Official Nvidia GTX 970 Discussion Thread

[LAwLz]

4. people are getting tired of the constant hate on nvidia and anyone who doesn't have an issue with their purchase.  It's o.k to hate what nvidia did (I have yet to read a post where someone says it was o.k), but not o.k to extend that opinion on to others as if they are somehow wrong because their perspective is different.

To add to this, we don't need like 10 threads about it. It was the same deal with bendgate. Everyone was making new threads left and right even though there were no new "news" about it (other than "person X says it is a disaster! Person Y calls person X a dumbass for talking about it!").

Keep everything in 1 thread.

[Guest Kloaked]

I have a 970 sitting in a customer build, so I'm going to take it out and put it in my system for science.

 

I'm going to put this shit to rest tonight.

[mr moose]

Its still proof that things run like shit when the GTX 970 goes over 3.5GB of vRAM. I actually had windows 7 unusable until I brought the vRAM usage under 3.5GB, and I'm using the classic theme.

 

We know that happens, we don't need proof, what we need is a clearer understanding of how bad and is vram the sole catalyst or does the GPU run out of HP at the same time? The settings and various other factors are not qualified in this video, and quite frankly anyone with movie maker could make a video like this and blame what ever component they wanted. 

[Dabombinable]

We know that happens, we don't need proof, what we need is a clearer understanding of how bad and is vram the sole catalyst or does the GPU run out of HP at the same time? The settings and various other factors are not qualified in this video, and quite frankly anyone with movie maker could make a video like this and blame what ever component they wanted. 

What I did was do an 640x480 2048MB memory burn with MSI Kombuster (all minimized once the burn was running), then loaded up Skyrim (it has Realvsiosn ENB and all of the recommended mods installed, and hits 1.9Gb of vRAM usage with just Skyrim running with the classic theme enabled-65MB of vRAM while idle).

[mr moose]

What I did was do an 640x480 2048MB memory burn with MSI Kombuster (all minimized once the burn was running), then loaded up Skyrim (it has Realvsiosn ENB and all of the recommended mods installed, and hits 1.9Gb of vRAM usage with just Skyrim running with the classic theme enabled-65MB of vRAM while idle).

 

did you mean to quote @Kloaked ?

[Dabombinable]

did you mean to quote @Kloaked ?

I was meaning that I was going to give a clearer understanding of how the vRAM issue can be. Heck, Skyrim at 2K DSR might be enough, because at 1080p with the same mods and settings, my GTX 650 Ti OC 2GB doesn't run out of horsepower, and a GTX 970 is a lot faster.

[Lewellyn]

I was going to say something but I didn't want to perpetuate this thread with a new person and then go over old ground.  So here's a tl:dr for those who seem to get the wrong Idea:

 

 

1.  Nvidia handed out the wrong specs concerning rops and cache to the reviewers,  there is no way to prove this was intentional and the debate has always been believe what you want, but there is ample reasoning to dispute any assumption.

 

2. No one in this thread is defending nvidia falsely representing their products,  we all hate it and no one has said otherwise.

 

3.  How major the issue with the ram is is solely in the eye of the consumer,  for many who did not look at the specs and purchased of the performance reviews it is a non issue.  for a few it is a major issue because they were hoping that based on the specs the GPU would perform better than it does now.  (again it's upto the individual to decide if this is nvidias fault or a fault of the consumer for making assumptions about future software and hardware requirements).

 

4. people are getting tired of the constant hate on nvidia and anyone who doesn't have an issue with their purchase.  It's o.k to hate what nvidia did (I have yet to read a post where someone says it was o.k), but not o.k to extend that opinion on to others as if they are somehow wrong because their perspective is different.

 

1. They (nvidia) and the brand manufacturers are still listing the card specs without any mentioning of the segmented VRAM. It doesn't matter that you can read a review on a tech website that has broken the news, the information has to be (should be) officially disclosed on the specifications page and on any retail seller's listing.

 

2. Honestly, that's bullshit. On the first pages of this thread (or similar threads) statements like "It's still a good card" and "Oh, placebo effect! It didn't just drop in performance you idiots!" are annoying straw man arguments because those aren't what the opposition has ever said. The opposition has never claimed that the card is not a good card anymore nor that the benchmarks are/were inaccurate, merely that the card was never tested for full utilization on those benchmarks and that nvidia is making bad arguments to still advertise as such to the current market. The benchmarks don't mean much if they didn't test for an issue that can arise because of a specific design. The analogies are annoying/repetitive, I know, but it would be the equivalent of stress testing a rope with a 20 pound weight and then rating it can handle 35 pounds, only to learn later that it will not handle 35 pounds. It can still handle 20 pounds, no one disagrees, but this is why I consider it a fallacious argument.

 

3. It's up to the consumer to decide whether the card is worth it, but the information has to be on the table. Consumers on forums and nvidia itself trying to pass the issue off as a non-issue by either claiming it won't affect performance (yes it can) or disingenuously stating "the card isn't powerful enough to use more VRAM anyways" is anti-consumer.

 

4. I agree.

 

edit: formatting, clarity

[jaggysnake57]

1. They (nvidia) and the brand manufacturers are still listing the card specs without any mentioning of the segmented VRAM. It doesn't matter that you can read a review on a tech website that has broken the news, the information has to be (should be) officially disclosed on the specifications page and on any retail seller's listing.

 

2. Honestly, that's bullshit. On the first pages of this thread (or similar threads) statements like "It's still a good card" and "Oh, placebo effect! It didn't just drop in performance you idiots!" are annoying straw man arguments because those aren't what the opposition has ever said. The opposition has never claimed that the card is not a good card anymore nor that the benchmarks are/were inaccurate, merely that the card was never tested for full utilization on those benchmarks and that nvidia is making bad arguments to still advertise as such to the current market. The benchmarks don't mean much if they didn't test for an issue that can arise because of a specific design. The analogies are annoying/repetitive, I know, but it would be the equivalent of stress testing a rope with a 20 pound weight and then rating it can handle 35 pounds, only to learn later that it will not handle 35 pounds. It can still handle 20 pounds, no one disagrees, but this is why I consider it a fallacious argument.

 

3. It's up to the consumer to decide whether the card is worth it, but the information has to be on the table. Consumers on forums and nvidia itself trying to pass the issue off as a non-issue by either claiming it won't affect performance (yes it can) or disingenuously stating "the card isn't powerful enough to use more VRAM anyways" is anti-consumer.

 

4. I agree.

 

edit: formatting, clarity

 

my god man let it go

[mr moose]

I was meaning that I was going to give a clearer understanding of how the vRAM issue can be. Heck, Skyrim at 2K DSR might be enough, because at 1080p with the same mods and settings, my GTX 650 Ti OC 2GB doesn't run out of horsepower, and a GTX 970 is a lot faster.

 

However it also trade blows with the 290x that isn't hampered with the vram issue indicating more than just ram could be limiting the card.  I guess it depends how you interpret this:

 

 

  NVIDIA’s performance labs continue to work away at finding examples of this occurring and the consensus seems to be something in the 4-6% range. A GTX 970 without this memory pool division would run 4-6% faster than the GTX 970s selling today in high memory utilization scenarios. Obviously this is something we can’t accurately test though – we don’t have the ability to run a GTX 970 without a disabled L2/ROP cluster like NVIDIA can

 

[Lewellyn]

my god man let it go

 

Great post, I'll take it into consideration. For now though, would you kindly not tell others what they should and shouldn't be talking about in an open forum. Instead leave it to the mods to discern what can and can't be posted within the guidelines outlined in the CoC.

[jaggysnake57]

Great post, I'll take it into consideration. For now though, would you kindly not tell others what they should and shouldn't be talking about in an open forum. Instead leave it to the mods to discern what can and can't be posted within the guidelines outlined in the CoC.

 

what how exactly did i tell you what you could or could not post?  seriously its more life advice get over it and move on

[othertomperson]

Btw here is an interesting video I came across today where someone showed the GTX 970 breaking down badly when hitting the 4 GB mark.

 

We know that happens, we don't need proof, what we need is a clearer understanding of how bad and is vram the sole catalyst or does the GPU run out of HP at the same time? The settings and various other factors are not qualified in this video, and quite frankly anyone with movie maker could make a video like this and blame what ever component they wanted. 

 

I'm pretty certain that there are other bottlenecks in this system. This isn't an issue that anyone has reported about this -- some people have been complaining about microstuttering and others have been saying that they cannot use beyond 3.5GB at all. None of these experiences agree with my own, so I would suggest the game itself may be a massive variable, but considering the sample size of games that even use this much of vram at 4K is very small there's very little concrete to go on at all.

 

And really no shit performance at 4GB is going to be poor. A 290X is not going to be giving a satisfactory experience with its frame bugger saturated either.

[mr moose]

1. They (nvidia) and the brand manufacturers are still listing the card specs without any mentioning of the segmented VRAM. It doesn't matter that you can read a review on a tech website that has broken the news, the information has to be (should be) officially disclosed on the specifications page and on any retail seller's listing.

 

2. Honestly, that's bullshit. On the first pages of this thread (or similar threads) statements like "It's still a good card" and "Oh, placebo effect! It didn't just drop in performance you idiots!" are annoying straw man arguments because those aren't what the opposition has ever said. The opposition has never claimed that the card is not a good card anymore nor that the benchmarks are/were inaccurate, merely that the card was never tested for full utilization on those benchmarks and that nvidia is making bad arguments to still advertise as such to the current market. The benchmarks don't mean much if they didn't test for an issue that can arise because of a specific design. The analogies are annoying/repetitive, I know, but it would be the equivalent of stress testing a rope with a 20 pound weight and then rating it can handle 35 pounds, only to learn later that it will not handle 35 pounds. It can still handle 20 pounds, no one disagrees, but this is why I consider it a fallacious argument.

 

3. It's up to the consumer to decide whether the card is worth it, but the information has to be on the table. Consumers on forums and nvidia itself trying to pass the issue off as a non-issue by either claiming it won't affect performance (yes it can) or disingenuously stating "the card isn't powerful enough to use more VRAM anyways" is anti-consumer.

 

4. I agree.

 

edit: formatting, clarity

1. except that the debate about the segmenting of memory is philosophical,  given many manufacturers simply choose the shortest best looking spec to promote it's not exactly out of the norm. If I am reading it right, the last 512M being separated is what allows the rest of the bus to operate at full speed, ergo some could argue that marketing it as separate cache would be just as false, however in reality, not change the actual performance of the card.

 

2. But the card is a good one and saying that it still performs exactly the same (in response to people claiming it was somehow gimped after sale) is not the same as saying people are happy with Nvidia for falsely representing their products.

 

3. If an owner with all the information now still believes it's a non issue, why is this a big problem for people to accept?  Many people didn't even know how many rops or cache it had when they bought it, so why should they be infuriated to find out a spec they weren't looking for, was never directly given should be all of a sudden incensed because it turned out to be wrong.

[mr moose]

I'm pretty certain that there are other bottlenecks in this system. This isn't an issue that anyone has reported about this -- some people have been complaining about microstuttering and others have been saying that they cannot use beyond 3.5GB at all. None of these experiences agree with my own, so I would suggest the game itself may be a massive variable, but considering the sample size of games that even use this much of vram at 4K is very small there's very little concrete to go on at all.

 

And really no shit performance at 4GB is going to be poor. A 290X is not going to be giving a satisfactory experience with its frame bugger saturated either.

 

Yes, In fact the pcper article actually states that how not all memory is accessed the same way.

 

 

It comes down to the way that memory is allocated by the operating system for applications and games. As memory is requested by a game, the operating system will allocate portions for it depending on many factors. These include the exact data space that the game asked for, what the OS has available and what the heuristic patterns of the software models deem at the time. Not all memory is accessed in the same way, even for PC games.

 

and

 

If a game has allocated 3GB of graphics memory it might be using only 500MB of a regular basis with much of the rest only there for periodic, on-demand use. Things like compressed textures that are not as time sensitive as other material require much less bandwidth and can be moved around to other memory locations with less performance penalty. Not all allocated graphics memory is the same and innevitably there are large sections of this storage that is reserved but rarely used at any given point in time.

 

 

I read this to basically say,  depending on the game, textures, OS, etc the results will be different, one game might work flawlessly right upto 3.8G while another chugs the second it hits 3.5G.  It just depends whether that last 512M contains speed sensitive data or not.

[zappian]

If this is common place and happens on all cards at 3.5G why didn't the reviewers mention it and why are there people showing the card going over 3.5 with much less stuttering?

 

Is it possible this is a rare or setup up video? There are no onscreen displays to verify the settings and FPS etc. 

 

Not trying to fuel the anti-debate, but I would like to see people be a little more critical of videos and arguments that appear designed to antagonize the situation.

 

Basically if games like shadow of mordor existed at the time to push VRAM this issue would be known MUCH earlier.

[othertomperson]

Basically if games like shadow of mordor existed at the time to push VRAM this issue would be known MUCH earlier.

 

...but it did. That game has been out for a while now and people have been playing and benchmarking 970s with it.

[Dabombinable]

However it also trade blows with the 290x that isn't hampered with the vram issue indicating more than just ram could be limiting the card.  I guess it depends how you interpret this:

 

I interpret it as "Nvidia should have just kept 3GB of vRAM, just like the 780Ti that it replaces".

[s3ns3]

Basically if games like shadow of mordor existed at the time to push VRAM this issue would be known MUCH earlier.

what issue? the card is fine and runs great. the issue is that nvidia wasnt transparent about the specs . gtx 970 is guilt free and its a very good gpu at a very good price

 

http://youtu.be/dwiuD1WtnNs?t=4m20s

[Lewellyn]

1. except that the debate about the segmenting of memory is philosophical,  given many manufacturers simply choose the shortest best looking spec to promote it's not exactly out of the norm. If I am reading it right, the last 512M being separated is what allows the rest of the bus to operate at full speed, ergo some could argue that marketing it as separate cache would be just as false, however in reality, not change the actual performance of the card.

 

2. But the card is a good one and saying that it still performs exactly the same (in response to people claiming it was somehow gimped after sale) is not the same as saying people are happy with Nvidia for falsely representing their products.

 

3. If an owner with all the information now still believes it's a non issue, why is this a big problem for people to accept?  Many people didn't even know how many rops or cache it had when they bought it, so why should they be infuriated to find out a spec they weren't looking for, was never directly given should be all of a sudden incensed because it turned out to be wrong.

 

I thought I was going to be giving you the benefit of the doubt, but now I see that it's no use. The arguments I have already rebutted are being repeated with the same casual zeal as if they have any merit.

 

"except that the debate about the segmenting of memory is philosophical" 

 

No it isn't. The design is admitted to by nvidia. The 'debate' is non-existent, the memory is segmented and has a lower bandwidth across the threshold.

 

"Many people didn't even know how many rops or cache it had when they bought it, so why should they be infuriated to find out a spec they weren't looking for, was never directly given should be all of a sudden incensed because it turned out to be wrong." 

 

The customer expects the entirety of VRAM to run at the same speed, unless told otherwise. Nvidia knows that the customer expects this and still continues to market the card as 4GB without a disclosure that the VRAM operates at two different speeds. Whether or not it will affect performance in the majority of scenarios is irrelevant and whether or not the customer actually needs to know the specifics of what causes the VRAM to be segmented is irrelevant. Continuing to market the card without disclosure is illegal. It is called false advertisement. Refunds, while ethically debatable, are not guaranteed by law due to inability to prove intentional deceit. An error in communication gives them a legal slide in providing compensation for those who have already purchased, however it does not allow a corporation to continue a false advertising campaign.

 

If someone wishes to buy the card now with the information, that's their prerogative. I'll repeat, "Consumers on forums and nvidia itself trying to pass the issue off as a non-issue by either claiming it won't affect performance (yes it can) or disingenuously stating "the card isn't powerful enough to use more VRAM anyways" is anti-consumer." Again, it is irrelevant whether or not the consumer didn't have or look for the information on the inner workings of the cache and ROPS, it is the end user effects that can happen because of it that are required. The specifics of why the rope can't handle the load are not necessary, only the information that the rope can break.

[mr moose]

I thought I was going to be giving you the benefit of the doubt, but now I see that it's no use. The arguments I have already rebutted are being repeated with the same casual zeal as if they have any merit.

 

"except that the debate about the segmenting of memory is philosophical" 

 

No it isn't. The design is admitted to by nvidia. The 'debate' is non-existent, the memory is segmented and has a lower bandwidth across the threshold.

 

"Many people didn't even know how many rops or cache it had when they bought it, so why should they be infuriated to find out a spec they weren't looking for, was never directly given should be all of a sudden incensed because it turned out to be wrong." 

 

The customer expects the entirety of VRAM to run at the same speed, unless told otherwise. Nvidia knows that the customer expects this and still continues to market the card as 4GB without a disclosure that the VRAM operates at two different speeds. Whether or not it will affect performance in the majority of scenarios is irrelevant and whether or not the customer actually needs to know the specifics of what causes the VRAM to be segmented is irrelevant. Continuing to market the card without disclosure is illegal. It is called false advertisement. Refunds, while ethically debatable, are not guaranteed by law due to inability to prove intentional deceit. An error in communication gives them a legal slide in providing compensation for those who have already purchased, however it does not allow a corporation to continue a false advertising campaign.

 

If someone wishes to buy the card now with the information, that's their prerogative. I'll repeat, "Consumers on forums and nvidia itself trying to pass the issue off as a non-issue by either claiming it won't affect performance (yes it can) or disingenuously stating "the card isn't powerful enough to use more VRAM anyways" is anti-consumer." Again, it is irrelevant whether or not the consumer didn't have or look for the information on the inner workings of the cache and ROPS, it is the end user effects that can happen because of it. The specifics of why the rope can't handle the load are not necessary, only the information that the rope can break.

And that's your opinion.

[Lewellyn]

And that's your opinion.

 

That's an interesting and entirely thought provoking response. Any other 5 word responses you want to give? Truly truly insightful. So fresh too!

[mr moose]

That's an interesting and entirely thought provoking response. Any other 5 word responses you want to give? Truly truly insightful. So fresh too!

 

4 words, not 5.. (technically 3.5 words, as the 4th word is a contraction and not whole words )   And if I can dismiss your entire argument by simply pointing out what many of us have been saying from the very start then you need to consider what that means.

 

We don't need walls of text for personal opinions, we need people to stop posting the same shit not understanding that it's their opinion and their opinion only. 

[Rohith_Kumar_Sp]

@ThatGuyWhoTwirlsHisPen there are facts and there are opinions , no one is arguing about the facts with you , we all know what nvidia did was uncalled for and just wrong , no is saying it was the right thing to do , but you are fighting against people's opinions. honestly if it were me, be the wise guy, let people have their opinions whether they want to support nvidia in the future or not , you don't have to force them by all the analogies and arguments. just me 2 cents, have a nice day.

[Lewellyn]

4 words, not 5..   And if I can dismiss your entire argument by simply pointing out what many of us have been saying from the very start then you need to consider what that means.

 

We don't need walls of text for personal opinions, we need people to stop posting the same shit not understanding that it's their opinion and their opinion only. 

 

And that's your opinion.

 

You don't understand how utterly dismissive that is and I doubt you ever will.

 

I count 5 excluding contractions while you count 4, and that is your opinion. Don't debate it, just a difference in opinion.

[mr moose]

And that's your opinion.

 

You don't understand how utterly dismissive that is and I doubt you ever will.

 

I count 5 excluding contractions while you count 4, and that is your opinion. Don't debate it, just a difference in opinion.

 

That's just it, I am not dismissing facts and I am allowing people to have their opinions.  which is the point many seem to have missed.  we know what Nvidia did, whether that is bad or not is upto the individual.  You cannot dismiss someone because in their personal opinion and experience they are not upset by what happened.  There are people in this thread literally calling anyone who likes the 970 an idiot, a moron for supporting a disgusting lying company like nvidia.