I also find that benchmarks alone don't show stability. I've had cards appear dead stable, only for them to slump in one of 3 games:
COD MW2
F1 2010
DIRT 2
In the racing games, the game has random 0.25second maybe once or twice a lap
In CODMW2, the game just freezes for a good 30+ seconds once in a specific level (Act 1, mission 2 iirc)
Originally Posted by storm20200 Always shocks me how massive current Nvidia GPU's are, are they bigger than LGA1366 processors?
The actual GPU is much smaller, its just a big heatspreader sitting on top of it.
Err.. Nehalem is 263mm^2 while GTX 580 (which is the same size as the 480) is 529mm^2
3.01 BILLION transistors versus 731 MILLION, the 480/580 should definitely be bigger.
How does the card's power management affect performance on an over-clocked GPU? Presumably there is a possibility that after several hours of playing something like Bad Company 2 at maximum settings and a high resolution that the VRM's may got too hot and the card's power management kicks in, resulting in lower performance than when you first launched the game?
furmark is pointless to run now unless you have used the gpu-z overclock flag to disable the power limiter (not read all of it yet to see if you did or not)
Rather pathetic over clock with minimal results at a higher price premium. Not really sure it would make a true difference for gamers. If they could remove the control that limits the current to keep the card from getting too hot..... ????
Originally Posted by Farting Bob That RAM overclock is a bit pathetic, but then Fermi doesnt seem to like fast RAM, it just likes having alot of it.
Yeah, we though that, the RAM overclock was a little pewny. It'd run quicker but we had a real problem with artefacts so we had to dial it back. The complexity of the 384-bit interface is probably a factor here.
afterburner has a user adjusted fan ramp in the options.. might squeeze some more.. in that pic- interesting the link is lit up? does that mean you can adjust the shaders free of the clock, like in the 200 series
Originally Posted by pistol_pete Extra vat on a £400 card would be... £10? Hopefully price wars would drop things more than that to cancel out the vat rise...
that's the thing. would price drop enough to wait or to buy it now and enjoy it? i'd say £1 per day of devaluation seems reasonable for these high end cards at beginning of its shelf life, so in 20 days for 6970 to come out, then another 30 for the price to drop, would gtx580 hit £350 before end of the year? i highly doubt it's going to drop that fast.
I would say wait until March 2011 and see which one to buy. Honestly there isn't a game around right now that's worth the upgrade.
However Crysis 2 will come out in March and I would really recommend buying a new GPU then.
Also save up for the best possible card - since you ideally want that to last 2+ years and not upgrade next year. Unless you're a money spending whore who likes upgrading every fortnight
Originally Posted by cgthomas I would say wait until March 2011 and see which one to buy. Honestly there isn't a game around right now that's worth the upgrade.
Originally Posted by cgthomas I would say wait until March 2011 and see which one to buy. Honestly there isn't a game around right now that's worth the upgrade.
*unless you have a 30 inch screen.
True that - but still March is just around the corner. I bet you could down scale your resolution :p
Originally Posted by Tattysnuc Just out of interest, but with Bit-tech having such a presence in Folding, why is folding not used to stress test and benchmark?
Oddly, in my (somewhat limited) experience of Folding, it's doesn't seem to stress the card hard enough to make it error the same way a really demanding game will (the two I think of here are actually Just Cause 2 and Civilisation V in DX11 mode rather than Crysis, which isn't too bad).
I recently had my GTX460s folding for a few days and they were rock stable at 875MHz (from stock 700MHz, didn't take them higher) with no voltage bump (the VID of these cards is 1.025v and 1.000v) when folding... but ask them to do Just Cause 2 or Civ 5 DX11 at that speed and voltage, and... well, JC2 just locked up as soon as the game loaded, and Civ 5 DX11 worked for about five minutes then took the computer down with it. Eventually I just went back to running them at stock as it was less hassle.
Originally Posted by Tattysnuc Just out of interest, but with Bit-tech having such a presence in Folding, why is folding not used to stress test and benchmark?
Also, what's the power draw like when overclocked?
It's reassuring to see that my watercooled twin GTX 480's aren't being outdone by too much performance wise.
The Folding @ Home application is not a benchmark - it's a scientific research distributed computing project. The creators of the F@H software have made it clear numerous times that their software is not to be used as a stress test or a benchmark for new or overclocked hardware. If a card or CPU EUE's several times duting a "F@H stress test" then those work units are wasted along with the Stanford work servers bandwidth and processing time.
Simply put, don't use F@H as a stress test app or benchmark even if it is the best thing for doing so. That's not what it exists for.
Comments 1 to 25 of 34
ReplyAlso, what's the power draw like when overclocked?
It's reassuring to see that my watercooled twin GTX 480's aren't being outdone by too much performance wise.
The actual GPU is much smaller, its just a big heatspreader sitting on top of it.
COD MW2
F1 2010
DIRT 2
In the racing games, the game has random 0.25second maybe once or twice a lap
In CODMW2, the game just freezes for a good 30+ seconds once in a specific level (Act 1, mission 2 iirc)
3.01 BILLION transistors versus 731 MILLION, the 480/580 should definitely be bigger.
Or do all GPU's do this to a certain extent?
Yeah, we though that, the RAM overclock was a little pewny. It'd run quicker but we had a real problem with artefacts so we had to dial it back. The complexity of the 384-bit interface is probably a factor here.
buy gtx580 now for £395? or wait until 6970 come out and buy next year when tax are up? (but prices might drop for both cards)
also, want to play HAWX 2 with tessellation, and have high doubts about 6970's abilities. so will almost definitely go for gtx580 anyway.
On a side note as the drivers mature the performance will just get better still.
Dont listen to me mate its your money.
However Crysis 2 will come out in March and I would really recommend buying a new GPU then.
Also save up for the best possible card - since you ideally want that to last 2+ years and not upgrade next year. Unless you're a money spending whore who likes upgrading every fortnight
*unless you have a 30 inch screen.
True that - but still March is just around the corner. I bet you could down scale your resolution :p
I recently had my GTX460s folding for a few days and they were rock stable at 875MHz (from stock 700MHz, didn't take them higher) with no voltage bump (the VID of these cards is 1.025v and 1.000v) when folding... but ask them to do Just Cause 2 or Civ 5 DX11 at that speed and voltage, and... well, JC2 just locked up as soon as the game loaded, and Civ 5 DX11 worked for about five minutes then took the computer down with it. Eventually I just went back to running them at stock as it was less hassle.
The Folding @ Home application is not a benchmark - it's a scientific research distributed computing project. The creators of the F@H software have made it clear numerous times that their software is not to be used as a stress test or a benchmark for new or overclocked hardware. If a card or CPU EUE's several times duting a "F@H stress test" then those work units are wasted along with the Stanford work servers bandwidth and processing time.
Simply put, don't use F@H as a stress test app or benchmark even if it is the best thing for doing so. That's not what it exists for.
-
« Previous
-
1
-
2
-
Next »
Discuss in the forums