HardOCP looking into the 970 3.5GB issue?

Status
Not open for further replies.
They shouldn't be given time to fix the problem, if it exists. We need a witch hunt.
 
Nobody is asking for a witch hunt, what would be good is having a reputable review site do some testing to prove whether or not this is a quantifiable problem across the 970 GPU models.
 
Huh interesting. Nvidia is actually checking on this problem?

IMO i would rather have them fix the SLI vcore bug lol.
 
DASHlT said:
Huh interesting. Nvidia is actually checking on this problem?

IMO i would rather have them fix the SLI vcore bug lol.
Click to expand...

I'm not sure there is a voltage bug unless I am missing something?
As far as I recall, the "problem" is that 2 identical (ish) cards in SLI both use different voltages to do the same frequency.

The temperatures each card runs at are not the same.
The voltage is dynamic, it scales to remain stable.
The temp difference can result in a different voltage being needed for each card.
 
Nenu said:
I'm not sure there is a voltage bug unless I am missing something?
As far as I recall, the "problem" is that 2 identical (ish) cards in SLI both use different voltages to do the same frequency.

The temperatures each card runs at are not the same.
The voltage is dynamic, it scales to remain stable.
The temp difference can result in a different voltage being needed for each card.
Click to expand...

This has been an ongoing issue since release. If I do not uncheck sync SLI My games will crash as 1 card will have too low of a voltage. It is a known problem.

Anyway. Right AMD and Nvidia can both go to hell with driver issues etc. Bring back 3dfx!
 
convexion said:
If you want to keep voltages more in line between cards, flash a custom BIOS. I'm using one I customized myself for temp/volt efficiency based off of the instructions here:

http://www.overclock.net/t/1517316/...d-980-firmware-zosons-h2o-and-air-custom-bios
Click to expand...

Trust me I know this. I have been dealing with this since the begining.

As you see we have all been trying to troubleshoot this issue. http://hardforum.com/showthread.php?t=1838369 (look who started the thread 3 months ago)

My point is, users shouldn't have to fix an issue Nvidia caused.
 
DASHlT said:
This has been an ongoing issue since release. If I do not uncheck sync SLI My games will crash as 1 card will have too low of a voltage. It is a known problem.
Click to expand...
Ah I see.
Is not syncing SLI undesirable?

op, forgot to mention.
I'd like to see more info on the thread topic as well.
I'm glad NVidia are taking a look at it.
 
Nenu said:
Ah I see.
Is not syncing SLI undesirable?

op, forgot to mention.
I'd like to see more info on the thread topic as well.
I'm glad NVidia are taking a look at it.
Click to expand...

If syncing SLI. My top card is 1.212v, bottom card is 1.168v, and I crash in numerous games when the load hits 99%.

If Un Syncing and putting the top card at +30 mhz bottom card at 0(which helps fixing the issue)
My top card is 1.212v, and bottom is 1.223v and I do not crash.

This keeps my GPU core speeds @ 1329 on both cards, and my crashing in games goes away.

Now Back on topic
 
I agree with you that it shouldn't be on us to fix it, but it is fixable at least unlike this VRAM problem. I'm not sure from your post whether you're having any success with a custom job, but I see we're using similar cards so here's the BIOS I'm using, maybe it can help you with a baseline:

https://mega.co.nz/#!6UISXC4L!LR2JlSxRBaM5i_ip2zC_3CIUkgdGoQXmUdyx_U6qqB8

Note that the clocks will be lower than those listed in my sig, 1493/3800 which is a pretty safe place to start and will give you very similar volts between the two cards. Mine register at 1.268/1.275 steady at full power limit which is as close as I've come to unifying the two cards' voltage. Anything you squeeze out after that can be done through AB should you choose to. FYI, I'm using a HAF 932 with very good case cooling and the CPU is water cooled, I hit 80C at load on the top card which is, I think, acceptable using air cooling.
 
Last edited:
What good timing, my replacement 970 was installed today. Yay buggy hardware!
 
Is this affecting all 970s or just some? I thought there was another thread where someone mentioned that it's not all 970s being affected.
 
xIronCrossx said:
Is this affecting all 970s or just some? I thought there was another thread where someone mentioned that it's not all 970s being affected.
Click to expand...

There's a lot of confusion about that. It seems you need to turn off a lot of OS stuff that people usually don't turn off for the issue to show up on this particular benchmark, so some people may have mistakenly said it isn't on their 970s by not replicating the circumstances exactly.
 
Interesting but I actually haven't seen myself going over 2gb of VRAM yet on my gtx 970/1440p combo (yea, I game on Ultra but always without any AA unless it's an old/light game then I might push MSAA 2x but never more than that).

Not played Shadows of Mordor yet though, waiting for a big discount on that one. And Ryse too but not sure how demanding that game is gonna be.
 
kalston said:
Interesting but I actually haven't seen myself going over 2gb of VRAM yet on my gtx 970/1440p combo (yea, I game on Ultra but always without any AA unless it's an old/light game then I might push MSAA 2x but never more than that).

Not played Shadows of Mordor yet though, waiting for a big discount on that one. And Ryse too but not sure how demanding that game is gonna be.
Click to expand...

Pretty easy to do, Use 4xMSAA in dragon age, and it will starting using up vram.

Also Shadows of Mordor, Farcry 4, Watch dogs etc.

Not saying I have this issue, just I have seen it use alot
 
I'm not interested in Far Cry or Watch dogs or Dragon Age at all though :p

But of course I will find a way of testing it when I get back home, it's simply not an issue with the games I currently play (but I assume it WILL be an issue once I get Shadow of Mordor).
 
Last edited:
kalston said:
I'm not interested in Far Cry or Watch dogs or Dragon Age at all though :p

But of course I will find a way of testing it when I get back home, it's simply not an issue with the games I currently play (but I assume it WILL be an issue once I get Shadow of Mordor).
Click to expand...

Well it will be in the very near future :)
 
I saw this a couple of weeks ago, interesting. Hope they are able to resolve this with a BIOS update or driver.
 
Read about this possible issue a bit ago.

Haven't personally experienced this problem as everything I play (lately PlanetSide 2) seems to use 2GB or less at 1440p.

But, I'm hoping we figure out what the deal is and if there is an issue, nV comes clean and explains/fixes it.

This does make me worry about upcoming titles which will likely use and benefit from having so much VRAM.
 
Last edited:
Since I recently RMA'd my Asus 970 I am very curious how this will turn out. The rumor is that it effects all cards and there is some flex in when/where the performance tanks for each card.

My hope is a firmware or bios fix. If I start seeing version2 cards coming I'm going to be very upset if they don't swap.

I'm really glad I didn't buy that 4k monitor now...
 
I can replicate this problem as well using the benchmarks. I have 3x of the 970 (titan style cooler version). However, I can do the ram test just fine in kombuster and in shadow of mordor I can utilize over 3.5gb just fine. However, I have noticed it likes to stay around 3.5gb mark but eventually will go over into 3800ish.
 
I wonder if this explains why my PC now refuses to resume from sleep ever since upgrading to a 970.
 
This really pisses me off! First GPU upgrade for me since 2009 then this shit happens! :mad:
Looking into returning my card to Amazon right now. Guess I'll be stuck with my old 280 GTX until NVIDIA gets their shit together!
 
Jfoul said:
https://forums.geforce.com/default/...tx-970-3-5gb-vram-issue/post/4430032/#4430032
http://www.neogaf.com/forum/showthread.php?t=977645
https://www.reddit.com/r/pcgaming/comments/2s2968/gtx970_memoryvram_allocation_bug/
http://forums.guru3d.com/showthread.php?p=4991040
http://www.overclock.net/t/1535502/gtx-970s-can-only-use-3-5gb-of-4gb-vram-issue
http://www.computerbase.de/forum/showthread.php?t=1435408

Taken from Overclocker.net: (click the picture for a larger image)
Click to expand...
The bandwidth slowing down to a crawl above 3.2GB is really troubling... So it looks like this is a credible concern. Makes one wonder if NVIDIA deliberately did this to make 4k performance suffer on these cards when compared to the GTX 980.
 
I've not had this issue and I used over 3.8GB VRAM @ 1440p on AC Unity.

Is it only certain games? Certain Memory cfgs?

I have Samsung on my G1.
 
Any possible chance it's a problem with this rec.exe testing tool? It seems to be some homebrew thing and maybe a programming error?

I'm using an older driver (344.75) and it crashes before it can even finish the test. Shrug, seems possibly driver related.

Asus Strix 970:


JxZOJYA.png
 
DASHlT said:
Trust me I know this. I have been dealing with this since the begining.

As you see we have all been trying to troubleshoot this issue. http://hardforum.com/showthread.php?t=1838369 (look who started the thread 3 months ago)

My point is, users shouldn't have to fix an issue Nvidia caused.
Click to expand...

*cough*
https://forums.geforce.com/default/topic/777448/

*cough*
Posted 09/24/2014 11:32 PM

:) Custom BIOS fixes it, but it's a hindrance for overclocking on stock BIOS and pretty despicable they have yet to acknowledge it as a real issue. I actually tried to get it through in the launch interview that PCper was doing but the moderator of the chat kept insisting it was "normal behavior" :rolleyes: and wouldn't be asked to Tom Petersen. I got them launch day, too :( : http://hardforum.com/showthread.php?t=1834364
 
GoldenTiger said:
*cough*
https://forums.geforce.com/default/topic/777448/

*cough*
Posted 09/24/2014 11:32 PM

:) Custom BIOS fixes it, but it's a hindrance for overclocking on stock BIOS and pretty despicable they have yet to acknowledge it as a real issue.
Click to expand...


You tried to show fault about nvidias products on their own website and your shocked that they didn't approve of that? :rolleyes: Thats like going onto a Chevy dealership and screaming out to other customers that Fords are better.
 
Last edited:
Arct1c0n said:
You tried to show fault about nvidias products on their own website and your shocked that they didn't approve of that? :rolleyes: Thats like going onto a Chevy dealership and screaming out to other customers that Fords are better.
Click to expand...

Actually, PCper is a hardware review site, along the lines of HardOCP and Anandtech :p .

GoldenTiger said:
I actually tried to get it through in the launch interview that PCper was doing
Click to expand...

:)
 
I wonder if this is a case of the GPU being overloaded/taxed near 100% and not able to properly deal with memory transfers?
What happens if you run a less demanding game/benchmark with 8k textures, fill up the VRAM in essence and have a much lower load on the GPU?
 
Status
Not open for further replies.
Back
Top