10-08-2011 03:17 PM
I'm trying to track down a suitible PSU upgrade from the 625w AcBel, that was stock on my S20, to something in the 850w rangs (800 - 900). The 10-pin EPS (CPU power supply) on the MoBo is making finding anything aftermarket difficult. I see that the C20 has an 800w PSU. I'm trying to find out if it is compatible and if so the part #. The IBM Phone system has been a little les that productive so far.
Any product recomendation would be greatly appreciated (aftermarket or manufacturer spec).
Thanks for your time - NHC
BTW - The new GPU I'm trying to power is just a AMD 6970, it's crashing after about 45 min. of strenuous use. All drivers are updated, temps are good, and system checks all come up positive. Hence my hypothesis of being slihtly underpwered. The PSU calc. I ran came up with an estimated requirement of 560w so running the old 625 is cutting things a bit close.
Solved! Go to Solution.
10-10-2011 04:37 AM
welcome to the forum!
to answer your intial question... no, the C20 PSU won't fit. not only is the box a different size, bolt pattern, and fan layout but the wiring would be a mix of too long and too short in all the wrong places. thinkstation PSUs are designed specifically for these systems so there won't be a direct replacement out there.
gaming cards, as you probably already know, aren't supported on workstations for various reasons. their insanely high power draw is one of them. but, 45 minutes is a long time for a power-related crash and AMD themselves say the minimum req's for your GPU is a 550W PSU.
have you run openGL and DX11 stress tests to rule out drivers or software? furmark is openGL and and 3dmark11 is DX11. if you don't crash right away using one of these then it could be a driver/software issue or the card itself. this doesn't rule out the PSU entirely but might help narrow down the hunt.
what application are you running for 45 minutes that causes the crash?
10-10-2011 09:23 PM - edited 10-10-2011 09:29 PM
1st off, thank you so much for not only taking the time to respond, but also for giving me a strait and definative answer.
While running your recomended test I also spent a good portion of the day on the phone with various aspects of Lenovo & IBM's sales and tech support. About all I could gather is that you are indeed correct; nowhere could I or any of the people on the other end, dig up any acceptible upgrades. At least nothing documented anyway.
Now, that aside, here are my test results:
Old GPU: 5770 scores: for open GL & DX11
Furmark score on 1080 benchmark: 1038 (17 FPS, 60,000ms)
Furmark Burn In test: 1600 x900 -22FPS ave, held stable @ 75
Deg. ran for 45 Min w/second monitor
3D Mark11 score P2828
New GPU: 6970 scores: for open GL & DX11
Furmark score on 1080 benchmark: 1725 (25 FPS, 60,000ms)
Furmark Burn In test: 1920 x 1080 res @ full screen -29FPS
ave, held stable @ 83 Deg. Set on auto,
fan never broke 60%, highest temp spike was 87 deg. ran for 1hr w/second
3D Mark11 score P5269
Additionally, I went back and ran a custom fan curve through
MSI Afterburner. With the new setting the system stabilized at 77 deg with a
fan speed of 66%. there were no spikes to speak of. I though the additional power consumption of
the fan might be the proverbial "straw", but there was no event after
45 min. of operation (furmark).
When I reinstalled the new card, I roled back to a slightly older AMD driver. Some of the smaller glitches I was getteing (like screen tearing on monitor #2), cleared up along with some artifacting (for lack of a better description) that I was getting in 3d CADD models. BF:BC2 (game) still crashed the system, and I have just scrubbed the system clean of all thing Battle Field and am currently redownloading it from E.A. along with a fresh install of Origin.
As an additional test, I reinstalled Metro 2033 (A extremely demanding game) and just ran it for 2.5 hours on very high settings without incident. Even at 95+% GPU usage the fans and temps stayed in the 60s and 70s respectfully.
I'm now hypothesizing that with all the card flipping I did initially, I left a Beta video driver in and/or corrupted a driver. I'm not at all sure why BF:BC2 would fail and lock up the system (It's been very stable for over a year), but we'll see what the new install does.
Since it, looks like the PSU stays, my only consideration there is that I need to test it to be sure it hasn't degraded over time. I'm looking into it now, but if you have any advice on a sound method for testing (particilarly under load), I'm all ears.
I'm truly greatful for the advice and links you provided, particulary regarding the non-compatitibility of the other PSUs. That probably saved me a headache. I'll keep you posted, mabey we can add this to your "resolved" pile. And someone else can be spared hours of fruitless search & speculation.
Thanks again - NHC
BTW - This machine was for CADD only when I bought it but has turned into my primary desktop (I don't do much drafting now) - hence Frankenstien's build.
10-11-2011 05:06 AM - edited 10-11-2011 05:10 AM
awesome. that's an exceptional amount of testing and definitely points more toward this being a driver issue rather than something power-related. having gone down a similar road myself (not with games but rather with desktop GPUs), i can certainly appreciate the headaches involved here.
i'm not sure how finicky AMD drivers are but do know that nvidia fermi-based cards are ridiculously picky with how they're installed, updated, removed, etc. if you still have issues in your current setup then my next suggestion would be to find a spare hard drive and test a clean install of windows with known-stable WHQL drivers.
while i'm not a gamer by any stretch of the imagination, i can appreciate one's desire for running a desktop card in their workstation. hop on the autodesk forum and you'll find a healthy amount of users running DX-based geforce and radeon cards purely due to autodesk's shift from openGL to DX a few years ago. the so-called "gaming" cards outperform traditional workstation cards by a huge margin. i've personally tested a handful of quadros (FX4800, Q600, Q4000, and Q5000) against a geforce GTX 570 and 560 Ti in autodesk inventor 2012 and 3ds max 2012 and the GTX cards performed flawlessly with WHQL drivers. the quadros were slow in comparison, including the venerable Q5000. the GTX cards proved stable with large, complex assemblies, finite analysis and mold simulations. so, for all the years i've given people slack over running gaming cards in their thinkstations, i'm soon to eat crow and join them—but definitely not to run games. openGL apps like solidworks, catia, pro/e (now creo) absolutely need quadro and fireGL/firepro cards though. there's no free lunch in that case unless the rest of the industry shifts to DX as well.
anyway, give your setup some time to burn through the new drivers and see if things remain stable. if you find a good driver combo then stick with it and don't upgrade unless absolutely necessary. also, stick with WHQL-certified drivers, skip the betas, and install any new drivers as cleanly as possible.
regarding your current PSU, i'm sure it's perfectly fine. even though gaming cards can be power-hungry, workstation PSUs are designed to output clean power 24/7 under stressful loads like rendering, analysis, simulation, etc. i highly doubt you damaged yours in any way.
good luck. keep us posted on your results.
(edit: added info)
10-19-2011 12:31 PM
I can just about guarantee you that power is not your problem. From what I can find, the 6790 requires 2 x 6pin power connectors. These connectors supply ~75W each. With 75W from the board (by spec), this card theoretically maxes out at 225W. Based on the specs of the card, I would say the max power is actually well below that.
S20 can easily support a 225W card, along with another standard 75W card as well. So power on this rail is plentiful, and is unlikely to be the cause of your problem....especially since I don't think your card is anywhere near 225W.
As Erik stated, drivers is where you want to focus. Desktop/gaming cards aren't workstation class cards, are are well known to be more buggy/crash-prone (though ideally not as bad as what you're experiencing).
10-20-2011 10:23 PM
Thank you both,
Problem solved, it was actually a software conflict. Turns out the anti-cheat/hack monitoring software in some games conflicts with certain system monitoring applications. Often enough the result is a graphically induced lockup.
In my case it was MSI afterburner, which I was using to document the new GPU's performance & reset the cooling fan to a more gradual curve, and "punkbuster" EA's anti cheat software. I guess there are many variations and combinations of this conflict; this one happens to be pretty well documented (with some digging).
I substantiated it by installing all superfluous programs and the reinstalling the ones I actually use, one at a time.
psuturtle - I only make this correction for the benefit of future readers who may stumble upon this little issue. For the record, the GPU in question was a 6970 which requires one 6 and one 6+2 connection (incidentally exactly what the S20's power supply offers).
I know I'm improperly using this machine, but I can now (after some more marathon gaming and stress tests) confidently say the S20 works extremely well with this unsupported GPU. Additionally, game developers are starting to write applications that can actually take advantage of hyperthreading, It's exciting for us cross discipline PC users.
Needless to say there will be no future upgrades for this machine, but it has been phenomenal for four years and I expect to be able to stay one the leading edge of software usage for at least another year or more. By then maybe I'll splurge for a gaming rig.
Thank you and this community again for taking the time to consider my problem - NHC
10-23-2011 04:41 PM - edited 10-24-2011 09:32 AM
speaking of naming conventions...
s20 s30 (edit: looking ahead on this one)
10-24-2011 09:05 AM
Yeah I've complained about that one too. You can see how far that got me
I "think" we (being the ThinkStation team) were first in using that naming convention going back to the S10, but I could be wrong.
Are you sure there's a thinkpad s20? I knew about the ideapad s20, but didn't know there was a thinkpad s20.