-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
970's having performance issues using 4GB Vram - Nvidia investigating
- Thread starter kitch9
- Start date
- Status
Soldato
- Joined
- 17 Aug 2003
- Posts
- 20,168
- Location
- Woburn Sand Dunes
It's 3.5.1.0 1280x720, no AA 
lol Orangey you're not helping
I'm just saying if it works fine on my machine then its probably not a reliable way to test for this issue
lol Orangey you're not helping
I'm just saying if it works fine on my machine then its probably not a reliable way to test for this issue
Last edited:
It's 3.5.1.0 1280x720, no AA
OK. How are you changing the vram?
Have any tech sites chimed in with their own results (preferably using review samples)?
I imagine it would be quite the scandal if the review samples showed to be good throughout the whole memory range.
Otherwise whilst I agree it's a problem that should be fixed, it's not the end of the world. The 970 you've bought still has the same merits that you presumably bought it for.
I imagine it would be quite the scandal if the review samples showed to be good throughout the whole memory range.
Otherwise whilst I agree it's a problem that should be fixed, it's not the end of the world. The 970 you've bought still has the same merits that you presumably bought it for.
Soldato
- Joined
- 17 Aug 2003
- Posts
- 20,168
- Location
- Woburn Sand Dunes
No on the 660 cards it was because of their 192bit bus being split into memory fractions.
After some further testing heres an image of how the increase in memory usage is affecting framerates, and cpu memory usage. This needs to be spread around and shown to Nvidia.
[/URL][/IMG]
As you can see the framerate drops significantly when the memory usage goes above 3072mb. I tested with MSI Kombuster 3.5.1.0 (GPU memory Burner tests - 1024mb, 2048mb and 3072mb). The windows task manager shows a rather sharp increase in system memory usage when the 3072mb testis started.
The spikes and fluctuations on the 3072mb test in the MSI afterburner are due to huge freezing/slow loading and a long pause before the MSI logo starts.
As you can see the framerate drops significantly when the memory usage goes above 3072mb. I tested with MSI Kombuster 3.5.1.0 (GPU memory Burner tests - 1024mb, 2048mb and 3072mb). The windows task manager shows a rather sharp increase in system memory usage when the 3072mb testis started.
The spikes and fluctuations on the 3072mb test in the MSI afterburner are due to huge freezing/slow loading and a long pause before the MSI logo starts.
Last edited:
ok fair enough thanks
I do wonder if it can be driver level issue though as thought drivers are pretty unified these days
what would be done differently in driver vs 980 ?
Same results here roughly. Considerable drop in FPS when using the highest VRAM test. (MSI Gaming GTX970 OC'd)
spread the word guys. This is serious stuff.
Same thing is happening with my Titan, though. I go from 242 to 90 odd.
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 37,804
- Location
- block 16, cell 12
Would be interesting to see what happens on a 3 amd 6gb ati card.
Interesting. Maybe Kombuster works by increasing detail or something hence the lower frames. But the main issue is that my gtx 970 takes significantly longer (about 1 minute) to load the 3072mb test while the other 2 tests load within 2 seconds. This is indicative of some bandwidth problem over the 3GB region as found from the original tests.
How long does your Titan take to load?
Or a problem with the rest of your PC, this can also effect loading times when the workload gets high.
Oh it's instant mate. No waiting at all.
I tried to run that other benchmark but got a Cuda DLL error. I really CBA looking for DLLs.
I have a 3gb 7990 (well 6gb, but 3gb per card) can possibly run it on that later.
I've heard a lot of rumours that the 780 does this too, so it could be a more widespread issue, or, not an issue at all.
Just a matter of waiting now and seeing if any better tests come through. I have a feeling some one is going to have to write a specific test to get to the bottom of this though.
- Status