RTX as graphic acelerator

[iskander87]

Hey updated my work pc to 12 gen and I'm buying a card tomorrow, an RTX 2070, can I keep using the integrated UHD770 as video output and keep the nvidia just as an acelerator?

[igormp]

13 minutes ago, iskander87 said:

Hey updated my work pc to 12 gen and I'm buying a card tomorrow, an RTX 2070, can I keep using the integrated UHD770 as video output and keep the nvidia just as an acelerator?

Why would you do that? Depending on what you're doing (specially if it includes some form of graphics output), you'd hinder performance since the 2070 would need to pass the framebuffer back to the 770 for display, generating a bottleneck.

 

But yes, you can do that.

[Middcore]

There is zero reason to do this. 

[iskander87]

44 minutes ago, igormp said:

Why would you do that? Depending on what you're doing (specially if it includes some form of graphics output), you'd hinder performance since the 2070 would need to pass the framebuffer back to the 770 for display, generating a bottleneck.

 

But yes, you can do that.

yeah I know, for visor draw etc it would be hindering the performance, but I want it for vray render and EmberGen, denoise, topaz image upscaling etc, so I am thinking on it more as an accelerator than anything else

[igormp]

17 minutes ago, iskander87 said:

yeah I know, for visor draw etc it would be hindering the performance, but I want it for vray render and EmberGen, denoise, topaz image upscaling etc, so I am thinking on it more as an accelerator than anything else

Yes, it is possible, but we aware that you won't gain much from doing this.

[iskander87]

8 minutes ago, igormp said:

Yes, it is possible, but we aware that you won't gain much from doing this.

well, there are softwares I can't even run without an nvidia card present, and I imagine he power consumption, heat etc. will be lower when not using the card I'm I wrong?

[iskander87]

13 minutes ago, igormp said:

Yes, it is possible, but we aware that you won't gain much from doing this.

 

[igormp]

25 minutes ago, iskander87 said:

well, there are softwares I can't even run without an nvidia card present

Sorry if I didn't make it clear: I vouch for you to use the nvidia gpu as your output and main GPU, and not only having your iGPU.

26 minutes ago, iskander87 said:

and I imagine he power consumption, heat etc. will be lower when not using the card I'm I wrong?

It'll be irrelevant.

16 minutes ago, iskander87 said:

 

1- That's only valid for video editing, where the built-in decoder will be faster than the 3090's since it won't need to transfer data through the PCIe bus.

2- This video is somewhat misleading, Nvidia's ENCODER is way better than Intel's.

3- For the stuff that you mentioned, mostly ML stuff, the iGPU is useless.

[iskander87]

10 minutes ago, igormp said:

Sorry if I didn't make it clear: I vouch for you to use the nvidia gpu as your output and main GPU, and not only having your iGPU.

It'll be irrelevant.

1- That's only valid for video editing, where the built-in decoder will be faster than the 3090's since it won't need to transfer data through the PCIe bus.

2- This video is somewhat misleading, Nvidia's ENCODER is way better than Intel's.

3- For the stuff that you mentioned, mostly ML stuff, the iGPU is useless.

how come irrelevant? can you point me to somewhere validating your statement? 

another thing... what's "ML stuff"?

[igormp]

5 minutes ago, iskander87 said:

how come irrelevant? can you point me to somewhere validating your statement? 

GPUs are really efficient. Even if it's plugged in, but not connected to a display, they'll idle in their lowest clock state (using 10~20W).

And when you do have it plugged in, it'll also idle in the same state in its 2D-clocks mode (with the same 10~20W, maybe a little bit higher due to the extra vram usage), so you won't be saving much in that case.

 

8 minutes ago, iskander87 said:

another thing... what's "ML stuff"?

Machine learning. Topaz uses machine learning (which in turn requires cuda) underneath for its upscaling.

[iskander87]

6 minutes ago, igormp said:

GPUs are really efficient. Even if it's plugged in, but not connected to a display, they'll idle in their lowest clock state (using 10~20W).

And when you do have it plugged in, it'll also idle in the same state in its 2D-clocks mode (with the same 10~20W, maybe a little bit higher due to the extra vram usage), so you won't be saving much in that case.

 

Machine learning. Topaz uses machine learning (which in turn requires cuda) underneath for its upscaling.

and what do you think will be faster a gpu rendering or the same gpu rendering and handling the display gui etc at the same time?
talking about FStorm, V-ray, Redshift, and EmberGen
and speaking of gui, which system do you think will be more fluid, one connected through an igp with another gpu doing stuff or the same connected through the same gpu that it's doing the stuff?

besides I can't imagine a 200 watts gpu consuming the same doing nothing than working no matter how light the work is
 

[iskander87]

14 minutes ago, igormp said:

GPUs are really efficient. Even if it's plugged in, but not connected to a display, they'll idle in their lowest clock state (using 10~20W).

And when you do have it plugged in, it'll also idle in the same state in its 2D-clocks mode (with the same 10~20W, maybe a little bit higher due to the extra vram usage), so you won't be saving much in that case.

 

Machine learning. Topaz uses machine learning (which in turn requires cuda) underneath for its upscaling.

bear in mind that some gpu/igp needs to drive this gui and handle those little balls that represent the materials to be used on the scene... with a  6700k and 2 x gtx 1070 made a huge difference to have the monitor connected to the board instead of any of the cards

[igormp]

1 hour ago, iskander87 said:

and what do you think will be faster a gpu rendering or the same gpu rendering and handling the display gui etc at the same time?
talking about FStorm, V-ray, Redshift, and EmberGen

The difference in performance should be less than 5%, so pretty negligible.

1 hour ago, iskander87 said:

and speaking of gui, which system do you think will be more fluid, one connected through an igp with another gpu doing stuff or the same connected through the same gpu that it's doing the stuff?

Again, that's a non-issue.

1 hour ago, iskander87 said:

besides I can't imagine a 200 watts gpu consuming the same doing nothing than working no matter how light the work is

Do you know what? Buy the GPU since you need it anyway, then you can try it for yourself and see which performs better and uses less power

[iskander87]

yeap like 20 seconds less per frame on 1 minute render frame, that sounds more than 5pc

and I don't have way to measure power but unless rendering, afterburner manages to turn off the fans and its like 38c while working on 3ds max x vs 48 and the fans running when the cable is on the card

 

so not pointless at all don't you think?