Official Nvidia GTX 970 Discussion Thread

[Majestic]

Thats why 290x cfx is faster at 4k than 970 SLI.

 

Because the extra VRAM and higher bit BUS makes no difference.

 

Correlation =/= causation. 780 Does even worse, and that has a 384 bitbus. 290 and 290x also have the same bitbus.

I don't think the memory is the determining factor, nor can it be verified since we don't have actual memoryuse numbers.

 

Because if we grab another game, like crysis3, which uses lots of memory;

http://www.hardocp.com/article/2014/11/19/nvidia_geforce_gtx_970_sli_4k_nv_surround_review/5

 

The same can't be said. You're just cherrypicking results.

[zappian]

Correlation =/= causation. 780 Does even worse, and that has a 384 bitbus. 290 and 290x also have the same bitbus.

I don't think the memory is the determining factor, nor can it be verified since we don't have actual memoryuse numbers.

 

Because if we grab another game, like crysis3, which uses lots of memory;

http://www.hardocp.com/article/2014/11/19/nvidia_geforce_gtx_970_sli_4k_nv_surround_review/5

 

The same can't be said. You're just cherrypicking results.

 

The 290x crossfire is still faster in that game.

 

[Majestic]

The 290x crossfire is still faster in that game.

 

 

 

But the 780 SLI is faster than the 290X, which has 384bit 3GB versus 512Bit 4GB. Your argument is invalid, sorry.

This could also very well be within the range it exceeds 3.5GB because of improper heuristics. 

 

It does not give 512bit 4GB any tangible benefit over 3GB or 3.5GB in anyway way or form. So that's NOT a selling point of the 290X, which is why you're making a bigger deal out of it than is called for.

 

Which is also why I call you out on it, everytime.

[zappian]

But the 780 SLI is faster than the 290X, which has 384bit 3GB versus 512Bit 4GB. Your argument is invalid, sorry.

 

I thought we were talking about the 970.

 

And the 780ti has higher bus than the 970 while having more raw performance than the 290x.

 

4k performance :

 

780ti SLI < 290x  CFX < 970 SLI.

 

If the game uses more than 3 GB of vram(crysis doesnt push much the VRAM i have that game):

 

290x CFX < gtx 970 SLI < 780 TI SLI

[Majestic]

I thought we were talking about the 970.

 

No, we are talking about whether the 512BIT 4GB memory buffer has any benefit over the 3GB 384bit GTX 780 or 3.5GB 224bit GTX 970. 

Since it scores LOWER than the 780 in some cases, that theory is proved incorrect. Meaning something else is determining performance.

[Gunjob]

But the 780 SLI is faster than the 290X, which has 384bit 3GB versus 512Bit 4GB. Your argument is invalid, sorry.

This could also very well be within the range it exceeds 3.5GB because of improper heuristics. 

 

It does not give 512bit 4GB any tangible benefit over 3GB or 3.5GB in anyway way or form. So that's NOT a selling point of the 290X, which is why you're making a bigger deal out of it than is called for.

 

Which is also why I call you out on it, everytime.

290 vs 780 are run at lower setting than the 970 vs 290x if you look at the graph again. They compare them in pairs. 

[zappian]

No, we are talking about whether the 512BIT 4GB memory buffer has any benefit over the 3GB 384bit GTX 780 or 3.5GB 224bit GTX 970. 

Since it scores LOWER than the 780 in some cases, that theory is proved incorrect. Meaning something else is determining performance.

 

 

290 vs 780 are run at lower setting than the 970 vs 290x if you look at the graph again. They compare them in pairs. 

 

Like GunJob mentioned the 290 and 780 TI run at lower settings.

So the 290x is still faster.

And the 290 cfx keeps up with the 780 ti SLI , a much priecier solution.

Read the whole page before lashing out at me.

[Majestic]

290 vs 780 are run at lower setting than the 970 vs 290x if you look at the graph again. They compare them in pairs. 

You're right for the Crysis 3 benchmark. But the 290 also has 512bit 4GB memory. Still disproves his argument and he's overstating the importance of the memorybuffer.

 

Again, i'm not arguing 290 vs 970 vs 780. Just tackling is argument about the importance of the memorybuffer.

[zappian]

You're right for the Crysis 3 benchmark. But the 290 also has 512bit 4GB memory. Still disproves his argument.

 

No because it keeps up with me much pricier 780 TI SLI solution.

 

Hell it even wins because it has higher minimum FPS.

 

 

High bus is important at 4k.

[Majestic]

No because it keeps up with me much pricier 780 TI SLI solution.

 

Hell it even wins because it has higher minimum FPS.

 

It's not a TI....

 

If you look at 1920x1080, you can see them at the same distance to each other as here in 4K. Meaning, the memory buffer doesn't equate for jack shit.

So stop overstating it's usefullness.

 

It's the same as people arguing for 8cores on FX CPU's. Overstating their multithreading benefit.

 

1600p http://www.guru3d.com/index.php?ct=articles&action=file&id=13423&admin=0a8fcaad6b03da6a6895d1ada2e171002a287bc1

1080p http://nl.hardware.info/reviews/5621/13/nvidia-geforce-gtx-980--970-review-incl-ultra-hd-benchmarks-crysis-3

 

Same distance. Memorybuffer isn't important. Atleast not as important as you make it to be.

[zappian]

It's not a TI....

 

If you look at 1920x1080, you can see them at the same distance to each other as here in 4K. Meaning, the memory buffer doesn't equate for jack shit.

So stop overstating it's usefullness.

 

It's the same as people arguing for 8cores on FX CPU's. Overstating their multithreading benefit.

 

I meant 780 . The 780 SLI is more expensive than 290 CFX.

And it wins at 4k ( the 290 wins i mean)

 

It must be for some reason.

 

Like the 290x cfx is cheaper and wins at 4k against 970 SLI.

[Gunjob]

You're right for the Crysis 3 benchmark. But the 290 also has 512bit 4GB memory. Still disproves his argument and he's overstating the importance of the memorybuffer.

In regard to the 290/780 result, I'd make the case that the 780 was and is just a better GPU. And is probably held back by a lack of VRAM/bus but this is one of those situations that pretty difficult to prove either way. Unless they release a new revision with the same core clocks and core count but more VRAM and enlarger Bus to account for the extra VRAM (looking at you cards with doubled memory) 

 

But I'm sure we can all agree at UHD and above res', a larger VRAM amount as well as high memory bandwidth is pretty key to not only a good frame rate, but a smoother experience in general. This is often shown when Nvidia's cards a pretty much king down at 1440p and below. But the gap is often closed in multi monitor and UHD+ by AMD. 

[Paragon_X]

Since when did a VaporX reach 56dBA?

 

 

You must clearly misunderstand what Mantle was intended to do, it was intended to lower CPU overheads and allow more work to be offloaded to the GPU, so people with low end CPU's would see more more performance than those with high end ones.

I didnt want to get involved in this but I lol'd they measure the dBA from 10cm, i cant read german but they got the same results as a link posted before noting/comparing the dBA of all r9 290/X. Also I have to mention that i own a r9 290 VaporX and I can barely hear it with the sidepanel open.

Also About this topic:

FACTS:

- Nvidia misrepresented the Specs of the card

- It fucks up people that wanted to SLI for whatever reason to play at a higher resolution or even DSR

- 970 is an awesome card still

 

If Nvidia gets unscathed with saying whatever the hell they want on the specs. Who can ensure me that my next purchase from them, or even AMD if they see that people don't give a fuck about precision on advertisement will be correct? 

 

They need to receive the message pretty clear so they will make sure this will not happen again.

[zappian]

In regard to the 290/780 result, I'd make the case that the 780 was and is just a better GPU. And is probably held back by a lack of VRAM/bus but this is one of those situations that pretty difficult to prove either way. Unless they release a new revision with the same core clocks and core count but more VRAM and enlarger Bus to account for the extra VRAM (looking at you cards with doubled memory) 

 

But I'm sure we can all agree at UHD and above res', a larger VRAM amount as well as high memory bandwidth is pretty key to not only a good frame rate, but a smoother experience in general. This is often shown when Nvidia's cards a pretty much king down at 1440p and below. But the gap is often closed in multi monitor and UHD+ by AMD. 

 

Its a better GPU at lower resolutions.

 

At 4k amd gpus are cheaper and faster.

 

That much cant be ignored.

[Majestic]

I meant 780 . The 780 SLI is more expensive than 290 CFX.

And it wins at 4k ( the 290 wins i mean)

 

It must be for some reason.

 

Like the 290x cfx is cheaper and wins at 4k against 970 SLI.

 

It doesn't win, it ties. Meaning nothing since by your logic it should have a clear advantage over the 780. And it just performs the same compared to the 780 in 1080p or 1600p.

Not talking about pricing, will you stick to one aspect please (added benefit of higher memorybus) and stop deflecting because you sense you're losing the argument again.

 

If what you said was true, what applied to the 970 and 290, should also apply to the 290 and 780. And it just doesn't.

 

That much cant be ignored.

 

 

It can, because I just did.

 

In regard to the 290/780 result, I'd make the case that the 780 was and is just a better GPU. And is probably held back by a lack of VRAM/bus but this is one of those situations that pretty difficult to prove either way. Unless they release a new revision with the same core clocks and core count but more VRAM and enlarger Bus to account for the extra VRAM (looking at you cards with doubled memory) 

 

But I'm sure we can all agree at UHD and above res', a larger VRAM amount as well as high memory bandwidth is pretty key to not only a good frame rate, but a smoother experience in general. This is often shown when Nvidia's cards a pretty much king down at 1440p and below. But the gap is often closed in multi monitor and UHD+ by AMD. 

 

Performance on 1080p and 1600p is similar. That's why i'm making the argument.

[zappian]

It doesn't win, it ties. Meaning nothing since by your logic it should have a clear advantage over the 780. And it just performs the same compared to the 780 in 1080p or 1600p.

Not talking about pricing, will you stick to one aspect please (added benefit of higher memorybus) and stop deflecting because you sense you're losing the argument again.

 

 

 

It can, because I just did.

 

 

Performance on 1080p and 1600p is similar. That's why i'm making the argument.

 

It does perform better the 290x CFX crossfire performs better than 970 SLI.

And the 780 SLI performs worse than the 290 SLI:

Same avg fps , but the 290 configuration boasts higher minimum fps and higher max fps.

 

 

Why do you think a card that performs worse at 1080p when you get to 4k performs better?

Better VRAM and higher bit BUS.

[mr moose]

The analogy is right since the card craps out after 3.5 GB .

 

People were simply not aware of that at release and are now.

 

Its still the same card , we simply know more about how it works now.

 

And we know the limitations of the card when we didn't before.

 

That much is accurate.

 

Cheers.

 

please, everyone knew the limitations of the card before purchase, there are many respectable reviews and performance benchmarks.  :rolleyes: 

[zappian]

please, everyone knew the limitations of the card before purchase, there are many respectable reviews and performance benchmarks.  :rolleyes:

 

Did you fully understand the memory architecture and limitations at launch?

Did you know that the card had less SM units and less l2 cache ?

 

Did you know that the card had performance problems over 3.5 GB of VRAM?

If we did this wound not  be an issue.

You are telling me we knew this at launch and understood this perfectly:

 

.

 

You must work at nvidia then.

[Majestic]

It does perform better the 290x CFX crossfire performs better than 970 SLI.

And the 780 SLI performs worse than the 290 SLI:

Same avg fps , but the 290 configuration boasts higher minimum fps and higher max fps.

 

 

Why do you think a card that performs worse at 1080p when you get to 4k performs better?

Better VRAM and higher bit BUS.

 

Yes, the 290x CFX, in this test, performs better than a 970 SLI. But it's not because it has any benefit for having a 512bit 4GB buffer.

Because if that were the case, the 290 would also score better than teh 780 and it doesn't. Don't give me semantics over a 5fps difference in min. framerate. That could've just as well been the CPU in that test. 

 

All this proves is that 3GB 384bit vs 512GB 512bit are not significantly different. 3.5GB 224bit is TBD after nvidia attempts to fix it with drivers. I'm going to guess it doesn.t.

 

By the way, 290x sits at 49fps average and 30min. Meaning 4K is also out of ITS comfortzone. So why recommend those cards for 4K anyway.

[zappian]

Yes, the 290x CFX, in this test, performs better than a 970 SLI. But it's not because it has any benefit for having a 512bit 4GB buffer.

Because if that were the case, the 290 would also score better than teh 780 and it doesn't. Don't give me semantics over a 5fps difference in min. framerate. That could've just as well been the CPU in that test. 

 

All this proves is that 3GB 384bit vs 512GB 512bit are not significantly different.

 

So why does it perform better then?

Since the 970 smacks the 290x at 1080p easily.

 

At least you agree with me than 290x CFX is the best 4k config.

 

970 is 256 bits lost.

 

780 is 384 bits lost.

 

290x ix 512 bits won

 

290 is 512 bits won.

Its not just in this test , look at the rest of the tests done.

 

It wins most of the  times the 290x CFX i mean.

The 970 only wins in watch dogs because thats a nvidia optimized title and a gameworks title.

[Majestic]

So why does it perform better then?

 

Not sure, but it's not the buffer.

 

Again, I wasn't arguing whether 290x or 970 or 780 was better for 4K. I disproved your argument that you absolutely need the 290x memorybus for 4K.

[zappian]

Not sure, but it's not the buffer.

 

Not sure.

 

Okay.

 

Its the magic powder amd puts in their cards so they perform better at 4k.

 

But lose at 1080p

 

We will stay with that conclusion i guess since you wont agree with me.

 

You haven't disproved nothing im afraid my theory is the memory buffer and indeed cards

with higher memory buffer win at 4k.

[Majestic]

Its the magic powder amd puts in their cards so they perform better at 4k.

 

Your proof doesn't suggest this. Why do you keep stating this as a fact when your own evidence doesn't support this. Don't be so daft please..

 

I've made my statement. Any sane person who reads it will know better. You can just stay in wonderland. I'm done with you.

[zappian]

Your proof doesn't suggest this. Why do you keep stating this as a fact when your own evidence doesn't support this. Don't be so daft please..

 

Im gonna go ahead and report you since its the fourth time you have used name calling in this argument,

 

Higher bus cards are wining at 4k.

 

Thats my theory.

 

Feel free to disagree with it , but we know that a higher bus helps at higher resolutions.

[Majestic]

Feel free to disagree with it , but we know that a higher bus helps at higher resolutions.

 

I did, but your cognitive dissonance is higher than the empire state building. You don't even realise how thick you're being, dunning-kruger in full effect.

 

Hence i'm done with you, for my own sanity's sake.