Hey folks,
I've been an AT reader for years, but I decided to create an account to respond to this alleged issue. On 5/7 I purchased an EVGA 7900GT KO P/N 256-P2-N564-AX (see
http://www.newegg.com/Product/Product.asp?Item=N82E16814130015 ) and just a few days later I received it. After about 2 days of intensive gaming, I began observing the following symptoms:
1. Graphical artifacting/defects/tearing
2. 3D Applications/games hanging
3. Windows BSOD
4. Monitor flickering (intermittant signal loss) when exiting a 3D app/game
I promptly contacted EVGA for a cross-shipped RMA. A few days later, I get a brand new retail-boxed card. It worked absolutely fine for 2 days--just like the first card--but then began exhibiting the same symptoms. At this point, I began to wonder if I was just really unlucky, or there was a more widespread problem (i.e. a design flaw affecting some, but not all 7-series video cards). A quick Google search found many results of people experiencing the same problems. Apparently some vendors have closed their message boards because of the deluge of complaints about these cards. It is important to note that the symptoms are the same in almost every case, as is the fact that these cards work perfectly for a matter of hours or days before they begin to fail. You only have to take a glance at EVGA's forums to realize the extent of these issues (see
http://www.evga.com/community/messageboard/forum.asp?FORUM_ID=32 ).
Of course, EVGA as well as other vendors have been quiet on the issue. In my case, when I called in to report my RMA replacement was exhibiting the same symptoms as the first, the CSR (customer service representative) immediately blamed my computer. The CSR stated that the 7-series are power-hungry cards and require a minimum of 22A on a *single* +12V rail. Nvidia's own "recommendation" is 22A, but it does not say whether this is combined or not... I would assume so. I told the EVGA CSR that my power supply is no cheapie, a Seasonic S12 500 Rev. A2 (cream of the crop according to SPCR, see
http://www.silentpcreview.com/article247-page1.html ) which is Nvidia SLI certified. The CSR claimed that SLI Certification is purely marketing, and further claimed that every RMAed card they received was just "fine" and that the problem lay entirely with the customer's equipment. Of course, I told the CSR that he was completely wrong, but I wanted to be sure (as I reasonably could be).
I decided to run some tests to see if the 7900GT was really as power hungry as EVGA claimed. Test equipment included a calibrated DMM (digital multi meter) and a P3 International Kill-A-Watt meter. Current measurements (in watts) are PEAK figures, i.e. worst-case scenario. During testing the +12V line was continuously monitored and the power supply's exhaust heat and fan speed observed.
My computer:
===============================================
AMD Athlon64 X2 4200+ @ 1.25V (TDP = 53W @ 1.35V)
1GB (2x512MB) OCZ Platinum Rev. 2 PC3200
Epox 9NPA+ Nforce4 Ultra
NEC ND-3540A DVD-RW
200GB Samsung P120 SATA3.0 HDD
Seasonic S12 500 Rev. A2 (+12V1 17A, +12V2 16A)
Test results:
===============================================
IDLE - 101W AC, 12.09V
RTHDRIBL - 155W, 12.09V
FEAR - 152W, 12.10V
FEAR TEST - 155W, 12.10V
3DMARK06 HDR1 Canyon Flight - 159W, 12.10V
3DMARK06 HDR2 Deep Freeze - 165W, 12.10V
So the maximum that this computer will pull when running a 3D application is 165W AC. At this load, the PSU was exhibiting no increase in fan speed and only luke-warm air was coming out of the exhaust. Measured +12V rail at 12.10 volts is well within ATX spec (less than 1% deviation, spec is +/- 5%) and hardly changed at all between idle and load conditions (+/- 0.01V).
If we assume the PSU is at least 80% efficient (and this is a conservative estimate according to Silent PC Review, see
http://www.silentpcreview.com/article247-page4.html ), then thats 165*0.8 = 132W DC. Just for the sake of argument, let's say all 132W DC is coming off a single 12V rail, that's only 132/12 = 11A! Not even close to the 22A that the CSR claimed I needed on a *single* rail! According to Xbit Labs (source:
http://www.xbitlabs.com/articles/video/display/evga-7900gt_4.html ), an EVGA 7900GT CO SC draws a maximum of ~57W DC, or just 4.5A... which would agree with my findings.
My suggestion to anyone else who is hearing, "your power supply isn't good enough? then here's your proof that you?re being misinformed (at best) or deceived (at worst). EVGA needs to better educate their CSRs that you don't need 22A for a single 7900GT if my whole computer draws a maximum of 11A. I'd say that 22A for an entire medium-range computer would be safe (see
http://www.silentpcreview.com/article28-page4.html ), but not necessarily required. Even a decent 300W power supply can provide 22A (see
http://www.newegg.com/Product/Product.asp?Item=N82E16817104907 ).
I'm still considering my options. Whether to keep trying to RMA in the hopes I'll get a working card, or to just send the original back to Newegg for a refund. It's too bad, because the 7900GT is a good card... when it works.
In any case, if some high-profile website like AT brings light to this issue, we'd likely see a much quicker resolution to these widespread problems. It?s likely reviewers didn?t see these issues because either they were lucky enough to get a good card, or they just didn?t spend enough time with the card (this seems more likely).