Polymer Replacement Issues

[JEWilson]

I have two boards and I would like to replace the wet lytics with
polymers (where or if possible)

One mobo (already recapped) has 10 of 6.3V 1200uF Pana FMs
Can I replace these with OSCON-SEPC 4v 560uF?

I can alway re-use the FMs I intend to replace in boards where
wet lytics are the only option.

The other mobo has yet to be recapped.

Whereas the above (1st board) is a reference design, the next and newer
mobo has less caps on it (as usual). I intend to follow the recap
method for the reference mobo even so, as it has the same silk
screen as the ref design.

In fact, I have another three boards which I keep as spares
and if ok to replace with poly's - will do these boards as well.

[willawake]

Re: Polymer Replacement Issues

you can measure the voltage across the caps with a multimeter while the board is running.

i dont see the point to recap with polymers except on vcore and vdimm out.

the general guideline is half capacitance is ok for polymer. in the case of vcore and vdimm out 4v is fine and used in practice on many boards. sepc will be excellent.

[JEWilson]

Re: Polymer Replacement Issues

Thx will

I recall you have mentioned this before in other posts.

It is in fact, Vcore and Vdimm only with which I seek to use these.

I've got about 50 of these SEPC

These are for a Gigabyte GA-7DXR - ref. mobo - already recapped
as stated. The other board is a Gigabyte GA-7DX+.
The spares are all Gigabyte GA-7DX.

Only use them, because they all have the ECC Reg.

As I understand it, they use +5V for VRM in these designs

[willawake]

Re: Polymer Replacement Issues

board specs say :

vcore 1.5V/1.525V/1.55V...1.85V
vdimm 2.5V/2.7V/2.9V

As I understand it, they use +5V for VRM in these designs

which is why to be careful where to put the caps, not on the input

[JEWilson]

Re: Polymer Replacement Issues

Y - if I have read this correctly,

the Input side has 4 of 10V 1200uF right beside the MOSFETs

thx

[willawake]

Re: Polymer Replacement Issues

it appears so but the GA-7DXR has strange layout. make sure to measure with multimeter prior to replacing with lower capacitance.

[davmax]

Re: Polymer Replacement Issues

you can measure the voltage across the caps with a multimeter while the board is running.

i dont see the point to recap with polymers except on vcore and vdimm out.

the general guideline is half capacitance is ok for polymer. in the case of vcore and vdimm out 4v is fine and used in practice on many boards. sepc will be excellent.

It would not be a good idea to recommend any significant reduction in capacitance on this forum. Capacitance does play a significant role. ESR is just not the only game in town.

See this thread: https://www.badcaps.net/forum/showthread.php?t=5032

[gonzo0815]

Re: Polymer Replacement Issues

The GA-7DXR is may be using only a 2 phase VRM, if this is the case i would although not recommend to lower total capacitance, as AMD K7 CPU`s are drawing a heck of current.
For 3phase VRM`s i think it may work well.
If you want to be on the safe side, i would certainly opt for 1000uF polymer capacitors.
In any way, you has to understand that reducing capacitance is always experimental.

If you decide to do the mod, i would be pleased to know your outcome.

[JEWilson]

Re: Polymer Replacement Issues

Well...

I have recapped the board, the GA-7DX+ as follows;

Replaced 4 of 3300uF 6.3V GSC with 10 of 560uF 4V Sanyo OSCON-SEPC
As Will stated, I have gone along the lines of the half capacitance method
or thereabouts, 10 of 1200uF = 12000uF (see below), 10 of 560uF = 5600uF.

In the ref. board GA-7DX+, there was 10 of 1200uf 6.3V GSC
So... the silkscreen and mobo layout is almost the same with both mobos.

As you surmised, the VRM is a two phase design and the input
is powered from the +5V rail (it is an old board but newly used) not
the +12V rail.

The I/P side of the VRM had 4 of 10V 1200uF GSC on the GA-7DXR
The I/P side of the VRM had 3 of 10V 1200uF GSC on the GA-7DX+
I replaced both mobos with 2 of 10V 1200uF and 2 of 10V 1500uF Pana FM

Obviously, the O/P side of the VRM and VDIMM are powered (now) by
the SEPCs.

Also, I replaced all of the 25V 330uF caps with Pana FM 16V 330uF.
These are obviously for the +12V power to the mobo.

If this is a particular problem, then it may be, I will have to replace the
560uF SEPCs with a higher capacitance value. As I have recapped it,
I do not like the prospect of taking the mobo out again for more recap
work.

BTW, the VIA Hardware Monitor v2.05 still reports VDIMM being too low,
this is reported as being between 1.07 and 1.09V

I'm going to measure it as time suits just to make sure

Is there anything else I should be aware of?

Thanks for the input

[willawake]

Re: Polymer Replacement Issues

looked like 3 phase on gigabyte site



sorry i should have said that it would be "experimental" mod.

what is happening now? did you stress test ok?

[JEWilson]

Re: Polymer Replacement Issues

Thx all

I should issue apologies.
Y - I am blind, just as well as I am getting new spectacles (y, really).

It is a 3 phase design - this is obvious from the pic and it is so
on the board. You can also see, the cap decals around the CPU
and beside the DIMMs, these are replaced with the SEPCs.

The I/P side of the VRM are beside the modem slot - in this pic, 3 are populated

Sorry - for the confusion
I am actually using this board to make these posts

Y - I have yet to stress test the mobo - will do with Prime95
Will report back results in due course

Thx again

[willawake]

Re: Polymer Replacement Issues

one solution to the lack of bulk capacitance noted by the naysayers would be to increase the capacitance heavily on input side. this would be quite interesting and yes experimental

do you intend to overclock the board? such experiments could be validated well if results before and after could be compared.

i dont believe the stress testing with prime95 would be a final validation though although it would certainly be useful, in respect of the reduction of capacitance you really need to heavily load the cpu and then idle repeatedly because this one of the functions of capacitors to support the cpu in load transients, not only for the smoothing function which is probably served better by the sepc.

[JEWilson]

Re: Polymer Replacement Issues

Y, Will

That is what I have done.
The ref. board (GA-7DXR) has 4800uF on the I/P side of the VRM.
I have replaced this with 5400uF 10V Pana FM on both the ref. board
and the GA-7DX+.

I have conducted the first hour of Prime95, which test if the
software is working properly. The board has stood up to this
with no problems.

Of course, further testing will be necessary.

Just a point, is there evidence that crapacitors such as GSC
and G-Luxon are really ambitious in respect of the specs. they
claim their capacitors support?

I believe, certainly the ESR claims are a fiction but does this
apply to claimed cap ratings such as voltage and capacitance also?

I state this to be so, as usually trying to find good replacement
caps in the same phy. form factor is almost always a problem.
This is a well documented fact in this forum.

Would you agree?

Of course, I will devise a heavy/light load schedule which may prove to
be a better indication of the effectiveness of this recap.

Thx

[gonzo0815]

Re: Polymer Replacement Issues

I think G-Luxon may meet the spec at the start, but longevity as we all know is not one of their strength.
If they have no concerns in selling bad caps, i would say they do not apply any safety margin on their capacitors.
Thus they may can offer CV products not met from serious vendors.
On the other hand, if you buy large quantity, even the better vendors will do custom parts.
But it seems that most of the hardware manufacturers do only stock cheap crap caps in large quantity.


The problem i have seen in my solid polymer recap of an Asrock K7s8xe+ where more frequent voltage spikes.
As i have had no Oscope at that time, i was not able to track this issue down.
Furthermore it was a highly over clocked Athlon XP and it was an 2 phase vrm probably already at the limit regarding bulk capacitance.

Increasing the input capacitance can not equal the loss of output bulk capacitance, as the limiting factor is the choke`s charge time.
Any way, a 3phase VRM has in most cases a far faster response then a 2phase design.

[JEWilson]

Re: Polymer Replacement Issues

Just to report back on progress with this experiement.

First, all things seemed well...
Over the last two days however, the mobo has had problems
starting.... It will power up ok, just has problems posting.

However, if I use a hard reset after a min with the system powered,
it will POST.

Symptom - I suspect the polymer replacement capacitance may
be a problem as has been suggested by others.

I have reviewed the OSCON SEPC datasheet and have found,
1. The 560uF 4V has an SV (service voltage) of 5.2V
2. The 820uF 2.5V has an SV of 3.3V - these would still be suitable.

So... I may want to raise the VRM O/P capacitance from;
1. As it is, 10 x 560uF 4V SEPC (5600uF) to;
2. As it originally was, 10 x 1200uF (12000uF) 6.3V Pana FM or
3. Raise the SEPC to 10 x 860uF 2.5V (8600uF).

If I pursue the latter I will require to source SEPC at 860UF
Where might I be able to get these?
Of course, it may be an alternative problem.

I should add this is a moderately loaded PC. It runs
1. ATI FireGL8800 AGP 4x 128MB
2. 2 of matched pair Infineon PC2100 512MB ECC Reg
3. Promise SX6000 6ch RAID
4. 3 of Seagate 7200rpm 80GB 2MB cache HDD
5. DVD/RW
6. Athlon XP-M Barton at 2.0GHz
7. 3COM PCI NIC
8. 5 of 12v Fans
9. 1 of WDC 7200rpm 40GB 2MB cache HDD
10. FDD

This is powered by an Enermax EG651P-VE FMA

I should add further, the separate thread at
https://www.badcaps.net/forum/showth...2&page=2&pp=20
details the replacement CPU for this mobo

Thanks for your consideration

[gonzo0815]

Re: Polymer Replacement Issues

I would suggest to order some Samxon X-Con 1000uf 4v. You can source them from http://www.phenos.nl. He does offer some Nippon Chemi-Con PS 820uf too, if nthe price of the X-con does not suite you ;-)

Regarding your problem, i think it could may be the very low ESR of the Oscon which could cause this start up issue. Still only a guess but may be this VRM can`t handle the high inrush current.
But sure, changing some of the caps may be one at any phase with an larger Lytic capacitor should proof if the issue is due to low esr or low capacitance.
May be you can just solder an additional lytic at the bottom side of the pcb, i mean in ghetto style ;-) just to check this out.
If the issue would be solved with a few more capacitors, it is an capacitance, not an ESR problem.

[willawake]

Re: Polymer Replacement Issues

how come first it was ok now its not
does not compute

[gonzo0815]

Re: Polymer Replacement Issues

Don`t get what you mean Willa.

[JEWilson]

Re: Polymer Replacement Issues

Just some update info

Started to lose BIOS settings, changed old BIOS battery with new one.

Also, as I used CrystalCPUID to set the mult of the Barton XP-M to x15
It implemented an overclock of 0.52%
see further
https://www.badcaps.net/forum/showth...2&page=2&pp=20

The RAM in situ at a stock CL of 2 by SPD had a max speed of 133MHz.
The slight overclock raised the FSB above this maximum.
In turn, I manually set the CL to 2.5 in the BIOS which sets the max FSB at 143MHz to accomodate the speed above 133MHz..

Q - Is it possible the software setting of the multiplier to x15 with
CrystalCPUID so far as it has raised the FSB etc has marginally raised
the voltages and other frequencies out of specification?

I will need to get a readout of these with a view to modifying them,
where or if possible. Any good utils recommended?

I've attached the dump of CPU-Z and it can be seen the effect of
changing the CL from 2 to 2.5

Ideally, I would simply lock everything to the stated speeds.
So....
FSB would be certain to be 133MHz
DDR would be certain to be 266MHz
PCI and AGP would be 33 and 66MHz respectively
Same for the voltages

Thanks (again)

[zandrax]

Re: Polymer Replacement Issues

0,52% overclocking is nothing at all: this is due to fsb tolerance (1 or 2 MHz on 133 is something like 2%, almost negligible).
I don't like overclocking: if you're careless, there are chances breaking something; moreover Athlons XPs were quite good at overclocking but heated a lot: mainstream liquid coolier were launched for Athlons and P4s.
Remember: if you rise fsb, the gain in bandwith should at leats be equal to loss in memory latence. Programs should run at same or higher speed to justify an overclock.

If you really want to, try something like 5% at first: overclock fsb from 133 to 140-142 MHz, set CAS latence from 2 to 2.5 or even 3 and lover cpu multiplier from 15 to 14. You should obtain a near standard cpu frequency (142 * 14 = 1988 MHz vs 133 * 15 = 2000 MHz) which doesn't pose a heat issue at all and you have an higher memory bandwith. Test your ram with memetset86+ for long (say 5-6 cycles after any fsb change) and revert to last stable fsb value when you get a memory error; then you can check performance improvement with SuperPI or other cpu benchmarking tools.
I think you should afford a 10% oc with standard voltages: increasing vcore or ram voltage more than 6-8% is never reccomendable because it increases heat and shorten chips lifespan (increases electromigration, so chips are unstable: they require an higher voltage to be temporary stable again, then are newly unstable; this degenerative process ends only with components death or irreversible damage).

Keep in mind: overclocking higher than 15% requires additional cooling and better equipment in most cases; moreover, extreme oc lasts for very small time (few hours) while your goal is having a rock stable plattform for months or, better, years.

Enjoy your hacking

Zandrax