By using this site, you agree to our Privacy Policy and our Terms of Use. Close
Pemalite said:
Bofferbrauer2 said:

I will concentrate on the 1440p data as the 1080p goes way too high in FPS and 4k is too harsh on early GPUs of this comparison.

  • HD 7970: 26 FPS
  • R9 380X: 29 FPS
  • RX 470: 36 FPS
  • RX 570: 43 FPS
  • RX 5600: N/A, but the 5600XT has 63 FPS, hinting at probably around 56 FPS and beating the Fury cards 
  • RX 6600XT: 87 FPS
  • RX 7600: 95 FPS
  • RX 7600XT: 101 FPS

So yeah, performance in this game more than tripled with the same amount of CU. Most of this is due to clock speeds (7970 Ghz edition: 1000 Mhz, 7600XT: 2755 Mhz boost), but there must have been some IPC improvements also.

What I'm surprised at is the gap between the 470 and 570, because it's functionally the same chip and the boost clock speeds differ only slightly (38 Mhz). Probably the higher TDP of 150W compared to 120W has allowed it to boost more and longer. Meanwhile one can clearly see the sharp increase in clock speeds with RDNA 2 resulting in doubling the FPS over the 570. Too bad the 5600 is MIA, it would have been a nice point of reference.

And now we'll have to wait and see it the 9060XT will be added to the chart.

Drivers, faster RAM and higher TDP.

That's where some big gains have been had... Especially during the Polaris era where AMD sought to overhaul their drivers so they weren't "second tier" to nVidia.
Now arguably better than nVidia in many aspects.
It wasn't unusual to see increases of 20% in some instances.

Interesting stuff that, despite the same number of functional units, we have had some decent gains, although... 26fps to 101fps or roughly 400% over a 13~ year period probably makes it seem less impressive.

Can confirm RDNA2 has double the performance over Polaris as the GPU I had before this was the RX580, which I sold for $600 AUD, which was the exact amount needed for the RX6600 during the Crypto craze. Haha But that was just slightly less than doubling.

Now if only we can get them clock normalized so we could analyze the IPC gains year on year, but that's probably a tall ask.

Higher TDP?

  • HD 7970: 250W
  • R9 380X: 190W
  • RX 470: 120W
  • RX 570: 150W
  • RX 5600: 150W
  • RX 6600XT: 160W
  • RX 7600: 170W
  • RX 7600XT: 190W

Until the RX 470 the TDP actually went down a lot for the same number of CU with a small clock speed increase.

Since then however, the clock speeds got ramped up significantly, and with it, the TDP also went up - but not that much that it would be the reason for the higher performance, though it certainly helped.

Jizz_Beard_thePirate said:

Nintendo Switch 2 clocks confirmed: NVIDIA Ampere GPU runs at 561 MHz in portable mode, max 1.4 GHz

https://videocardz.com/newz/nintendo-switch-2-clocks-confirmed-nvidia-ampere-gpu-runs-at-561-mhz-in-portable-mode-max-1-4-ghz

A lot more info in the video. Worth a watch.

So it's confirmed that the Switch 2 has a higher CPU clock on mobile than docked, I wonder why that is...

Jizz_Beard_thePirate said:

GeForce RTX 5090 Laptop GPU disappoints: barely beats RTX 4090 by 3% at 4K, no real gains at 1080p or 1440p

https://videocardz.com/newz/geforce-rtx-5090-laptop-gpu-disappoints-barely-beats-rtx-4090-by-3-at-4k-no-real-gains-at-1080p-or-1440p

NVIDIA reportedly removes POPCNT driver requirement, making RTX 5090 and Core 2 Duo pairing possible

https://videocardz.com/newz/nvidia-reportedly-removes-popcnt-driver-requirement-making-rtx-5090-and-core-2-duo-pairing-possible

Was to be expected for the laptop GPU, most performance gains of Blackwell seem to stem from higher TDP, which you can't do on mobile...

I wish I still had the old PC with the Core 2 duo E4300 and Radeon X1650 Pro (was my stepfather's PC, and despite the low-power hardware inside got hot as hell) and pair that thing with a 5090...