The 970 3.5gb bug

[Cheddle]

so all of a sudden, a few months after launch... This bug is found, showing that the first 3.5 gb of ram on the 970 is accessed differently than the last 500mb.

 

before this bug was found, every man and his feking dog reckoned the 970 was the best dam video card money could buy. And those that bought them LOVED them.

 

Now, after several months of enjoying these cards, this bug is found and all of a sudden everyone hates them, claims Nvidia are terrible and that this is unacceptable.

 

what... the... fek...

 

Why is it that something that has effected everyone since day one, and not been a problem, is all of a sudden a huge reason to hate on the 970 and Nvidia in general as soon as its discovered? nothing has changed, the card is still as it was when it came out...

 

the 970 took an arrow to the knee, no one wants to adventure with it any more.

[zappian]

Because nvidia hide it from the consumer.

 

Gabe newell will explain:

 

[Imakuni]

Why is it that something that has effected everyone since day one, and not been a problem, is all of a sudden a huge reason to hate on the 970 and Nvidia in general as soon as its discovered?

Because the Net is full of AMD fanboys, lurking around the good stuff.

[Razzaa]

so all of a sudden, a few months after launch... This bug is found, showing that the first 3.5 gb of ram on the 970 is accessed differently than the last 500mb.

 

before this bug was found, every man and his feking dog reckoned the 970 was the best dam video card money could buy. And those that bought them LOVED them.

 

Now, after several months of enjoying these cards, this bug is found and all of a sudden everyone hates them, claims Nvidia are terrible and that this is unacceptable.

 

what... the... fek...

 

Why is it that something that has effected everyone since day one, and not been a problem, is all of a sudden a huge reason to hate on the 970 and Nvidia in general as soon as its discovered? nothing has changed, the card is still as it was when it came out...

 

the 970 took an arrow to the knee, no one wants to adventure with it any more.

because everyone likes to scream that the sky is falling when most people are ignorant and dont even understand the issue to begin with.

[Tim Drake]

If I hear one more person say to the knee when it's IN the knee I am gunna flip..

 

But Nvidia already explained how it works? In games where that much Vram is needed, my gpu will use more than 3.5GB.

 

This is due to people being misinformed.

[Guest]

so all of a sudden, a few months after launch... This bug is found, showing that the first 3.5 gb of ram on the 970 is accessed differently than the last 500mb.

 

before this bug was found, every man and his feking dog reckoned the 970 was the best dam video card money could buy. And those that bought them LOVED them.

 

Now, after several months of enjoying these cards, this bug is found and all of a sudden everyone hates them, claims Nvidia are terrible and that this is unacceptable.

 

what... the... fek...

 

Why is it that something that has effected everyone since day one, and not been a problem, is all of a sudden a huge reason to hate on the 970 and Nvidia in general as soon as its discovered? nothing has changed, the card is still as it was when it came out...

 

the 970 took an arrow to the knee, no one wants to adventure with it any more.

 

Are you serious? It's a problem that doesn't affect everyone right now, yes. But it is a problem that will affect a great deal of people eventually and it has already affected many. Why are people pissed? Because Nvidia effectively dicked them over and made more than 10% of their available VRAM useless. What about SLI owners? What about games that simply need a ton of VRAM or other VRAM intensive tasks? This is a huge deal whether you like it or not, and the people affected have every right to be as pissed off as they are right now. It's false advertisement, it's dishonest and it's simply not okay to tell the affected people that it's okay because they don't notice it "yet". No, stop it.

[zappian]

You people are strawmans.

 

There are tons of people affected by the issue , just because your 970 is fine doesnt mean crap.

 

Read the countless reports of stuttering over 3.5 GB vram on guru 3d ,this forum and the nvidia board.

 

There is an issue , nvidia didnt disclose it and now people are rightfully mad.

 

There.

[Guest]

You people are strawmans.

 

There are tons of people affected by the issue , just because your 970 is fine doesnt mean crap.

 

Read the countless reports of stuttering over 3.5 GB vram on guru 3d ,this forum and the nvidia board.

 

There is an issue , nvidia didnt disclose it and now people are rightfully mad.

 

There.

Except that hitching/stuttering issues have hardly been confirmed to be caused by accessing the final pool of vram on the 970. Everything I've seen people posting that 'confirms' this issue has either been this 'nai benchmark', or a video of people running modern games at obscene settings, like 4k + high levels of msaa. no **** that isn't going to perform well on a 300 dollar single gpu. Ramping up settings like that you are going to hit other bottlenecks in the gpu far before you hit any mythical 'memory bug'.

People have basically started attributing any kind of performance issue they have with the 970 to this 'bug', with little to no conclusive evidence, except a 'benchmark' that's known to be completely flawed.

 

check this out, these are posts from the creator of the nai benchmark:

 

https://translate.google.com/translate?depth=1&hl=en&rurl=translate.google.com&sl=de&tl=en&u=http://www.computerbase.de/forum/showthread.php%3Ft%3D1435408%26p%3D16917097#post16917097

 

and https://translate.google.com/translate?hl=en&sl=de&tl=en&u=http%3A%2F%2Fwww.computerbase.de%2Fforum%2Fshowthread.php%3Ft%3D1435408%26page%3D7%26p%3D16912375%23post16912375&sandbox=1

 

The benchmark measures "actually" not the DRAM bandwidth but the bandwidth of the "global memory" s in CUDA. The global memory in CUDA is a virtual memory space, which includes not only the DRAM, the GPU but also DRAM areas of the CPU. A virtual memory space is always distinguished by the fact that there is virtual memory addresses. The virtual memory addresses are translated to a memory access * somehow * in the actual physical memory addresses of the DRAM. The advantage of such virtual memory space, among other things, that it allows swapping of memory regions. Just such a retrieval at the DRAM, the CPU seems CUDA here to make the slump in the benchmark.

The benchmark allocates as many blocks of the global DRAM memories within the GPU as possible. Then it measures the read bandwidth within each of these blocks. The measurement is only as long as the allocated blocks really are in the DRAM GPU good. In this case, memory bandwidth is measured relatively accurately.

The problem however, is that the benchmark is not fully owns the DRAM, the GPU itself. After running in the background Windows and various programs, all also claim some of the DRAM of the GPU. However, the virtual memory space guarantees the benchmark that it () except for the area of ​​the primary surface (?) May be used for almost the entire 4 GiByte. Therefore, the allocation of non-DRAM proposes initially not be out until the upper limit of about 4 GiByte. However, the GPU must now begin to swap memory areas. That is, when the CUDA application is running, should in the DRAM of the GPU, the corresponding data of the CUDA application. Does the character process of Windows or another application should in DRAM GPU their data is.

The GPU seems swapping, according to my first simple investigations of this benchmark is to make simple associative. That is, at a physikalsichen address in the GPU DRAM always the same CUDA data or DirectX data must reside. This leads to conflicts that manifest themselves in the collapse of the bandwidth of the global memories in such conflict areas.

The benchmark attempts to reduce this effect by repeatedly requesting the data in each storage area successively. That is the first demand "should" cause the benchmark a page fault. The page fault "should" copy the data on the GPU Page from the DRAM from the CPU to the GPU DRAM. Then the other global memory accesses would be carried out with the DRAM bandwidth. So at least the assumption on my part.

Interesting way, the GPU does not behave as expected of me. This makes the GPU, the corresponding data does not seem to upload into the DRAM of the GPU, but again to request directly from the DRAM CPU with each memory access in a page fault in CUDA. Characterized the benchmark measures total in such cases, more or less the swapping behavior of CUDA, and not the DRAM bandwidth.

The whole can verify by allowing any applications running in the background that consume a lot of DRAM from the GPU, allowing more swapping needs to take place is also easy. In this case, the benchmark collapses also.

 

Since my last post is kind of lost here before and another 100 teams post their results briefly for all in capital and Red:

 

 

My benchmark does not say what you believe it says. It is * NOT * to detect the presence of a VRAM error.

 

[zappian]

Except that hitching/stuttering issues have hardly been confirmed to be caused by accessing the final pool of vram on the 970. Everything I've seen people posting that 'confirms' this issue has either been this 'nai benchmark', or a video of people running modern games at obscene settings, like 4k + high levels of msaa. no **** that isn't going to perform well on a 300 dollar single gpu. Ramping up settings like that you are going to hit other bottlenecks in the gpu far before you hit any mythical 'memory bug'.

People have basically started attributing any kind of performance issue they have with the 970 to this 'bug', with little to no conclusive evidence, except a 'benchmark' that's known to be completely flawed.

 

check this out, this is a post from the creator of the nai benchmark:

 

http://www.computerbase.de/forum/showthread.php?t=1435408&page=13&p=16917097

 

I dont give much validity to the nai benchmark , im talking about user reports of stuttering when 

the vram usage goes over 3.5 GB in games.

 

The kind of fps drops and stutter that are usually very sudden and associated with high vram usage.

 

Nvidia itself said that they had two memory modules one .3.5 GB one and a quite slower 0,5 GB one thats low priority.

Thats the one that causes the stutter.

 

They failed to reveal this untill they were called out which is unacceptable.

 

Im quite well informed in this issue and to say nvidia did no wrong doing here is just something i wont agree.

[boboman342]

Is there no such program that will use a user defined amount of your gpu's ram without effecting its fps etc? For running the same game/settings with regular ram usage and synthetic ram usage?

[R_Golden]

The simple answer to this question is, people are dumb and like to bitch, its in their nature. Also... Fanboys.

[Guest]

I dont give much validity to the nai benchmark , im talking about user reports of stuttering when 

the vram usage goes over 3.5 GB in games.

 

The kind of fps drops and stutter that are usually very sudden and associated with high vram usage.

 

Nvidia itself said that they had two memory modules one .3.5 GB one and a quite slower 0,5 GB one thats low priority.

Thats the one that causes the stutter.

 

They failed to reveal this untill they were called out which is unacceptable.

 

Im quite well informed in this issue and to say nvidia did no wrong doing here is just something i wont agree.

and as I mentioned, most of these reports of 'stuttering past 3.5gb', are people running games with extremely demanding settings, like copious amounts of downsampling + msaa, intentionally trying to force the game to allocate more than 3.5gb of vram.

 

By doing this they are basically giving themselves a self fulfilling prophecy by using extremely intensive game settings to force the gpu to allocate more than 3.5gb of vram. You will bottleneck the entire gpu, which pretty much makes the entire test utterly inconclusive, because there's no way to tell if the stuttering is actually caused by accessing the last 500 mb of vram, or the more likely explanation of 'you are bottlenecking your entire gpu by running absurd settings'.

[zappian]

and as I mentioned, most of these reports of 'stuttering past 3.5gb', are people running games with extremely demanding settings, like copious amounts of downsampling + msaa. You are basically giving yourself a self fulfilling prophecy by using extremely intensive game settings to force the gpu to allocate more than 3.5gb of vram. You will bottleneck the entire gpu, which pretty much makes the entire test utterly inconclusive, because there's no way to tell if the stuttering is actually caused by accessing the last 500 mb of vram, or the more likely explanation of 'you are bottlenecking your entire gpu by running absurd settings'.

 

Nvidia already said that the last 0.5 GB were slower.

 

So its highly likely that the problem considering they confirmed it in their OFFICIAL statement.

 

No use in denying it when the company that makes the chip in the first place has confirmed it .

[Guest]

Nvidia already said that the last 0.5 GB were slower.

 

So its highly likely that the problem considering they confirmed it in their OFFICIAL statement.

 

No use in denying it when the company that makes the chip in the first place has confirmed it .

Nvidia in no way 'confirmed' that it causes stuttering issues, you are completely misrepresenting their statement.

[zappian]

Nvidia in no way 'confirmed' that it causes stuttering issues, you are completely misrepresenting their statement.

 

It confirmed it slower and lower priority.

 

Thus it is VERY likely than its the cause.

 

Just apply some basic logic here.

[Imakuni]

Nvidia already said that the last 0.5 GB were slower.

 

So its highly likely that the problem considering they confirmed it in their OFFICIAL statement.

 

No use in denying it when the company that makes the chip in the first place has confirmed it .

They also said that this was only 1~3% performance loss, compared to a card not affected by the issue. And the fact that it cost MUUCH less.

[Justhavocman]

The simple answer to this question is, people are dumb and like to bitch, its in their nature. Also... Fanboys.

So heres a ferari for 20k $ , but it was one mirror , oh and no windows , oh and it can only reach 100Mph , oh why are you complaining? You paid 20k for a ferari so shut up

Pretty much sums up every arguement about it not being a big deal ,

IT IS A BIG DEAL.

[zappian]

They also said that this was only 1~3% performance loss, compared to a card not affected by the issue. And the fact that it cost MUUCH less.

 

They didnt account frametime variance.

 

Or minimum fps caused by VRAM stutter , they cherry picked what they showed.

[zappian]

So heres a ferari for 20k $ , but it was one mirror , oh and no windows , oh and it can only reach 100Mph , oh why are you complaining? You paid 20k for a ferari so shut up

Pretty much sums up every arguement about it not being a big deal ,

IT IS A BIG DEAL.

 

This.

 

Only people that dont have a 970 can be objective in this issue.

Otherwise we have a bunch of morons justifying their purchase.

[Imakuni]

They didnt account frametime variance.

 

Or minimum fps caused by VRAM stutter , they cherry picked what they showed.

Okay. Grab a 980 and a 970, and test it yourself. Show your results, as well as your testing methodology.

 

Until that happens, though, the only test we have is Nvidia's. Anything else is just rant.

[zappian]

Okay. Grab a 980 and a 970, and test it yourself. Show your results, as well as your testing methodology.

 

Until that happens, though, the only test we have is Nvidia's.

 

Their testing is flawed and they know it.

 

If they had shown min-mum fps you would understand what i mean.

 

This doesnt affect me at all im just being objective ,go buy 4 970s for all i care.

[Justhavocman]

Okay. Grab a 980 and a 970, and test it yourself. Show your results, as well as your testing methodology.

Until that happens, though, the only test we have is Nvidia's. Anything else is just rant.

You cant be this ignorant..

[Guest]

They didnt account frametime variance.

 

Or minimum fps caused by VRAM stutter , they cherry picked what they showed.

Alright, so where's your definitive proof that frametime variance issues are caused by accessing that vram? How would one even benchmark that accurately to get results that aren't completely inconclusive? Because I can tell you one thing, the current benchmarking methods of 'load a game up with tons of downsampling and msaa to force it to allocate 4gb' are definitely not a conclusive way to expose a specific bottleneck, because you will certainly run into other bottlenecks too.

[Briggsy]

Its not like this is the first time nvidia have done this. Anyone who owns a 2GB 660/660 Ti, or a 1GB 550 Ti were in the same boat, and it was quickly brushed under the carpet and dismissed as the buyers fault for not knowing.

 

What makes the 970 a special case worth being upset about, is that no one expected a chunk of ram to be gimped on a 256 bus with a logical amount of Ram. if the 970 had a 192bit bus instead, everyone would have known immediately that 4GB was going to have issues. but with 256 its disturbing and makes people think twice about trusting nvidia at their word.

[Imakuni]

You cant be this ignorant..

Oh, sorry, I didn't see you had made tests yourself.

 

So, where are them?