The 970 3.5gb bug

[maizenblue]

I bought a 970 a few weeks ago and I've said all along that I don't see what all the hype was about, it just seems like a slightly cheaper gtx 780 to me. Now that I find out a huge chunk of my Vram is useless im not exactly thrilled. I was on the fence between a 970 and an r9 290x before all this, so I definitely would have went with the 290x had I known Nvidia was selling a neutered card.

 

Since I bought a reference 970 sold directly by Nvidia, im tempted to RMA the card, telling them there seems to be a problem with my cards VRAM and demanding they fix it.

 

The coverup is usually what gets people into trouble. It'll be interesting seeing Nvidia try to explain away VRAMGATE

[Whorax]

This really pisses me off actually. I didn't just spend money on SLI 970s to get fucked over...

[GrimNeo]

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.

Can you please provide proof that their testing is flawed. Others everything else you say I will have to ignore. Also can you please explain why you are even arguing if it does not affect you.

 

You cant be this ignorant..

Can you please provide proof.

 

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.

Exactly. This guy gets it.

 

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

 

So, where are them?

Exactly. No proof was made from the previous person.

[boboman342]

Nvidia confirms that the ram on the reference gtx 970 is segmented into a 208 bit bus for the 3.5GB and another 48bit bus for the 0.5GB. 

 

The gtx 980 does not suffer from any "issues" related to using over 3.5GB of vram, and nor do some after market gtx 970s. 

 

Is the ram on the custom pcb gtx 970s or reference gtx 980 non-segmented? Is all the ram on the same 256 bit bus?

 

These are the questions we need answers to!

[Guest]

Nvidia confirms that the ram on the reference gtx 970 is segmented into a 208 bit bus for the 3.5GB and another 48bit bus for the 0.5GB. 

 

The gtx 980 does not suffer from any "issues" related to using over 3.5GB of vram, and nor do some after market gtx 970s. 

 

Is the ram on the custom pcb gtx 970s or reference gtx 980 non-segmented? Is all the ram on the same 256 bit bus?

 

These are the questions we need answers to!

The vram on the 970 is NOT physically segmented, its all the same speed gddr5 and its all on the same 256-bit bus. The vram is 'logically' segmented at the firmware and/or driver level.

[boboman342]

What about the gtx 560 and gtx 660 then? Both have a labeled ram spec, but it is a (somewhat) lie.

[Guest]

What about the gtx 560 and gtx 660 then? Both have a labeled ram spec, but it is a (somewhat) lie.

the 660 did have an asymmetrical design with two separate buses

[Dabombinable]

Honestly, the memory bug never affects me as non of the games I play seem to use more than slightly over 2GB. And since I've gone from 256MB, 2GB and then to 4GB in the space of 12 months, I don't really care as the 970 as a massive improvement over my 650ti (which is light years ahead of the Quadro NVS 110M graphics that were in my old laptop-2006 graphics chip that can't play most games from 2001 and later).

[AmbellinaThePrise]

So.... My question here is, I have a build i'm throwing together money for right now... i7 4790k being the cpu, so i'm in no way worried about my processor causing any issues when gaming at all.

 

But should I actually avoid the 970 because of this and just go for a 290x? I was committed to buying a gsync enabled monitor, and everything seemed to be falling into place until hearing this. What do you all think?

[boboman342]

the 660 did have an asymmetrical design with two separate buses

From what Nvidia said: [spoiler=Nvidia Quote] The GeForce GTX 970 is equipped with 4GB of dedicated graphics memory. However the 970 has a different configuration of SMs than the 980, and fewer crossbar resources to the memory system. To optimally manage memory traffic in this configuration, we segment graphics memory into a 3.5GB section and a 0.5GB section. The GPU has higher priority access to the 3.5GB section. When a game needs less than 3.5GB of video memory per draw command then it will only access the first partition, and 3rd party applications that measure memory usage will report 3.5GB of memory in use on GTX 970, but may report more for GTX 980 if there is more memory used by other commands. When a game requires more than 3.5GB of memory then we use both segments.

 

We understand there have been some questions about how the GTX 970 will perform when it accesses the 0.5GB memory segment. The best way to test that is to look at game performance. Compare a GTX 980 to a 970 on a game that uses less than 3.5GB. Then turn up the settings so the game needs more than 3.5GB and compare 980 and 970 performance again.[/spoiler]

 

A "high priority" and "low priority" ram access firmware/software seems so weird to me to understand. The "benchmark" on pcper showed the vram speed near the bandwidth limit of a 48 bit bus at 6Ghz being 36GB/s when using "both seatmates" or more than 3.5GB of vram.

 

I assume this "issue" is just a software/firmware/board design thing because the 980 does not have this problem, nor do some custom pcb 970s.

[R_Golden]

@Imakuni @zappian @Justhavocman

[R_Golden]

Just like @Imakuni said, test it yourself...

[Dabombinable]

Just like @Imakuni said, test it yourself...

Yep, though as I said before, all of it can be accessed (I proved that myself), but the last 500MB is extremely slow.

[R_Golden]

Yep, though as I said before, all of it can be accessed (I proved that myself), but the last 500MB is extremely slow.

After passing 3.5gb I didn't notice any stuttering like others have pointed out. Not sure if it's a card to card thing.

[Guest]

From what Nvidia said: [spoiler='Nvidia Quote'] "The GeForce GTX 970 is equipped with 4GB of dedicated graphics memory. However the 970 has a different configuration of SMs than the 980, and fewer crossbar resources to the memory system. To optimally manage memory traffic in this configuration, we segment graphics memory into a 3.5GB section and a 0.5GB section. The GPU has higher priority access to the 3.5GB section. When a game needs less than 3.5GB of video memory per draw command then it will only access the first partition, and 3rd party applications that measure memory usage will report 3.5GB of memory in use on GTX 970, but may report more for GTX 980 if there is more memory used by other commands. When a game requires more than 3.5GB of memory then we use both segments.

 

We understand there have been some questions about how the GTX 970 will perform when it accesses the 0.5GB memory segment. The best way to test that is to look at game performance. Compare a GTX 980 to a 970 on a game that uses less than 3.5GB. Then turn up the settings so the game needs more than 3.5GB and compare 980 and 970 performance again." [/spoiler]

 

A "high priority" and "low priority" ram access firmware/software seems so weird to me to understand. The "benchmark" on pcper showed the vram speed near the bandwidth limit of a 48 bit bus at 6Ghz being 36GB/s when using "both seatmates" or more than 3.5GB of vram.

 

I assume this "issue" is just a software/firmware/board design thing because the 980 does not have this problem, nor do some custom pcb 970s.

All 970's have the 'problem', its the way its designed.

 

Here's a breakdown of what's happening on the 970:

 

1. each SM or SMM is a multiprocessing unit that consists of 128 cuda cores.

 

2. The GTX 970 is basically just a GTX 980 with a trio of SM units disabled, everything else is identical.

 

3. Each SM uses what's called 'crossbars' to access memory channels.

 

Since the GTX 970 has several SM's disabled, this results in more load on fewer crossbars when accessing memory channels, so this results in the 970 not being able to access all 4gb vram as efficiently as the gtx 980.

 

For this reason, nvidia logically segmented the gtx 970's vram into 3.5gb and 0.5gb chunks as a performance optimization. This way the first 3.5gb of vram on the 970 maintains the same ratio of SM units to VRAM as the gtx 980, and the gtx 970 can access this 3.5gb chunk with full efficiency.

 

There's really no 'memory bug' with the gtx 970, the card is designed this way. People also need to note that the gtx 970 is a good $200 less than the gtx 980, of course it isn't going to perform as well at very high resolutions as the gtx 980, if it did there would be no reason for the gtx 980 to exist.

 

People are basically whipping themselves up into a frenzy, because a card that is a cut down version of the gtx 980... performs like a cutdown version of the gtx 980. This was nothing unexpected, and the card still performs as well as initial benchmarks indicated.

[Dabombinable]

After passing 3.5gb I didn't notice any stuttering like others have pointed out. Not sure if it's a card to card thing.

I tested by using the memory burner in MSI Kombuster, since it actually lets your view how much is loaded in the vRAM in real time, you can see the speed difference. (I tried overclocking the vRAM as well, 450MHz didn't make a noticeable difference, and none of my games use more than 3GB).

[boboman342]

All 970's have the 'problem', its the way its designed.

 

Here's a breakdown of what's happening on the 970:

 

1. each SM or SMM is a multiprocessing unit that consists of 128 cuda cores.

 

2. The GTX 970 is basically just a GTX 980 with a trio of SM units disabled, everything else is identical.

 

3. Each SM uses what's called 'crossbars' to access memory channels.

 

Since the GTX 970 has several SM's disabled, this results in more load on fewer crossbars when accessing memory channels, so this results in the 970 not being able to access all 4gb vram as efficiently as the gtx 980.

 

For this reason, nvidia logically segmented the gtx 970's vram into 3.5gb and 0.5gb chunks as a performance optimization. This way the first 3.5gb of vram on the 970 maintains the same ratio of SM units to VRAM as the gtx 980, and the gtx 970 can access this 3.5gb chunk with full efficiency.

 

There's really no 'memory bug' with the gtx 970, the card is designed this way. People also need to note that the gtx 970 is a good $200 less than the gtx 980, of course it isn't going to perform as well at very high resolutions as the gtx 980, if it did there would be no reason for the gtx 980 to exist.

 

People are basically whipping themselves up into a frenzy, because a card that is a cut down version of the gtx 980... performs like a cutdown version of the gtx 980. This was nothing unexpected, and the card still performs as well as initial benchmarks indicated.

Nvidia forgot to directly inform us (as well as many tech reviewers) that this is the way a gtx 970 is supposed to "act" when more than 3.5GB of vram is being used BEFORE  everyone bought the cards and found out after the fact! Many people praise it for its 4GB of vram while only 3.5GB is the advertised speed as per spec.

 

I was going to buy a gtx 970.....I probably still will.....but I am sure people feel duped or mislead...

[R_Golden]

I tested by using the memory burner in MSI Kombuster, since it actually lets your view how much is loaded in the vRAM in real time, you can see the speed difference. (I tried overclocking the vRAM as well, 450MHz didn't make a noticeable difference, and none of my games use more than 3GB).

I honestly have no clue why people are blowing up about this. By the time you get a monitor that is 4k to actually need more VRAM you have probably upgraded your GPU, and if you have a 4k monitor why would you get a 970 and not a 980. It kind of puzzles me the lack of logic people place into their arguments, They use numbers and theories to defend an argument rather than using a practical, objective scenario.

[zappian]

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.

 

We need to know minimum fps numbers.

 

Because a stutter is a sudden change in fps producing very low fps for a split second.

 

Thats why minimum fps are important.

[EpicNeuron]

when i use arround 3,7/8 ( shadow of mordor) i don't have frame drops.. sometimes but that's normal i think?

[Justhavocman]

I honestly have no clue why people are blowing up about this. By the time you get a monitor that is 4k to actually need more VRAM you have probably upgraded your GPU, and if you have a 4k monitor why would you get a 970 and not a 980. It kind of puzzles me the lack of logic people place into their arguments, They use numbers and theories to defend an argument rather than using a practical, objective scenario.

Games at 1440p already max out 4Gigs of VRAM  .

[Dabombinable]

Games at 1440p already max out 4Gigs of VRAM  .

And to think at 1080p I use just over 2GB.....

[zappian]

Games at 1440p already max out 4Gigs of VRAM  .

 

Shadow of mordor for example goes trough 4 GB of VRAM very fast at that resolution.

 

People are ignoring nvidias latest fuck up but if AMD fucks up in the future they will bring out their pitchforks.

 

Got to love double standards.

 

People are still mad because AMD launched bad drivers like 5 years ago.

 

via Imgflip Meme Maker

[Dabombinable]

Shadow of mordor for example goes trough 4 GB of VRAM very fast at that resolution.

 

People are ignoring nvidias latest fuck up but if AMD fucks up in the future they will bring out their pitchforks.

 

Got to love double standards.

 

People are still mad because AMD launched bad drivers like 5 years ago.

Some people can ignore it because it doesn't affect them. Simple.

[zappian]

Some people can ignore it because it doesn't affect them. Simple.

 

Just because it doesn't affect YOU doenst mean it inst a problem.

 

Did i ignore the artifacts on the r9 series just because it didnt affect ME and my particular card.

 

Hell no, i gave AMD hell when that happened , a lot of 270x cards are affect by artifcating.

 

Just saying, most people here have been giving me shit about the bad drivers AMD launched 5 years ago 

but this nvidia debacle , shhh just push it under the rug because reasons.

Everytime i reccomend a r9 280 in these forums the response is:

 

 

AMD HAS SHIT DRIVERS YOU MORON.

 

Every damn time , which isnt true anymore.