GTX 970 flaw

Status
Not open for further replies.
I honestly can't believe there are people OK with this on these forums and actually think it's not an issue. It's like if a car manufacturer advertised this car seats 6, but then doesn't mention that there are only enough seat belts for 5. Saying, "sorry you shouldn't have that 6th person in there because there aren't enough seat belts", after I bought it already, does not absolve them for falsely giving me the wrong information. Seriously the whole point of the issue is Nvidia took away every buyer's ability to choose wisely when they gave the Official Specs.

What annoys me most is, my ability to step up to a 980 passed. If I had known about this issue 4 weeks ago I would have moved to the 980. Nvidia effectively blocked me from being able to do this by withholding pertinent information which they apparently have known about since the release of the card.
 
When I play Shadow of Mordor at 1080p (no 4k dsr etc), I run fine for 20 minutes with 70-90 fps and the vram hovering around 3.5.

Then, the vram creeps up a little as I play and goes into the 3.6 - 3.8 range. I begin the notice lag, stutters and hitching. I look at the on screen fps and see that around 80 fps is still being shown.

When I am 3.5GB or lower, smooth gameplay. More than 3.5GB, same fps but stutters.

I have the same vram/fps/stutters relationship in other games too.

You're playing it wrong ;)
 
:eek: +1

I sincerely hope that this problem is fixed by nVidia. I was hoping to replace my aging 290 with the 970 at a future date. When someone pays a premium for a new GPU, you paid for 4GB, not 3.5 & change. Another example of satiating the share holder before consumer. This issue should not be a red vs. green thing. It's about company vs. consumer.

Do try to stay away from being an early adapter. It's just not worth it...hardware and software.

and for the AMD fanbois out there...ne pas stirrez le shitpot. No product from any company is worth the rage from your fellow man.
 
I'm giggling..

Can't imagine why someone would be giggling about other users having a potential issue here :confused:?

I want to see some FCAT results for sli setups on this issue... that'll put it to rest as to whether it's anything to care about at all or not, really.
 
Can't imagine why someone would be giggling about other users having a potential issue here :confused:?

I want to see some FCAT results for sli setups on this issue... that'll put it to rest as to whether it's anything to care about at all or not, really.

Please lighten up. I'm amused by the [H]umor but not any more happy about the possible issue than you are.
 
Fortunately it looks like there are some sites that are doing some analysis and better benchmarks to try to discern any impacts, so we should know some things a lot better by, say, the end of the week.
 
Can't imagine why someone would be giggling about other users having a potential issue here :confused:?

I want to see some FCAT results for sli setups on this issue... that'll put it to rest as to whether it's anything to care about at all or not, really.

My 970 runs every game I throw at it without issue.
 
Ok ;)... if I cover your next coffee at your namesake place, will you forgive me? :p :D

starbuck1.jpg


wut
 
I find nvidia's latest response to be unsatisfactory.

I'll be eagerly awaiting the third-party benchmarks that are incoming. Not that I'm likely buying this generation anyway.
 
This is looking really bad for nVidia...now they are outright saying the specs are different? LOL.

Doesn't make me want/need to sell my 970 but goddamn...get your shit together.
 
If nvidia have lied about the amount of ROPS etc. then surely that is illegal and people who bought a 970 should be getting a partial refund or similar... Still happy with my 970 but you cannot say "this has 64 rops" if it actually has 56... That is false advertising..
 
Anandtech (Ryan Smith) has an article about it.

In short this is probably why it gets noticeable (stutters?).
GTX 970 can read the 3.5GB segment at 196GB/sec (7GHz * 7 ports * 32-bits), or it can read the 512MB segment at 28GB/sec, but not both at once; it is a true XOR situation. Furthermore because the 512MB segment cannot be read at the same time as the 3.5GB segment, reading this segment blocks accessing the 3.5GB segment for that cycle, further reducing the effective memory bandwidth of the card. The larger the percentage of the time the crossbar is reading the 512MB segment, the lower the effective memory bandwidth from the 3.5GB segment.

Check out the article, it's pretty informative at least compared to PCPer's NV statement.

http://www.anandtech.com/show/8935/geforce-gtx-970-correcting-the-specs-exploring-memory-allocation
 
Anandtech (Ryan Smith) has an article about it.

In short this is probably why it gets noticeable (stutters?).


Check out the article, it's pretty informative at least compared to PCPer's NV statement.

http://www.anandtech.com/show/8935/geforce-gtx-970-correcting-the-specs-exploring-memory-allocation

It wouldn't cause stutters. You access data from that memory intraframe, each frame. Therefore the worst it can do is slow each frame equally. Whatever is in there is being accessed the same way each frame, or not at all. No opportunity for stutter since it's not as if one frame is accessing and the next frame not accessing.

Whoever is stuttering has an unrelated issue.
 
Last edited:
So the card only has 56 ROPS and less cache than the GTX 980? Sorry if old news but I thought they were the same.

http://www.pcper.com/reviews/Graphi...Full-Memory-Structure-and-Limitations-GTX-970
I assume you made a typo as it says 52 ROPs not 56. Techreport had pointed that out months ago but nothing was mentioned about the 3.5 GB issue at that time.

EDIT: Hmm Anandtech is saying 56 ROPs...

EDIT 2: It is actually 56 ROPs and pcper has now edited the specs to reflect that.
 
Last edited:
I'm dismayed with this site. I have watched this once great resource turn in to the BBQ Pit Boys of hardware enthusiast sites.

You are too busy making anti-Apple click bait news items and catering to redneck pc gamers, than standing by your loyal readers. You should be giving Nvidia both barrels, just like you used to do when you still cared about what you do. I bought my 970 after reading the review on this site, and since then I've noticed the complete lack of any real pc industry news and views.

I'm out of here. Thanks for all the fish
 
ummm...

what exactly are Redneck PC Gamers?

is there a difference between them and us Yankee PC Gamers?
 
I'm dismayed with this site. I have watched this once great resource turn in to the BBQ Pit Boys of hardware enthusiast sites.

You are too busy making anti-Apple click bait news items and catering to redneck pc gamers, than standing by your loyal readers. You should be giving Nvidia both barrels, just like you used to do when you still cared about what you do. I bought my 970 after reading the review on this site, and since then I've noticed the complete lack of any real pc industry news and views.

I'm out of here. Thanks for all the fish

g7mb.jpg
 
I have absolutely no cares about this issue. My Nvidia GPU would have to rape my wife while pouring sugar in my gas tank before I would go to AMD again.
 
It wouldn't cause stutters. You access data from that memory intraframe, each frame. Therefore the worst it can do is slow each frame equally. Whatever is in there is being accessed the same way each frame, or not at all. No opportunity for stutter since it's not as if one frame is accessing and the next frame not accessing.

Whoever is stuttering has an unrelated issue.

Many people getting stuttering have been forcing settings that would make even a GTX 980 stutter. I saw guys putting a modded skyrim at 8x MSAA and watching it drop to 15fps at a specific part of a town and saying "see! it stutters!". I'm sorry, but any card would stutter there. The engine itself has some issues with stuttering.

I bought my 970s because of the price bracket and the performance I expected from them. I did not lose performance because someone feels slighted about specs. I didn't look at ROPs or SMMs etc. I expect most don't either when making a decision, they probably jump right to the performance numbers. I looked at how it performs at my resolution in games I care about and it is fine on that end.
 
I feel sorry for the folks that are experiencing problems. (game attempting to use >3.5GB) But I feel there is a bit of over reacting going on. Nvidia needs to do something to make this right. But I do not feel these cards are now suddenly junk and unusable. I must admit to being glad I am still with my GTX 780 for now.....

Kid
 
Hell, I still don't don't know what a SMM or ROP really is or does

They don't matter. When you see forum posters outraged about disabled SMMs or the number of ROPs, that's your cue that they don't understand the problem and are just blurting out buzzwords:

"Before people complain about the ROP count difference as a performance bottleneck, keep in mind that the 13 SMMs in the GTX 970 can only output 52 pixels/clock and the seven segments of 8 ROPs each (56 total) can handle 56 pixels/clock. The SMMs are the bottleneck, not the ROPs."

"Second to that, it turns out the disabled SMMs have nothing to do with the performance issues experienced or the memory system complications."

http://www.pcper.com/reviews/Graphi...Full-Memory-Structure-and-Limitations-GTX-970

Accusations that Nvidia lied about the number of ROPs is a red herring. If the 970 had 64 ROPs like the 980 does, it wouldn't change anything about this issue; it would still be bottlenecked by the disabled SMMs.
 
They don't matter. When you see forum posters outraged about disabled SMMs or the number of ROPs, that's your cue that they don't understand the problem and are just blurting out buzzwords:

"Before people complain about the ROP count difference as a performance bottleneck, keep in mind that the 13 SMMs in the GTX 970 can only output 52 pixels/clock and the seven segments of 8 ROPs each (56 total) can handle 56 pixels/clock. The SMMs are the bottleneck, not the ROPs."

"Second to that, it turns out the disabled SMMs have nothing to do with the performance issues experienced or the memory system complications."

http://www.pcper.com/reviews/Graphi...Full-Memory-Structure-and-Limitations-GTX-970

Accusations that Nvidia lied about the number of ROPs is a red herring. If the 970 had 64 ROPs like the 980 does, it wouldn't change anything about this issue; it would still be bottlenecked by the disabled SMMs.

how many green NVidia dollars do you get paid for posts like this?
 
They don't matter. When you see forum posters outraged about disabled SMMs or the number of ROPs, that's your cue that they don't understand the problem and are just blurting out buzzwords:

"Before people complain about the ROP count difference as a performance bottleneck, keep in mind that the 13 SMMs in the GTX 970 can only output 52 pixels/clock and the seven segments of 8 ROPs each (56 total) can handle 56 pixels/clock. The SMMs are the bottleneck, not the ROPs."

"Second to that, it turns out the disabled SMMs have nothing to do with the performance issues experienced or the memory system complications."

http://www.pcper.com/reviews/Graphi...Full-Memory-Structure-and-Limitations-GTX-970

Accusations that Nvidia lied about the number of ROPs is a red herring. If the 970 had 64 ROPs like the 980 does, it wouldn't change anything about this issue; it would still be bottlenecked by the disabled SMMs.

Only because NOW we know that the ROPS/L2/MC are not intrinsically linked to each other. Nvidia's review documentation specifically stated they were.

Edit- Whoever stated that about the SMMs feeding the ROPs... well they are wrong.
 
Last edited:
Status
Not open for further replies.
Back
Top