Carzy Zarx’s PC Gaming Emporium - Catch Up on All the Latest PC Gaming Related News

[Jizz_Beard_thePirate]

haxxiy said:

A worrying trend: flagship GPUs have increased power consumption by ~15% per generation (in real-world scenarios, although the result of estimating it using TDP is similar). This is happening against a backdrop of decreasing gains in efficiency with rare exceptions due to architectural improvements.

So far it has been used to average ~50% gains per generation for the last decade, but ahead of us lies a bunch of nodes with worse power characteristics all the way to the end of CMOS die shrinking.

Will we see a 5090 with a 560W TDP to get to that 40% improvement? A 6090 keeping said TDP but consuming closer to it? A 7090 going for 650W? Or will consumer pushback and/or a novel innovative architecture save us from going that path?

It is why Nvidia, AMD and Intel are all pushing upscaling technologies along with Frame amplifiers like DLSS 3. The ceiling of gen on gen improvements is fast approaching and there's only so much you can do with the architectural side with minimal node improvements. But on the software side, there's infinite possibilities. They have fixed a lot of artifacting issues already with the DLSS 3 update and as with DLSS 2, it will only continue to get better until eventually, it becomes widely accepted like DLSS 2 has been.

Bofferbrauer2 said:

Captain_Yuri said:

A: It is a pretty small improvement because a 6800U has a tdp of 28 watts. So if we compare the 28 watt 6800U to the 25 watt 7940HS, that's like 8-9% improvement. 6900HS also has a tdp of 35 watts... So in order for the 7940HS to get 2791 score, it needed to be bumped up to 54 watts... So that's a 54% increase in wattage for a 22% increase in performance... That's horrid... I expect better from AMD, that's for sure.

B: Also a 3050 has nearly double the time spy score so it's not really close...

That's also some Intel logic you go there. Sure it can beat everything last gen at 25 watts but by 1-8% is pretty terrible...

The Steam Deck does not have a Vega 8 GPU. I have no idea where you got that from. It has 8 RDNA 2 CUs and the APU runs at 15 watts. The point was that because of how inefficient the 7940HS is, Valve might be better off going with RDNA 2 solution or waiting for RDNA 4 to see if there will be bigger efficiency gains for the Steam Deck 2.

A: Keep in mind that the 6800U also has more bandwidth to play with (LPDDR5-6400 compared to LPDDR5-5600), which certainly skews this in favor of the 6800U. And despite having more TDP headroom (28W vs 25W) and more memory bandwidth, the 6800U gets beaten by the 7940HS by about 9%.

B: I was wrong about the 3050, I thought it was weaker than that considering how bad the desktop version is, mea culpa on that one.

C: I meant RDNA2, no idea why I wrote Vega 8 there, maybe by association since it uses Zen 2 (it was pretty late). Still, Steam Deck just lingers between 1500-1700 points depending on settings, which is way below what the 780M achieves. Even at 15W, there would still be a healthy uplift, especially if paired with faster memory, as I'm fairly sure the 780M could go quite a bit higher if the Bandwidth wouldn't strangle it.

A: Imo at best the uplift of increasing memory bandwidth will be a few % because of how power limited the APU is. We see this because increasing the power netted big gains even though it's at the same memory speed. So even if we are generous and say 15% gen on gen uplift at 25 watts, I think it's pretty disappointing. Perhaps RDNA 4 will fix a lot of RDNA 3s short comings as Lovelace certainly fixed a lot of Amperes shortcomings after Nvidia went dual fp32 for the first time.

c: I think for Valve, if we assume the uplift is 15% with RDNA 3 compared to RDNA 2, it may be cheaper for them to go with RDNA 2 and spending the saved budget on a bigger battery thus increasing the tdp to 25 watts along with other improvements. Of course, this really depends on the difference they negotiate. If the difference in cost by going with the latest is big vs last gen, then the 15% uplift may not be worth it but if it's small, then sure. Or just wait for RDNA 4 and see how that pans out.