna fans have ceramic shafts.

Heh, so much for their "long lasting" shaft then. Because it is broken on both of them. I swear it the shafts looked plastic, though. I have to see if one of them is still in the bag somewhere. I don't think I saw it when I took out the fans for a picture.

so it appears from the benchmarks, your board's agp and pci-e x16 slot are not crippled in anyway. even tho the agp slot goes through the southbridge, it doesnt use a pci to agp bridge. thus the agp slot runs at full 8x speed without any performance penalty. my guess was right. in fact, even better. both the agp and pci-e x16 slot can run video cards simultaneously!

Yup, I found that out when I put a faulty GeForce 7900 gs in the PCI-E slot while using a Radeon 7000 (?) to view whether the PC even detects the 7900 gs.

Only problem with the 939Dual-SATA2 is that it is a very very finicky board. Granted I found it in a thrown out PC. Various reviews online, however, mentioned the same thing. In particular, many people had their PCs freeze up randomly when using the AGP slot - both with ATI and nVidia video cards. I had the same exact problem as well. I started using mine with one of the aforementioned 9700 video cards. During games, it would lock up and PC would restart. Sometimes, the mobo would also boot-cycle several times before booting. I ruled out RAM and PSU, as well as the system caps (recapped all important capacitors on the board). The issue persisted and only got worse over time. Eventually, the SATA ports stopped working properly, followed by the sound a few months afterwards. I got the board to work now by using the IDE ports and disabling onboard SATA and sound. So I started experimenting with overclocking. And this board is good for that - 2.6 GHz on the CPU with stock voltages, no problem.

so i guess the only solution is to adopt a "dual driver" model for your system. install the newer drivers when playing recent games and then uninstall the newer drivers and (re)install the older drivers when playing older games. tho this can be a hassle as it requires a couple of reboots to uninstall and reinstall the drivers.

Yeah, that's not something I would do. As much as I love messing with hardware, I absolutely hate mucking with software. For me, the software just has to work and do its job. And I especially dislike hunting for drivers. For this reason, when I find valuable drivers, I make sure to save them and make a backup somewhere too. Some drivers, I probably have backed up on like 5 PCs - because finding them really took me some time and effort and I am not willing to do that over again.

[Wester547]

Re: XFX GeForce 6800 Xtreme – an eBay special (with LONG story)

I know, it is crazy if you think about it. When the R300 came out, no game could make it sweat at all. Heck, not even flex a muscle. It wasn't until 2004 or so when Valve came with Half-Life 2 (and perhaps Unreal Tournament 2004 and Far Cry) that the R300/360 finally started to see some games use its power. I remember back in high school, a classmate of mine saved his lunch money for almost 2 years to buy a Radeon 9800 Pro. When Crysis came out in 2007, he was really happy because he could run it without having to upgrade his PC.

Not sure about UT2004 - it was basically another revision of UT2003. Far Cry was a very flawed game, but it certainly one upped Half-Life 2 and DOOM 3 as far as beating them to the punch with respect to graphical splendor. But to really enjoy Far Cry's graphics, at least IMO, you need a card that supports Shader Model 3.0 and HDR. HDR is one of the few useful graphical advancements. It adds an incredible layer of immersing atmosphere and realism. Shader Model 2.0 cards are capable of a more rudimentary form of HDR, demonstrated by HL2's Lost Coast (and the Source Engine as of May 2010) and the rthdribl demo. Crysis was another graphical milestone. Definitely not a perfect game, but enjoyable for what it was. You really needed a GeForce 8800/9600 or Radeon HD 2900/3850/3870 card at the time in order to make it playable and so as to retain the awe-inspiring visuals.

You are right.
The R300 and R360 used about 30 and 50 Watts under idle and full load respectively. But that's still quite a bit for those heatsinks they had. Moreover, both the Radeon 9700 and 9800 used the same yellow or pink "chewing gum" stuff as the thermal compound, which is quite terrible. And to top it off, the GPU had a metal shim around it to keep the heatsink spaced evenly above the GPU die. The problem is, the heatsink never really touched the GPU die. So the thermal compound layer was quite thick. I presume all of this is what led to the demise of many of these video cards. So far, I have revived 4 out 8 that I bought. One started artifacting again, though - but this was with a Radeon 9800-style heatsink. So I can only assume the original heatsinks just don't quite cut it.
I think these do, though
https://www.badcaps.net/forum/showpo...&postcount=239

Well, I thought what killed these cards was the fact that the Infineon BGA RAM chips ran so hot that, even being leaded solder, that so many thermal cycles would eventually do them in. At least it seemed to me that under full load, the RAM chips ran hotter than the GPU die itself (too hot to keep your finger on for sure). At least that's what happened to mine - one day wild artifacts started appearing even in 2D mode. I didn't bother to "reflow" it because I don't believe in temporary solutions like that - there is no permanent solution for bad BGA/RoHS solder, on motherboards or graphics cards. The bright side is that the card came with good capacitors - 470uF 10V Rubycon ZAs, 470uF 6.3V Nichicon NA SMD polymers, and a Chemicon PXA 180uF 16V polymer, as well as decent inductors and FETs. So those components could always come in handy for another repair job. Honestly, I think the GPU die could take the heat because it's leaded solder, even if it's of the BGA package, though I suppose with enough flexing and thermal cycles that the GPU die could eventually fail too. I noticed that even with the stock thermal paste and heatsink that the GPU heatsink does get hot under full load. So if the GPU die is running much hotter, eventual failure seems to be par for the course.

Another issue with these cards is the stock ADDA hyprobearing fans that come with it. They are not lubed well from the factory so they seize prematurely and expedite the failure of these cards. And they're only spinning at 7V by default, I think, so they aren't spinning fast enough to keep the card cool (though at least they're less noisy that way).

How do you find these things?! SMD electrolytic caps have always baffled me.

I guess since these are Sanyo, they get to stay, then .

I meant Suncon, not Sunco. I really hate typos. Sanyo/Suncon make good electrolytics for the most part. Those AX SMT capacitors are not suitable for very low ESR applications. It's not easy to identify them and source a datasheet for many SMD electrolytics because so many of them look generic - they could easily be mistaken for one another.

[ChaosLegionnaire]

Re: XFX GeForce 6800 Xtreme – an eBay special (with LONG story)

The R300 and R360 used about 30 and 50 Watts under idle and full load respectively. But that's still quite a bit for those heatsinks they had. Moreover, both the Radeon 9700 and 9800 used the same yellow or pink "chewing gum" stuff as the thermal compound, which is quite terrible. And to top it off, the GPU had a metal shim around it to keep the heatsink spaced evenly above the GPU die. The problem is, the heatsink never really touched the GPU die. So the thermal compound layer was quite thick. I presume all of this is what led to the demise of many of these video cards. So far, I have revived 4 out 8 that I bought. One started artifacting again, though - but this was with a Radeon 9800-style heatsink. So I can only assume the original heatsinks just don't quite cut it.
I think these do, though
https://www.badcaps.net/forum/showpo...&postcount=239

yea, i recall back in the day, a well-known modding site, adrian's rojak pot wrote a tutorial on how to ghetto mod a p4 heatsink for the 9700/9800 pro. the 50w tdp of the r300 series made it almost as hot as a cpu. now, if the situation were reversed and u were to try that puny r300 heatsink on a p4 cpu, ppl will think u are crazy.

For now, I do not have a dedicated fan for the video card. Instead, what I do is I take a 120 mm case fan and put it at the bottom of the case to direct airflow towards the motherboard. This creates enough air movement around the video card to keep it cool. And while the Radeon 9700 doesn't have an internal temperature probe to tell me at what temperature the GPU core is running, I have tested the same heatsink and 120 mm fan setup on my eVGA GeForce 6200 that I fixed a while back. With average room temperature of about 25C, the GeForce 6200 ran about 47C idle and 55C max under load (3D games). Using my highly temperature-sensitive finger , the GeForce 6200 heatsink got only mildly warm. The Radeon 9700 appeared to run only slightly warmer than that - probably around 40C when I touched it. So I am guessing my modded Radeon 9700 probably runs around 60C max under load. (But again, that is truly a guess.) Anyways, that heatsink should be enough to keep those Radeon 9700 video cards from artifacting again. IIRC, the Radeon draws about 30 and 50 Watts under idle and load conditions, respectively. So I think that heatsink and *some* airflow from a nearby fan around should be enough to keep the card cool. Eventually, though, I might add a dedicated fan on this heatsink. Right now, I just don't have many suitable fans for that.

thats okay. because what u are doing is 100% correct for prolonging the life of those cards so it is fine to leave it as it is.

the reason? the rumour mill around the web is that ati designed the entire pcb area of the 9700/9800 to act as a heat dissipator for the gpu. so to prevent the bga solder balls from detaching, u have to keep the entire pcb surface cool. (i have read around the web that hot pcbs cause premature bga solder failure) this is done by blowing a fan at the pcb to keep it cool which is exactly what u did.

i have also proven this theory right as far as i am concerned with my own radeon 9800se softmodded to a pro with an ati silencer rev2 using shin etsu thermal compound and copper bga ramsinks (both sides) with an 80mm thermaltake smart case fan II screamer blowing at the card in a micro-atx case overclocked to 432/378 which lasted me 7 years of 12x7x365 usage. quite impressive life i got out of my card before it failed i must say.

assuming momaka doesnt abuse his cards like i do, i think your stock of seven 9700 cards should last u at least half a century. now im jealous lol... wish i could have more 9700/9800 cards. cause im down to my last 9800 pro card...

Yeah, that's not something I would do. As much as I love messing with hardware, I absolutely hate mucking with software. For me, the software just has to work and do its job. And I especially dislike hunting for drivers. For this reason, when I find valuable drivers, I make sure to save them and make a backup somewhere too. Some drivers, I probably have backed up on like 5 PCs - because finding them really took me some time and effort and I am not willing to do that over again.

yea, i use the third party omegadrivers and they just work great for me. never experienced any of the problems ppl were complaining about with ati catalyst drivers. i just think those ppl simply had crappy boards, psus or just had a w3rd combination of hardware.

Well, I thought what killed these cards was the fact that the Infineon BGA RAM chips ran so hot that, even being leaded solder, that so many thermal cycles would eventually do them in. At least it seemed to me that under full load, the RAM chips ran hotter than the GPU die itself (too hot to keep your finger on for sure). At least that's what happened to mine - one day wild artifacts started appearing even in 2D mode. I didn't bother to "reflow" it because I don't believe in temporary solutions like that - there is no permanent solution for bad BGA/RoHS solder, on motherboards or graphics cards. The bright side is that the card came with good capacitors - 470uF 10V Rubycon ZAs, 470uF 6.3V Nichicon NA SMD polymers, and a Chemicon PXA 180uF 16V polymer, as well as decent inductors and FETs. So those components could always come in handy for another repair job. Honestly, I think the GPU die could take the heat because it's leaded solder, even if it's of the BGA package, though I suppose with enough flexing and thermal cycles that the GPU die could eventually fail too. I noticed that even with the stock thermal paste and heatsink that the GPU heatsink does get hot under full load. So if the GPU die is running much hotter, eventual failure seems to be par for the course.

yepp that is right. rumour mill has it again that infineon ram doesnt seem to like heat and is poor for these type of high heat applications with a hot pcb. so no surprises it fails! thats why i only have samsung and hynix ram for the 9800 cards i have.

and it also appears that momaka has infineon vram for at least one of his 9700 cards! i strongly advise momaka to ghetto bga ram sinks as well for his 9700 cards that have infineon vram... there, more modding work for u to do, sir! this modding is getting addictive, isnt it?

[momaka]

Re: XFX GeForce 6800 Xtreme – an eBay special (with LONG story)

I guess we have officially turned this thread into a Radeon 9x00 discussion. Oh well .

Crysis was another graphical milestone. Definitely not a perfect game, but enjoyable for what it was. You really needed a GeForce 8800/9600 or Radeon HD 2900/3850/3870 card at the time in order to make it playable and so as to retain the awe-inspiring visuals.

Yes, the original Crysis looks absolutely amazing, even to this day. But that's about all of the positives. I tried playing it at a friend's house (he had this Alienware PC with a 8800 GTX and C2Q - a dream PC for 2007 ), but the gameplay just seemed too repetitive. And the game just seemed to lack depth and story.

Well, I thought what killed these cards was the fact that the Infineon BGA RAM chips ran so hot that, even being leaded solder, that so many thermal cycles would eventually do them in. At least it seemed to me that under full load, the RAM chips ran hotter than the GPU die itself (too hot to keep your finger on for sure).

You know, I used to think the same thing, but I am not so sure anymore. In particular, the last time I tested one of my Radeon 9700 card, the RAM seemed to run at quite a normal temperature. I'll definitely have to test that again. Perhaps, the RAM on that card has not "degraded" quite yet. Usually, as the RAM ages, the capacitor memory cells in it become more leaky, making the RAM run hotter. Eventually a point is reached where the cells loose charge too fast and so they don't retain the data too well. That's when you can start getting artifacts. The solution is to increase the voltage. However, all that will do is make the RAM even more hotter. So it's a downward spiral process until the RAM dies.

Now, if my memory serves me right (oh wait, is that a pun?! kekeke ), I think I remember reading that the Radeon 9500/Pro and 9700 (non-Pro) all use slower Infineon RAM, whereas the 9700 Pro uses Samsung RAM. I'm not sure about the 9800. But either way, I think the Infineon RAM did indeed have some problems back then, and actually all the way up until the Xbox 360. In fact, some of the RROD problems are blamed on bad Infineon (now Qimonda) RAM. I've seen this at the repair shop where I used to work at. I replaced the GPU two times on one Xbox 360, and it was still finicky. Then I noticed that the RAM ran extremely hot. Unfortunately, I don't remember what brand it was, but it could have well been Qimonda.

I didn't bother to "reflow" it because I don't believe in temporary solutions like that - there is no permanent solution for bad BGA/RoHS solder, on motherboards or graphics cards.

Well, the Radeon 9700 and 9800 video cards have leaded solder, so it is worth a try. A proper reflow (with flux) can actually last a very long time - sometimes even as much as the card lasted from the factory. The real problem here is heat and that these cards are just not cooled well with that puny heatsink.

I noticed that even with the stock thermal paste and heatsink that the GPU heatsink does get hot under full load. So if the GPU die is running much hotter, eventual failure seems to be par for the course.

The die always runs hotter than the heatsink feels. That's a given.
The stock thermal compound is way too thick on these cards. Again, that's due to the metal shim that the GPU has. I guess ATI thought it would be a good idea to protect the GPU heatsink from cracking the die by putting on those metal shims. The truth is, those metal shims raise the heatsink too much. So the thermal compound layer ends up being a lot thicker than it needs to be.

And also, the stock thermal compound is absolute junk when it dries. I had to chisel it away from the die with a small screwdriver on some of these 9700's - that's how tough it had gotten. Moreover, the stuff that I couldn't remove with a screwdriver (due to being to close to a die corner or GPU PCB) I had to remove with Acetone. Yes, Acetone! Because even 99.9% IPA wouldn't do anything to it.

Another issue with these cards is the stock ADDA hyprobearing fans that come with it.

Yup, the fans on these video cards become seized quite often. I don't know if they are all ADDA, I didn't take note (I did fix one on my classmate's 9800 Pro - same guy that saved his lunch money for a year.)

And they're only spinning at 7V by default, I think, so they aren't spinning fast enough to keep the card cool (though at least they're less noisy that way).

Well, Radeon 9500/9700 cards have 12V going straight to their fan connector. IDK about the 9800. Probably the same, though.

I really hate typos.

They are inevitable in my posts . Sorry!

It's not easy to identify them and source a datasheet for many SMD electrolytics because so many of them look generic - they could easily be mistaken for one another.

Which is why I almost never bother either. At least not for the very generic-looking ones.

yea, i recall back in the day, a well-known modding site, adrian's rojak pot wrote a tutorial on how to ghetto mod a p4 heatsink for the 9700/9800 pro. the 50w tdp of the r300 series made it almost as hot as a cpu. now, if the situation were reversed and u were to try that puny r300 heatsink on a p4 cpu, ppl will think u are crazy.

I know, right. What's up with that? Why can we get huge oversized heatsink for the CPU but not the GPU? People panic when their CPU runs near or above 60C. Yet, for GPUs nowadays, an operating temperature of 70-80C is considered quite acceptable.
Just never mind the BGA and that lead-free solder, it's all dandy.

assuming momaka doesnt abuse his cards like i do, i think your stock of seven 9700 cards should last u at least half a century.

Well, I *only* have three working at the moment. And I also have three of them that I likely won't be able to repair: one had a dead RAM MOSFET and fails to detect in a PC, another freezes the PC at POST, and the last one I used for parts without analyzing it much, since I got these back in the day way before I knew anything about reflowing or even electronics for that matter. So *at best*, I think I might be able to get five running.

I haven't overclocked any of them. Though, if I do get another one running, I might try putting a bigger heatsink on it and overclocking the GPU core only (probably up to 9700 Pro speed, but no more). If anything, I never overclock RAM, especially if it's running at the speed it was designed for. Oh, and especially not Infineon - they like to die even when you don't abuse them, so I can only imagine what happens if you do. I also don't like overvolting RAM.

thats why i only have samsung and hynix ram for the 9800 cards i have.

Yes, Samsung and Hynix are the best IMO as well. And Micron, too. But I don't think they make video memory. Even Elpida is okay. Infineon really is the only one I had poor luck with - especially for desktop RAM.

They should have changed their name to Inferior RAM instead of Qimonda .

this modding is getting addictive, isnt it?

Yes , because it allows me to explore my creativity. I think that is something really important to keep your life interesting and not get bored doing the same stuff over and over.

[goodpsusearch]

Re: XFX GeForce 6800 Xtreme – an eBay special (with LONG story)

I hate myself for letting my ALL IN WONDER 9800 SE (modded to 9800 PRO) die due to lack of experience about electronics...

It started showing some vertical lines with artifacts, but if I touched it a bit it would stop. So it was definitely a broken solder issue. But I knew nothing about all that back then so I kept using it like that, with the stock cooler and the RAM chips bare and burning hot and 2 months later the vertical lines got permament

Still, I own the card, but it sits in its box waiting for repair

[Wester547]

Re: XFX GeForce 6800 Xtreme – an eBay special (with LONG story)

I guess we have officially turned this thread into a Radeon 9x00 discussion. Oh well .

Sorry, didn't mean to derail the thread.

Yes, the original Crysis looks absolutely amazing, even to this day. But that's about all of the positives. I tried playing it at a friend's house (he had this Alienware PC with a 8800 GTX and C2Q - a dream PC for 2007 ), but the gameplay just seemed too repetitive. And the game just seemed to lack depth and story.

Well, I also enjoyed the soundtrack that Inon Zur composed for the game. In fact, I enjoyed it enough to buy the special edition just to own that score. Otherwise, yes, the game was rather passable. Not quite as cringe-inducing as Far Cry, though. The dialogue in that game was so poorly written that it was laughable.

You know, I used to think the same thing, but I am not so sure anymore. In particular, the last time I tested one of my Radeon 9700 card, the RAM seemed to run at quite a normal temperature. I'll definitely have to test that again. Perhaps, the RAM on that card has not "degraded" quite yet. Usually, as the RAM ages, the capacitor memory cells in it become more leaky, making the RAM run hotter. Eventually a point is reached where the cells loose charge too fast and so they don't retain the data too well. That's when you can start getting artifacts. The solution is to increase the voltage. However, all that will do is make the RAM even more hotter. So it's a downward spiral process until the RAM dies.

Well, one of my 9500 Pros has had really hot running RAM from the factory. But again, the RAM it has is....

Now, if my memory serves me right (oh wait, is that a pun?! kekeke ), I think I remember reading that the Radeon 9500/Pro and 9700 (non-Pro) all use slower Infineon RAM, whereas the 9700 Pro uses Samsung RAM. I'm not sure about the 9800. But either way, I think the Infineon RAM did indeed have some problems back then, and actually all the way up until the Xbox 360. In fact, some of the RROD problems are blamed on bad Infineon (now Qimonda) RAM. I've seen this at the repair shop where I used to work at. I replaced the GPU two times on one Xbox 360, and it was still finicky. Then I noticed that the RAM ran extremely hot. Unfortunately, I don't remember what brand it was, but it could have well been Qimonda.

...Infineon. The inferior Infineon RAM probably runs hotter than the Samsung and Hynix/Hyundai RAM. Also, the RAM is cool as a cucumber to touch in 2D mode - only 3D mode make it run very hot.

Yup, the fans on these video cards become seized quite often. I don't know if they are all ADDA, I didn't take note (I did fix one on my classmate's 9800 Pro - same guy that saved his lunch money for a year.)

The OEM ATI ones come with ADDA hyprobearing fans. AD0512MX-G70 is the model number (rated for 4300RPM, 50x50x10mm). Some of them come in 7 blade variants, others 9 and 11 blade variants. Also, I meant came with them, not came with it.

Well, Radeon 9500/9700 cards have 12V going straight to their fan connector. IDK about the 9800. Probably the same, though.

It doesn't seem that way to me (I thought I read somewhere that +7V was the default voltage). I unplugged the fan from the header and connected it straight to the +12V output of a Hipro PSU (so it had good regulation for the +12V rail and no bad caps in that one), and it was much louder and was moving more air than it was just connected to the header. Maybe they limit the current going to the fan in order to slow it down somehow whilst still providing it with +12V.

They are inevitable in my posts . Sorry!

I should have been clearer. I hate it when I make typos, no disrespect to others - once the time to edit expires it makes an informative post look bad IMO.

They should have changed their name to Inferior RAM instead of Qimonda .

Well, Infineon/Qimonda RAM isn't quite as bad as Elixir or Corsair RAM. Or GSkill for that matter. But I guess that would be stating the obvious.

[ChaosLegionnaire]

Re: XFX GeForce 6800 Xtreme – an eBay special (with LONG story)

Only problem with the 939Dual-SATA2 is that it is a very very finicky board. Granted I found it in a thrown out PC. Various reviews online, however, mentioned the same thing. In particular, many people had their PCs freeze up randomly when using the AGP slot - both with ATI and nVidia video cards. I had the same exact problem as well. I started using mine with one of the aforementioned 9700 video cards. During games, it would lock up and PC would restart. Sometimes, the mobo would also boot-cycle several times before booting. I ruled out RAM and PSU, as well as the system caps (recapped all important capacitors on the board). The issue persisted and only got worse over time. Eventually, the SATA ports stopped working properly, followed by the sound a few months afterwards. I got the board to work now by using the IDE ports and disabling onboard SATA and sound. So I started experimenting with overclocking. And this board is good for that - 2.6 GHz on the CPU with stock voltages, no problem.

again, that sounds like a power delivery issue with the mosfets on the board to me. not the first time im hearing this on mainboards. the 9700/9800 and gf 5 series of cards at the time are known for being very harsh at sucking power from the mobo's 3.3v rail and sometimes the 5v rail on the board or molex connector.

on one of my country's computer hardware forums, hardwarezone, one guy complains about his 9700 or 9800 card always killing every mobo he goes through within a few months. if he doesnt use the 9700/9800 card, the mobo keeps running fine...

this has led me to strongly believe that either poor mosfet power delivery design to the agp slot (agp rev3 specs say design must support a max of 40w through the agp slot) and/or badcaps on the mobo cause problems with those agp cards that are power hungry.

if the 3.3v/5v mobo mosfets fail or arent up to the task, it likely will result in inadequate power to the ram, chipset or onboard peripherals since those require power from the minor rails. typically, i've heard of either the onboard sound or the nic dying first.

And also, the stock thermal compound is absolute junk when it dries. I had to chisel it away from the die with a small screwdriver on some of these 9700's - that's how tough it had gotten. Moreover, the stuff that I couldn't remove with a screwdriver (due to being to close to a die corner or GPU PCB) I had to remove with Acetone. Yes, Acetone! Because even 99.9% IPA wouldn't do anything to it.

have u tried using arcticlean two step tim remover? there's no tim it cant remove that i've heard of. i've also had no problem myself removing the stock chewing gum tim on my 9800 cards using arcticlean. its quite cheap in the usa too. performance pcs is selling it for around 6 bucks or sumthing and i just checked, its sold out atm at ppc...

over here, arcticlean is being sold at twice its cost in the usa. lucky u... better get it as soon as they restock it!

I know, right. What's up with that? Why can we get huge oversized heatsink for the CPU but not the GPU? People panic when their CPU runs near or above 60C. Yet, for GPUs nowadays, an operating temperature of 70-80C is considered quite acceptable.
Just never mind the BGA and that lead-free solder, it's all dandy.

yes, these days i've seen high-end video cards running at temps exceeding 90°C and everyone including the manufacturer/distributor insists its fine!

tsk tsk the throwaway culture these days... they just only care till it lasts until the next upgrade cycle...

You know, I used to think the same thing, but I am not so sure anymore. In particular, the last time I tested one of my Radeon 9700 card, the RAM seemed to run at quite a normal temperature. I'll definitely have to test that again. Perhaps, the RAM on that card has not "degraded" quite yet. Usually, as the RAM ages, the capacitor memory cells in it become more leaky, making the RAM run hotter. Eventually a point is reached where the cells loose charge too fast and so they don't retain the data too well. That's when you can start getting artifacts. The solution is to increase the voltage. However, all that will do is make the RAM even more hotter. So it's a downward spiral process until the RAM dies.

errr, the correct thing to do in that case would be to see if u can tinker with the advanced ram timings via overclocking utilities. the timing u are looking for would be the refresh rate, refresh interval or tREF. try to increase the refresh rate/tREF or decrease the refresh interval, depending on how u wanna look at it, until the artifacts disappear.

but im guessing u hate messing with software like that just to fix such a problem... keke...

Yes, Samsung and Hynix are the best IMO

well for me having had both types of vram, i'd say mhz/ghz wise, hynix is the best followed by samsung a close second.

the hynix ram i've had on my 9800s can hit a max overclock of 378mhz(756mhz ddr) with copper ramsinks. compared to the samsung (also with the same copper ramsinks) which hits only a slightly lower 371mhz (742mhz ddr).

i also have hynix AFR-T0C gddr5 ram (nominal speed 2ghz sdr or 4ghz ddr) on my 1gb gtx 460 and it overclocks to 2200mhz (4400mhz ddr) naked without any ramsinks plus artifact-free in occt gpu test. a 10% error-free overclock on such hot running gddr5 vram without using ramsinks is quite splendid i must say. hynix 4tw!

[goodpsusearch]

Re: XFX GeForce 6800 Xtreme – an eBay special (with LONG story)

hynix = Hyundai?

[momaka]

Re: XFX GeForce 6800 Xtreme – an eBay special (with LONG story)

Okay, I got a bit of news...
The ICs in those burned fans from the XFX 6800 above and 7600 gs might actually be good. Even the protection diode on the 6800 fan board is still good. The stator windings on the 6800 also might have been good (or not completely bad yet, at least). However, as I was trying to do some "fixing up" on the stator for the 6800 fan, I accidentally tore the stator core (with the windings) out. One of the wires snapped at a bad place, so I can't solder it back in without a complete rewind of the stator. And I can't use the stator from the 7600 gs fan, because it has a different windings pattern *and* the windings are burned.

So most likely, I will get rid of the windings on the 7600 gs fan and rewind the stator, then test it on the 6800 fan board. With some work and ghetto modding, I might actually be able to revive one of these fans.

Worth the time? Heck no! But I do tend to have some free time on my hands now while applying for jobs. (Which is why I am on BCN so much). Besides, why not? Maybe i learn something along the way .

I hate myself for letting my ALL IN WONDER 9800 SE (modded to 9800 PRO) die due to lack of experience about electronics...
...
Still, I own the card, but it sits in its box waiting for repair

Remove the through-hole electrolytic caps, and give it a reflow - both the RAM and the GPU. Make sure to use flux too. Probably not a good idea to bake in the oven since there are too many SMD electrolytic caps on the board that could pop. I've had good success on these with heat gun / hot air.

Sorry, didn't mean to derail the thread.

Not on you really, so no worries. Actually, I'm glad we can have some off-topic discussion .

The dialogue in that game was so poorly written that it was laughable.

And that's what really threw me out of it. Kind of like Batman's "evil" voice .

Also, the RAM is cool as a cucumber to touch in 2D mode - only 3D mode make it run very hot.

Hmm. Maybe that's why I thought mine ran cool. I'll have to re-check.

It doesn't seem that way to me (I thought I read somewhere that +7V was the default voltage).

Nope. Fan is connected straight to 12V rail on AGP connector.

I double-checked yesterday before making my post, since your post raised some doubts about my memory (which has started to deteriorate now, much like Inferior RAM).

I should have been clearer. I hate it when I make typos, no disrespect to others - once the time to edit expires it makes an informative post look bad IMO.

Yes, I know what you meant . I'm just saying that I don't sweat it anymore, because I can never catch all of my mistakes.

Well, Infineon/Qimonda RAM isn't quite as bad as Elixir or Corsair RAM. Or GSkill for that matter.

Now, if you want to go down the chute, you might as well pick Buffalo. And no fried wings included with that order!

again, that sounds like a power delivery issue with the mosfets on the board to me. not the first time im hearing this on mainboards.

Nope.

No board with AGP that I know of has MOSFETs between AGP power pins and power supply. PSU power goes straight to AGP connector. However, there may be filter caps on the board to reduce EMI from the video card. And I suspect this is the issue with my 939Dual-SATA2 - either inadequate filtering or bad routing of the traces. The former may be fixable by adding caps if the traces for the power lines going to the AGP are on an outer layer of the board. I haven't checked.

on one of my country's computer hardware forums, hardwarezone, one guy complains about his 9700 or 9800 card always killing every mobo he goes through within a few months. if he doesnt use the 9700/9800 card, the mobo keeps running fine...

I highly doubt the card is killing the PC. Board and/or power supply might have issues. Without pictures, stories like that should be taken with a grain of salt at the very least. No offense to that guy.

if the 3.3v/5v mobo mosfets fail or arent up to the task, it likely will result in inadequate power to the ram, chipset or onboard peripherals since those require power from the minor rails. typically, i've heard of either the onboard sound or the nic dying first.

No, that's not right.

If the RAM or NB/SB don't get adequate power, board won't POST or will crash. Simple as that.

I suspect the audio and SATA ports on mine died because of BGA issues with the SB. Both the NB and SB run extremely hot on the 939Dual-SATA2. This is actually the reason I now have the 120 mm fan there. Originally, I just ran the board as is and with my Radeon 9700 with its replica 9800 heatsink. It wasn't until the board severely started crashing that I noticed how hot the NB and SB ran. But by that time, it was too late. I probably have to reflow it someday.

have u tried using arcticlean two step tim remover?

Yes, and that stuff is good
However, a large container of Acetone is about the same price here. Plus, Acetone is useful for other stuff too. We always have some in the garage, so I never find a need for the ArctiClean stuff myself. Only used it at a repair shop I worked in a long time ago.

tsk tsk the throwaway culture these days... they just only care till it lasts until the next upgrade cycle...

Yup. Unfortunately. That's why I like to save older hardware - most of it is just built better (sans crappy electrolytic caps).

but im guessing u hate messing with software like that just to fix such a problem... keke...

I'd rather unsolder the RAM and replace it than play with it like that .
Unfortunately, I don't have any compatible RAM on my junk boards that I can use for the Radeon 9700/9800. It's a different package from what the Xbox 360 and GeForce 6 and 7 series use. On that note, Xbox 360 RAM is compatible with some of the GeForce 6 and 7 series cards (and in some cases, like my still dead 7600 GT, the Xbox 360 RAM is a direct replacement - already tried it ).