Tested a game with 5K DSR + 4x Super Sample...

[m6tzg6r]

And on the menu it was running like 0.1fps, it was so slow i thought the PC had froze, but it was just processing frames so slow it looked frozen.

Anyway my VRAM was maxed at 3GB, my motherboard ram had used 7.5GB out of the 8GB i have installed, and SSD pagefile was using 15GB. Ive never seen numbers like that all at once.

So i wonder what resolution the game was pumping out with the game res on 5K with 4x Super Sample.

With 1440p, 4x Super Sample gives me 5K res, but i cant calculate what 4x Super Sample on top of a 5K res equals 8K or something?

 

[64K]

Well, if you're talking about rendering a 5K screen at 4X DSR then we're talking about rendering around 59 million pixels or the equivalent pixels of 28.5 1080p screens and then down scaling back to a 5K frame. I'm not surprised at your FPS. I can't imagine how much VRAM you would need to do that. I'm certain that your system RAM is being used to offload to from your GPU and maybe your disk drive as well.

Edit: I misread you post OP. You were using a 1440p screen at 4X DSR. That would be rendering about 14.75 million pixels or the equivalent pixels of 7.1 1080p screens so I guess it was the combination of that and running 4X SS on top of it that murdered your frame rate.

 

[RCoon]

DSR + SuperSampling = stupid idea. Pick one or t'other, not both. Either that or try it on CS 1.6 and enjoy a fluid yet "fine" experience.

 

[m6tzg6r]

DSR + SuperSampling = stupid idea. Pick one or t'other, not both. Either that or try it on CS 1.6 and enjoy a fluid yet "fine" experience.

Yeah i only did it for fun to see what would happen. None of my older games have Sampling options, so the best they can do is 5K, only newer games have Sampling so that was all i could test.

It was just a test to see what would happen if you combined both of those brutal settings together.

 

[qubit]

@m6tzg6r I think what you did there was epic. I love testing extremes as well. It's also impressive that it didn't crash.

Which game was it?

 

[erocker]

Yeah, I was about to say.. The fact the computer didn't actually crash is pretty amazing.

 

[m6tzg6r]

Sniper Elite Zombie Army Trilogy. Video card temps were also not bad, i think because the frame rate was like <1fps the video cards were not really working at all so temps were not high.

Also i was only in the game menu, i wasnt even in game, so if i was ingame things would have been worse, if it would have actually let me load into the game that is.

 

[Vayra86]

Basically you have managed to saturate all bandwidth within your system, that is why your cards are not as hot. Everything's busy pumping data through and around your memory, both within GPU and between GPU > RAM.

 

[rtwjunkie]

Well, if you're talking about rendering a 5K screen at 4X DSR then we're talking about rendering around 59 million pixels or the equivalent pixels of 28.5 1080p screens and then down scaling back to a 5K frame. I'm not surprised at your FPS. I can't imagine how much VRAM you would need to do that. I'm certain that your system RAM is being used to offload to from your GPU and maybe your disk drive as well.

Edit: I misread you post OP. You were using a 1440p screen at 4X DSR. That would be rendering about 14.75 million pixels or the equivalent pixels of 7.1 1080p screens so I guess it was the combination of that and running 4X SS on top of it that murdered your frame rate.

More @64K math stuff. It's everywhere on here! I love it!

 

[qubit]

Sniper Elite Zombie Army Trilogy. Video card temps were also not bad, i think because the frame rate was like <1fps the video cards were not really working at all so temps were not high.

Also i was only in the game menu, i wasnt even in game, so if i was ingame things would have been worse, if it would have actually let me load into the game that is.

Oh the cards were working hard all right. Basically, the GPUs are going round and round in loops all the time. What determines your framerate is how fast it goes round that loop, which changes from moment to moment as the scene changes. In your case, they had such an enormous amount of data to process for each loop, that they took a very, very long time to go round it.

I'm not sure why they weren't that hot, but I do know that using a higher resolution mode makes a graphics card run noticeably hotter as uses more vram.

 

[kn00tcn]

Oh the cards were working hard all right. Basically, the GPUs are going round and round in loops all the time. What determines your framerate is how fast it goes round that loop, which changes from moment to moment as the scene changes. In your case, they had such an enormous amount of data to process for each loop, that they took a very, very long time to go round it.

I'm not sure why they weren't that hot, but I do know that using a higher resolution mode makes a graphics card run noticeably hotter as uses more vram.

er what, using vram has nothing to do with temps, amount of processing has to do with temps, BUT ALSO the higher the framerate, the higher the temps when the gpu usage is theoretically the same

if the pagefile showed 15gb usage, then maybe it was constantly thrashing data to disk, which means gpu usage goes closer to zero as it's waiting for data (same idea as being cpu limited, gpu usage goes down)

 

[qubit]

er what, using vram has nothing to do with temps, amount of processing has to do with temps, BUT ALSO the higher the framerate, the higher the temps when the gpu usage is theoretically the same

if the pagefile showed 15gb usage, then maybe it was constantly thrashing data to disk, which means gpu usage goes closer to zero as it's waiting for data (same idea as being cpu limited, gpu usage goes down)

Using more vram certainly does increase temperature as this is my experience and I wasn't even looking for it. It's dead easy to test, too. Just run something graphically intensive at, say, 800x600 and then 1920x1080 and compare temperatures between the two.

Again, it's not so much the framerate that will heat up the GPU (assuming vsync off, of course) but the kind of calculations it has to do. The framerate is simply determined by how long it takes to go round the loop, with long loops obviously taking longer. Easy example: Unreal Tournament 2004 at 250fps+ doesn't generate all that much heat, while Furmark at 50fps hits the temperature limit and the card starts throttling. I've actually measured these things, hence I know this for a fact.