Gigabyte EP43-UD3L stuck in POST

[eccerr0r]

Re: Gigabyte EP43-UD3L stuck in POST

Now do I risk buying a new CPU for this board instead of using this 3.4GHz PresHott as it's all I have spare.
Really don't want to invest anymore into S775 if I can help it... uugghh...

[zunasthegreat]

Re: Gigabyte EP43-UD3L stuck in POST

Τhats a fine board there man, I'm typing right now from a EP45-DS3R. Just get a ASM1061 SATA card for cheap if onboard SATA are busted. Some cheap modded Xeon will also do wonders.

[momaka]

Re: Gigabyte EP43-UD3L stuck in POST

Now do I risk buying a new CPU for this board instead of using this 3.4GHz PresHott as it's all I have spare.

Depends what you want the system to do.
If the CPU usage is expected to be moderate to high, you might want to grab a Core 2 Duo for next to nothing on eBay. There are a lot of choices for under $4 shipped to your door. You'll make that back in electricity savings in 2-3 months or less (depending on how you run the machine - i.e. 24/7 or just intermittently.) But if that system will be idle most of the time, there won't be that much of a difference between the Prescott and a C2D, unless you go with some really low-power C2D, like E5200 (which is a rather "slow" chip on a 45 nm process... and I put quotes around slow, because it will blow the P4 away still.)

Here's a 2.9 GHz E7500 for $3.90 shipped!
https://www.ebay.com/itm/Intel-Core-.../264870786372?
This will runs circles around the P4 all day when it comes to modern browser use.

[eccerr0r]

Re: Gigabyte EP43-UD3L stuck in POST

I fleaed a cheap Core2Quad, planning on sticking it into my PVR, and then swiping that machine's Core2Duo for this scratch machine in case the board is still no good. The C2Q should match the PVR's 1333 FSB M/B so I'll get both a FSB and core boost on that machine, as it's a PVR it's up and thus can help out on the distcc load. (For some reason my computers love asking my PVR for distcc help when there are faster ones available...)

Hoping this C2Q stays cooler (when idle) than the apparently preproduction C2D I have.

Oh and this test machine will be pretty much in S5 most of the time, until the production machines (like the PVR) croak... it'll be the swapout instead of a P3 or something ...

BTW, back to the Gigabyte... Strange I had two Gigabyte S775 boards (acquired new) and both have a problem when sometimes when you reboot, the Ethernet (yay RTL8169) doesn't enumerate and thus doesn't show up to the OS at all. Annoying, wonder if this has something to do with its flakyness...

[eccerr0r]

Re: Gigabyte EP43-UD3L stuck in POST

Damn, started hanging at the same spot again with the Core2, though seems to be prior to ram check. Ugh... worked better with the pres-hott...

[eccerr0r]

Re: Gigabyte EP43-UD3L stuck in POST

Ahhh WTF. The machine gets all flaky when I put in PC2-3200 or PC2-4200 if I don't have the PresHot installed. PC2-5300 seems OK and I don't have any PC2-6400s to try...

Mysteries continue...

[eccerr0r]

Re: Gigabyte EP43-UD3L stuck in POST

As for the artifacting HD3650... that will come back for sure, IME. Give it a few heat cycles or put under 3D load.

Oops, no wonder why the "HD3650" is "working" ... it's mistaken identity, the working board is an older X800XL ... which is probably about as good as the HD3650.

I swapped in a FX5200 to see what it does... yeah this thing is indeed slower but sucks up 10 less watts when idle and about the same power as the X800XL idling... when under load.

The X800XL seems to be able to eat 50W from idle to high power...

And I couldn't get the real HD3650 to artifact again either... it's about the same power as the X800XL idle but only increases 25W at full load instead of 50W... and they both do about the same in sproingies and pixelcity, and at least allows webgl to work, unlike the much older and less capable fx5200 that disables webgl...

[eccerr0r]

Re: Gigabyte EP43-UD3L stuck in POST

Oh...no... found a weird new observation: I found that if I had a USB2 HDD plugged in or plugged in at boot, it would freeze.

However a USB keyboard is fine...

What gives...

[momaka]

Re: Gigabyte EP43-UD3L stuck in POST

Ahhh WTF. The machine gets all flaky when I put in PC2-3200 or PC2-4200 if I don't have the PresHot installed. PC2-5300 seems OK and I don't have any PC2-6400s to try...

Mysteries continue...

Is this board powered from the Antec PSU you just posted about recently with bad 5VSB? If so, maybe the PSU just has more bad caps. Also, being an Intel board... any chance you are running a stock heatsink with push pins? I've seen those kind of intermittent problems due to this.

Oh...no... found a weird new observation: I found that if I had a USB2 HDD plugged in or plugged in at boot, it would freeze.

However a USB keyboard is fine...

What gives...

Bad caps in the PSU x2??? Check 5V and 5VSB rails for sure.

Oops, no wonder why the "HD3650" is "working" ... it's mistaken identity, the working board is an older X800XL ... which is probably about as good as the HD3650.

Ah, that would explain it then.
Yes, on average the X800XL will be about the same as the HD3650, though it depends on the 3D program you're using. For example, the X800XL should crush the HD3650 in older (DX 9.0b or earlier) games that rely more heavily on "raw" power (i.e. higher memory bandwidth, and more ROPs and TMUs.) In contrast, the HD3650 might do better in newer games (relatively speaking) based on DX 9.0c. But all in all, the X800XL is a higher-tier card, despite being 3 generations older.

I swapped in a FX5200 to see what it does... yeah this thing is indeed slower but sucks up 10 less watts when idle and about the same power as the X800XL idling... when under load.

Sounds about right.
The FX5200 is like 10-15 Watts TDP max. The slower-clocked ones will happily run with a tiny passive cooler and never get too hot.
By the way, is that FX5200 with a PCI-E connector? If so, those are actually kind of rare to see now - mostly because by the time PCI-E became popular, no one cared to get such a low-end card for PCI-E. The GF 6200 and especially 7300 were much better entry-level cards at the time, despite still being quite sucky overall.

The X800XL seems to be able to eat 50W from idle to high power...

Sounds about right x2.
According to some sites that tested its power draw on the motherboard (via current shunts), it's a 64W TDP card that idles around 25-ish Watts. So when you factor PSU inefficiencies, ~50W difference is about where it should be.

And I couldn't get the real HD3650 to artifact again either... it's about the same power as the X800XL idle but only increases 25W at full load instead of 50W...

Yup. HD3650 is rated around 40-50W, depending on core clocks (some manufacturers wanted to get away with cheaper coolers, so they downclocked the card, and thus lowered the TDP that way too.)

unlike the much older and less capable fx5200 that disables webgl...

Well, the FX5200 is really ancient stuff.
It makes for a good basic 2D adapter for office work and that's it. It's too weak for HD video acceleration (Mpeg2/DVD is about as good as it will do.) And in terms of 3D... despite nVidia listing this card as being DX9-compatible, the FX5200 GPU doesn't actually support all DX9 features and is really just a DX8.1 card with a few extra features from DX9 that allow it to be just adeqaute to run Aero interface on Vista/7.

I still like the FX5200, though - the AGP cards actually do quite well for a Windows 98 retro build and older games.

[eccerr0r]

Re: Gigabyte EP43-UD3L stuck in POST

Hmm... don't know, this is another Antec, actually it's probably the first Antec that I ever posted about, a 300W, though I've seen the same behavior on other PSUs.

Another issue is that I think I saw that PC3200 and PC4200 aren't technically "supported" on this board but not clear, perhaps due to RAM/FSB multipliers. Dunno. Anyway with the faster PC5300 (now dual channeled with a PC6400 at PC5300 speeds) the machine is actually rock solid, at least I was able to run a compile job overnight, and it even ran out of ram, stressing the disk subsystem as it thrash swapped - again fully stable. (the HS/F is the Zalman, which would require board removal to remove the whole assembly.)

Actually the FX5200 is a PCI, not even sure they have a PCIe version. Also noticed I have a FX5200 AGP though it has no chance of working in this MB.

Really should offload some of these old GPUs... too much garbage lying around. Radeon 9250 (PCI) to be completely trounced by the X800XL and HD3650 ...

[momaka]

Re: Gigabyte EP43-UD3L stuck in POST

Really should offload some of these old GPUs... too much garbage lying around. Radeon 9250 (PCI) to be completely trounced by the X800XL and HD3650 ...

Well, don't chuck them in the trash. Some of these old GPUs' value will increase in time as they get more rare, even if they are low end. Also, low-end for current gen could be top-of-the-line for few (or more than a few) generations back. The FX5200 and Radeon 9200/9250 are a good example of this. Although they are absolutely dog-slow even for XP -era games, they run quite well with anything DOS all the way up to late Windows 9x games. Heck, on that note, the Intel "Extreme" Graphics 2 (i865 IGP) on my P4 Dell Optiplex was laughably slow, even for when it came out (mid 2000's). But here is the funny part: a lot of late 90's games (in my case, mainly Need For Speed High Stakes and Half-Life, among others) run at nearly 60 FPS constant at 1024x768 in 32-bit color with that IGP. This is comparable (if not better) performance to Glide mode on older 3DFX Voo-Doo cards, which have been selling for crazy-high prices in the last 5-10 years already.

So my suggestion is to either keep those old GPUs... or if not, sell them or give them to someone that can use them. IMO, though, it's always a good idea to keep a spare PCI GPU or two in the parts drawer. Can be useful for flashing BIOS-bricked video cards (if IGP is not available) as well as troubleshooting other problems.

[eccerr0r]

Re: Gigabyte EP43-UD3L stuck in POST

Yeah, might have to fleabay them or something.
Think currently I have:
R9250 PCI
R9250 AGP
Rage Pro Turbo AGP
Rage Pro Turbo PCI
FX5200 AGP
FX5200 PCI
FX6200 AGP (maybe dead)
MX420 AGP (maybe dead)
MX440 AGP
R8500AIW AGP
Trio64 PCI
P650 PCI

Testing them is another problem, especially the P650 which has no Linux drivers, I really want to get rid of that card. But yeah I have too much junk and not enough $ to buy a Ryzen, so would be best to trade off...

(Oh, btw, I have a voodoo3 pci somewhere I think too... heard these were like made of gold now or something?)

[ChaosLegionnaire]

Re: Gigabyte EP43-UD3L stuck in POST

yeah i could use a radeon 9250 pci but only if its the 128 bit version with 4 mem chips on each side of the pcb and it has dvi. most built by ati radeon 9250 cards i've seen only come with vga and tv video out and no dvi.

[momaka]

Re: Gigabyte EP43-UD3L stuck in POST

Just had a thought regarding your USB ports... do the crashes happen only with the front USB ports on your machine or is it with all USB ports?
If it's only the front USB, what kind of case are you using?
Some cheap cases have poor shielding on the USB cables (or none at all on the really garbage ones), which can make USB detection and operation very sporadic.

(Oh, btw, I have a voodoo3 pci somewhere I think too... heard these were like made of gold now or something?)

Yup, pretty much.

Most Voo-Doo 3 stuff easily sells for $100+. Depends on demand and version of the card of course. But very rarely I've seen them dip below $100. Just find out what your card is exactly on eBay and watch some auctions end for the same card. That should give you a good idea how to price it.

Yeah, might have to fleabay them or something.
Think currently I have:
R9250 PCI
R9250 AGP
Rage Pro Turbo AGP
Rage Pro Turbo PCI
FX5200 AGP
FX5200 PCI
FX6200 AGP (maybe dead)
MX420 AGP (maybe dead)
MX440 AGP
R8500AIW AGP
Trio64 PCI
P650 PCI

That Radeon 8500 AIW is a nice card and could get you $15-20 or more depending on demand and if you have any of the original accessories with it. I've actually seen them sell for close to $30 sometimes. But since demand for them is "sporadic" at best, you really have to see when someone lists one on eBay and lost of "bids" start showing up. When that happens, you put up your card for sale while there are still people interested in buying one.

As for the other cards... well the 9250 PCI and FX5200 PCI are the ones that sell fairly regularly, though not for that much - usually $10-15. As for the other lower-end cards... the market is fairly saturated with those (especially the GF MX 400 series and ATI Rage anything)... so they may not sell anytime soon and not for much either. Better off keeping them as test/parts cards than anything else.

Oh and lastly, the GeForce 6200 AGP - see what kind of GPU it has. If it's a 6200LE with an NV43 (same GPU as 6600, but down-clocked and possibly cut/defective shaders and ROPs), it may be a lot more worthwhile. You can usually tell the NV43's from the latter low-cost native AGP 6200 cards, as the NV43's are native PCI-E and will have a bridge chip (second heatsink under the GPU one) as well as an AGP 3.0 -only connector (single notch cutout on the back only.) The low-cost 6200 cards have a small GPU chip with usually passive cooling and "universal" AGP connector. The NV43 is too hot for passive, so they come with a fan in most cases.

But yeah I have too much junk and not enough $ to buy a Ryzen, so would be best to trade off...

I wouldn't buy a new PC right now. Heck, even the used PC market is pretty hectic, thanks to both Covid and silicon chip shortages.

If you're trying to buy a lot of firepower for little money, a Dell Precision T3500 workstation might do the trick for ya. High-end 4-core (8 thread) and mid-tier 6-core (12 thread) 1366 CPUs are still pretty cheap on the used market. A 6-core in a T3500 with a good overclock can get you big bang for your buck (and I don't mean that in a Powmax -style way. ) I got a used T7500 (dual socket 1366) station like that last year, and it's a monster. I got really lucky with the pricing, though, because it was auction-style listing, but freight shipping (required for properly shipping one of these big boys) was billed separately and the seller happened to be not too far away from where I was. So I asked if they do local pickup, which they were more than happy to do for me, since that would save them time and labor on packing the thing. This ultimately gave me a lot more "bid power" over anyone else, as I could take the price over $100 easily and still be cheaper than a BIN listing going for $200+ with "Free" shipping... but I didn't need to jack up the price. For some reason, the description was very limited and under-represented what the station was. The pictures were accurate, though. When I saw the 2nd socket riser was included in one of the pictures, I immediately knew it was a good deal.

Anyways, long story short... if you get one of these older Xeon 1366 workstations, try to find one near you or maybe even drive out to pick it up and save on shipping, if the seller lets you. Moreover, these workstations take cheaper ECC/registered DDR3 and are usually pretty damn stable machines... hence why they use them as *workstations*. You can do all your compiling all night and day and never worry that it will crash midway. Those 1366 CPUs can be a little power-hungry, though, at high clock speeds (be it stock or over-clocked.)

[eccerr0r]

Re: Gigabyte EP43-UD3L stuck in POST

Alas no, the back panel ports were also crashing the machine. I don't think I have any front panel on this case...

Well, the R8500AIW was my main GPU for a long while -- it was fast at the time, bought it specifically for Linux compatibility (at least the 8500, not the "AIW" portion). I still have the original box and even the pod for it. Probably should let it go, not sure if I'll ever use it again.

The MX420 was my previous GPU prior to the R8500AIW. I just tested it -- it seems to be dead. I remember it artifacting badly prior to replacing it, though the 3650 seems to still work after its bout with artifacting...

I'm pretty sure the FX6200 is a cheap version, it had blown caps and a passive heatsink, though it appears to at least be working for some applications - though artifacting on WebGL apps.

I need to locate the voodoo3. I think I gave it to my parents, I'm sure they still have it somewhere.

I don't think I have enough money to buy a ryzen anyway unless I get lucky with unloading all of these cards and perhaps other crap. And if I somehow get the voodoo. At least as of right now I think this board is "stable" enough as a compute server with the quad in it...right now I have a G810 in it I believe though I could use the 3650HD. Just that it's RAM limited, it has a hodgepodge of 3.5GB RAM right now, and Gentoo with less than 4GB is dicey on 64-bit.

I really should put the quad in my PVR and put its duo on this machine, however... then again really should stick more ram in the PVR and have it be a better compute farm component...

[eccerr0r]

Re: Gigabyte EP43-UD3L stuck in POST

Ahh... sort of got everything where they need to go, except RAM ... not enough DDR2 to go around.

PVR now has 4G and the quad... but C-states seems to crash the machine. The duo works fine with C-states. Blame the CPU?

If the core temperature sensors are correct... the machine is running cooler now when idling (duo is 65nm -> 45 nm quad CPU...but double core count) Could not get the temperature below around 55C with the duo, but now with the quad it idles around 35C or so, just like the other quad I have despite the other one having much better case cooling and it being an S.

So the duo went into this much more featureful EP43-UD3L, and again it seems to be working fine. Overclocked the duo to 3GHz, and it still seems to run fine... Odd thing is that the idle temperature is cooler in this machine! Idles around 45C, around 10C cooler. Weird, what gives. Wonder if the CPU likes the zalman hs/f more than the stock cooler where the quads don't care either way... hmm.

*shrug* weird...

[momaka]

Re: Gigabyte EP43-UD3L stuck in POST

PVR now has 4G and the quad... but C-states seems to crash the machine. The duo works fine with C-states. Blame the CPU?

Perhaps V_droop or something else on CPU V_core is causing the instability, since quad core is much more power-hungry under load... thus going between high load and low load / C-states may be causing too many voltage fluctuations on the V_core, and thus instability. Or it could be something else too.

Ah well, don't worry - you won't be saving that much power with the C-states off, unless of course the machine runs 24/7/365. Only then 3-5 Watts can be a few $ saved per year, depending on your utility rates.

Could not get the temperature below around 55C with the duo

Idle or at full load?
If at full load, I wouldn't be worried. But for idle, that's too high - either improperly mounted HS (cheap push-pins coming to bite you back? ), inadequate HS, wrong fan profile (unlikely), or bad TC between CPU heatspreader and die (check CPU model, and if core is not soldered to heatspreader, it may be time to open it and change the TC in there.)

but now with the quad it idles around 35C or so, just like the other quad I have despite the other one having much better case cooling and it being an S.

At idle, the temperature won't be too different, because all these Core 2 CPUs tend to have a similar dissipation. With load is when you will see a difference, depending on rated CPU TDP (and on that note, Intel lists many of their quads as "95" Watts... but "Intel TDP" != max power dissipation and rather only "average" maximum power... whatever that means.)

So the duo went into this much more featureful EP43-UD3L, and again it seems to be working fine. Overclocked the duo to 3GHz, and it still seems to run fine...

What's the S-spec or model number on that C2D?
3 GHz isn't bad, but you may be able to push it higher, depending on what model it is.

Odd thing is that the idle temperature is cooler in this machine! Idles around 45C, around 10C cooler. Weird, what gives. Wonder if the CPU likes the zalman hs/f more than the stock cooler where the quads don't care either way... hmm.

Yeah, ya think?!

The "tall" stock coolers (with the copper inserts) are OK for the C2Ds. I wouldn't use them on the C2Q's, though - those need beefy coolers if you want your temps to stay under 60-65C under max load. On that note, some of the stock Intel Pentium D coolers are pretty good for that - particularly those that came with the Pentium D 8x0 series, as those are 130 Watt TDP CPUs that really will peak up to 150 Watts at max load. (Then again, how surprising is that, given Intel slapped essentially two Netburst dies on there, lol.)

Oh, and the "short" Intel stock coolers - garbage for CPUs over 30-40 Watts TDP, IMO.

[eccerr0r]

Re: Gigabyte EP43-UD3L stuck in POST

C2D was not stable at all at 3.2GHz stock voltages, it's a bit less unstable at 3.3GHz with upped voltages but nevertheless still fails. I haven't tried 3.2GHz with upped voltage. 3GHz fine at stock voltages however, I think I'll keep it at this speed for now. This processor had been running idle at 55C or so on the stock heatsink on the other machine (µATX PVR+Stock HSF) where it's been running at 45C idle on the full tower case (Full ATX+Zalman HSF). Still don't get why the C2Q is at 35C regardless in the PVR or full atx with their respective HSFs. (The Zalman doesn't fit in the PVR...)

Actually both this 65nm C2D and the 65nm C2Q I had a while ago were quite warm idle according to the onchip temp sensors. Don't know why. The 45nm C2Qs are much cooler at idle for whatever reason. I don't have any 45nm C2Ds.

Now I really need to do something about RAM, just so I can have enough RAM in the PVR for it to be a better compute farm server...

The "stock" HSF I think it's the "mid sized" Intel HSF with copper insert. There is a "huge" one that was for the P4, and there's a tiny one for the reduced power one, and that's the one in the other quad and it still maintains low temperatures until it's pushed, which is expected.

[ChaosLegionnaire]

Re: Gigabyte EP43-UD3L stuck in POST

iirc, i believe penryns (45nm core 2) have c1e enhanced halt support so this allows them to idle at a very low tdp. conroes (65nm core 2) dont have this, so they idle at a much higher tdp.

[momaka]

Re: Gigabyte EP43-UD3L stuck in POST

This processor had been running idle at 55C or so on the stock heatsink on the other machine (µATX PVR+Stock HSF) where it's been running at 45C idle on the full tower case (Full ATX+Zalman HSF). Still don't get why the C2Q is at 35C regardless in the PVR or full atx with their respective HSFs.

Read my response above again.

In addition to C1E state that Chaos mentioned below, I think most (or all?) the 45 nm CPUs also come with "soldered" heatspreaders on the CPU core, whereas many 65 nm chips didn't.

Actually both this 65nm C2D and the 65nm C2Q I had a while ago were quite warm idle according to the onchip temp sensors. Don't know why. The 45nm C2Qs are much cooler at idle for whatever reason. I don't have any 45nm C2Ds.

The idle temperatures are _irrelevant_. You kneed to know if the CPU will stay cool under max (or at least very heavy) CPU load - that is, at least 5 degrees under the rated max T_case. Most Core 2 CPUs are rated for around 65-75C, so you should stay under those temperatures (check the sSpec number on your CPU and see what T_case it is rated for.) Exceeding the T_case will lead to shortened CPU life and instability.

The "stock" HSF I think it's the "mid sized" Intel HSF with copper insert.

That's the one I would use only on the Core 2 Duo CPUs, despite it being used on some lower-power Core 2 Quad CPUs. In elevated room temperatures, and especially in a PC case (even if it did have good ventilation), you can expect those stock heatsinks to run the CPU near max temperature. So the "mid-sized" ones with the copper insert are only good enough for C2D, in my experience - mostly because in the summer, parts of my house will approach close to 29-30C inside. I've tried running one of my "low-power" C2Qs (a Q8200) one on of these heatsinks in the summer, and it was idling around 55-58C. As soon as I loaded a Youtube video - even on 360p or 480p (which is no feat for the C2Qs... still) - the core temperatures would spike above 60C and roam around 65C. I never even tried pushing that system with a synthetic CPU test for more than a minute, because the temperature would get to 65C after no more than 20-30 seconds and kept going up.

Thus, the "oversized" Pentium 4 / Pentium D ones is what should be used with the Quads, at the very least.

and there's a tiny one for the reduced power one, and that's the one in the other quad and it still maintains low temperatures until it's pushed, which is expected.

That's terrible. You shouldn't be using that small heatsink on any Core 2 Quad at all - it's not up to spec for it! Heck, it's barely up to spec for a 45 nm Core 2 Duo, and not so much for a 65 nm Core 2 Duo. So on a Core 2 Quad, regardless if it's a 45 nm or 65 nm, you'll be destroying that CPU eventually... unless you have freezing temperatures where the computer is located... then you might be just barely OK. And I'm not making this up - I've actually done a ton of CPU HS tests with LGA775. The small heatsinks are trash for 65 Watt CPUs. Anything higher than 40 Watts will be pushing them very hard. Actually, one of the reasons we started seeing a lot of failed 6th gen and 7th gen i5 CPUs at one of the computer places where I used to work was precisely this - Intel bundled these same "tiny" coolers with many 65W CPUs. And being that the 6th and 7th gen i5/i7 CPUs don't have soldered cores to the heatspreader, they would easily hit 80+C on the core with AIDA64 (some riding right around the T_j max.) Sure, if you never push your CPU load much past idling, it may last for a while. But most of those machines we saw with dead i5 and i7 CPUs were from budget gaming machines (some of them OEM PCs with an added GPU by the user, others built by various e-tailers or custom-built by the user... so it's not like only the CPUs from one particular builder failed), thus suggesting high CPU load killed them over time due to temperatures not being in spec. I actually witnessed one of these CPUs as it was dying - would boot fine and work fine, but occasionally the system would crash... and that's while idle. When pushed with load, it was a crash rodeo. Took us a while to find out the problem, as not a single one of us was expecting the CPU to be the problem. When we swapped the CPU and the issue disappeared, we were like :wft: . Swapped back the old one, just to confirm, and sure enough the issues re-appeared.

Now granted these older Core 2 CPUs are built on much bigger lithography, so they tend to be a little more durable, IMO. But still, I think you're playing with fire (literally) by using those small heatsinks.