agent_x007
New member
When Nvidia lost "G" from "GPU" - my thoughts about GTX 750 (Ti) a.k.a. GM107
Hello everyone.
I recently stumbled opon a problem we didn't had so far (and it's quite scary for me), but I start from the begining :
In Q1 2014, Nvidia launched a new product called GM107, or as they called it :
"1-st generation Maxwell GPU".
Everything seems great :
It was consuming a lot less than previous GK107 (or Kepler based GPU), and was still giving us performance gain.
But there was a small problem with it.
In whitepaper (LINK) we read :
U can now say : So what, Kepler was similar, right ?
And that is almost true.
But then we get to scary part :
Few months ago Nvidia launched second generation Maxwell, with VXGI, MFAA, better memory compresion algorithm, DX12, etc.
And ALL Kepler based GPU's do support DX11, or TXAA graphic funtions, and since GTX980/970, that is NOT the case with Maxwell based GPUs.
End result :
About a week ago Nvidia relesed to public new tech demo for "Maxwell architecture" called :
Apollo 11 : LINK
It uses new VXGI technology to get better image quality.
BUT this demo will NOT run on NOT Maxwell GPUs.
AND GM107 is NOT a Maxwell GPU, according to this official NV demo.
So my problem is : What is the meaning of having a naming scheme for GPU architectures, if GPUs from the same architecture will not support the same graphic functions ?
It's GPU - so it stand as "Graphics Processing Unit", right ?
"G" is first and should be most important from NV perspective.
But based on this, GM107 is not a new generation GPU.
It's a new generation "PU", or "Processing Unit".
Or basicly : A re-think of doing the same thing's, just better.
That's why, to me it's not enough to call GM107 a "new generation GPU" and Nvidia agrees aparently with me by locking Maxwell Demo to only 2-nd gen. Maxwell cards.
Conlusion :
What I'm afraid is that Nvidia start's locking graphic features to markets they target.
For example : DX12, VXGI, etc. will be able locked to mid to high end cards because... architecture naming is irrelavent (+ they get more ca$h from this).
And I don't want, for example "Big Maxwell" to be called a "Pascal architecture GPU", when it won't support the same features as rest of "Pascal architecture" lineup.
How about U guys ?
I may be overreacting here, sure. I just deeply concerned about what future NV products will look like from a graphic feature perspective (or GPU perspective).
Thank U all for reading.
Hello everyone.
I recently stumbled opon a problem we didn't had so far (and it's quite scary for me), but I start from the begining :
In Q1 2014, Nvidia launched a new product called GM107, or as they called it :
"1-st generation Maxwell GPU".
Everything seems great :
It was consuming a lot less than previous GK107 (or Kepler based GPU), and was still giving us performance gain.
But there was a small problem with it.
In whitepaper (LINK) we read :
So we get all the advances in architecture but without any new graphics functions.From a graphics features perspective, our first-generation Maxwell GPUs offer the same API functionality as Kepler GPUs.
U can now say : So what, Kepler was similar, right ?
And that is almost true.
But then we get to scary part :
Few months ago Nvidia launched second generation Maxwell, with VXGI, MFAA, better memory compresion algorithm, DX12, etc.
And ALL Kepler based GPU's do support DX11, or TXAA graphic funtions, and since GTX980/970, that is NOT the case with Maxwell based GPUs.
End result :
About a week ago Nvidia relesed to public new tech demo for "Maxwell architecture" called :
Apollo 11 : LINK
It uses new VXGI technology to get better image quality.
BUT this demo will NOT run on NOT Maxwell GPUs.
AND GM107 is NOT a Maxwell GPU, according to this official NV demo.
So my problem is : What is the meaning of having a naming scheme for GPU architectures, if GPUs from the same architecture will not support the same graphic functions ?
It's GPU - so it stand as "Graphics Processing Unit", right ?
"G" is first and should be most important from NV perspective.
But based on this, GM107 is not a new generation GPU.
It's a new generation "PU", or "Processing Unit".
Or basicly : A re-think of doing the same thing's, just better.
That's why, to me it's not enough to call GM107 a "new generation GPU" and Nvidia agrees aparently with me by locking Maxwell Demo to only 2-nd gen. Maxwell cards.
Conlusion :
What I'm afraid is that Nvidia start's locking graphic features to markets they target.
For example : DX12, VXGI, etc. will be able locked to mid to high end cards because... architecture naming is irrelavent (+ they get more ca$h from this).
And I don't want, for example "Big Maxwell" to be called a "Pascal architecture GPU", when it won't support the same features as rest of "Pascal architecture" lineup.
How about U guys ?
I may be overreacting here, sure. I just deeply concerned about what future NV products will look like from a graphic feature perspective (or GPU perspective).
Thank U all for reading.
Last edited: