- Joined
- Sep 2, 2006
I have kept up with the thread at Xtreme since I have one of these coming. If I remeber correctly, all the dead cards have been flashed except for 4 of them. That was a company that has some in their machines and they put 4 of them in Xfire (2 x 2 in 2 different machines). They died after that but the single machines are still working. As to it being the Qimonda chips, it was not the chips themselves that died in the 8800GT's. My GTS has the same chips but the power circuitry has been beefed up vs the GT, hence no problem with the GTS's. I would wait awhile before blaming the ram chips.
[edit]
BUt I do admit I may not be fast to list my GTS on the classifieds, just in case
Your right about the qimonda chips in the 8800gts but the qimonda chips in the 8800gts were also not gddr5. As for the memory controller dieing like it did in the 8800gts from overvoltage I don't think that is the problem here with the 4870's has people are killing the cards without touching the voltage and besides the 4850's memory overclocks pretty well.
The ironic thing is that most of the time the cards are fine, it's just that the BIOS is telling to card absurd things (to do). Like one thread on XS mentions a guy who had a good flash, the card failed the next day. He used it in slave mode (booted up under different card) then viewed GPU-Z on the card. It was clocked upwards of 20000MHZ!!! Basically, without the proper knowledge of EXACTLY what bit in the bios does/affects, things can go crazy. It is also worth noting that most failures are SOFTWARE failures, which can sometimes be recovered.
GPU-z reports that because the registers on the card aren't initialized and everything gets reported back to GPU-z as 0xFFFFFFFF which ends up as those funky numbers. IF those registers aren't even initialized yet there's no way its gonna even make it to read the registers and set the clock.
The main reason for these card failures is because people flash and mod these cards without testing it properly determining it's limits, I don't think it is a RAM flaw or design issue at all. Once Rivatuner brings us an update we can test the cards the proper way before flashing the bios with OC profiles
I don't think people are going past the limit but as far as experimental bios flashing seems to be the problem as there a a bunch of similar posts to this: http://www.xtremesystems.org/forums/showpost.php?p=3100116&postcount=2