Radeon R9 270X vs GeForce GTX 760

Radeon R9 270X vs GeForce GTX 760 - GPU Duel (ep. 34)

Рет қаралды 2,659

Күн бұрын

Пікірлер: 7

@gamecomparisons 2 жыл бұрын

The HD 7870 Ghz edition is still running fine in my Athlon X4 860K system. I only stopped using it because I got a deal on the PowerColor RX480 with 8GB of VRAM back in 2018 or so. Otherwise I would still be using it. Thanks for making this, it is good to see the same tier of Graphics card tech tested these days.

@CompatibilityMadness 2 жыл бұрын

Good test. Interesting they are so close together considering they respective architectures. GCN may work better in late DX11/DX12 games (as Kepler DX12 implementation isn't best, and Vulkan is just plain bad). Something to think about : Shouldn't "Unified SP" on Kepler be changed to "Cuda Cores" ? PS. Putting PCI-e 3.0 usage is nice (I'm guessing Radeon didin't provide this data ?), but I would like to see it below VRAM parameters (since it makes all GPU parameters on same level vs. Radeon cards). EDIT : Any chance of retest on Windows XP :D ?

@hardwaremuseum3532 2 жыл бұрын

Honestly - I expected bigger lead of CGN. Yes it is somewhat better in FC5 and Control (both should be "new" enough), but I expected more. Perhaps the 16x HQ AF has bigger performance hit for AMD GPUs. The Application is universal so it is either PS + VS or Unified SP. Cuda cores is just marketing name for the same thing... and I don't like marketing names :) Bus usage - yes, this is NV only feature. I believe not even RDNA2 can show this parameter. As for the XP - sorry, no chance. I have only few GPUs remaining to test and after that I will start to work on CPU benchmarks. Maybe i will make similar videos of CPU gaming performance.

@CompatibilityMadness 2 жыл бұрын

@@hardwaremuseum3532 Fair enough. I think DX11 is mostly the reason they are close (Vulkan and DX12 would fly on GCN). I understand no XP tests, but CPU tests should be interesting so I will wait for that to come :) EDIT [SP vs CC] : I understand naming all units in DX10 and later an "SP", but this can bring up many confusion about comparisons. I get that from instruction support, or in this case performance in applications, they are *almost the same (*counting here DX12/Vulkan support), in which case - the same naming is fine. BUT, I prefer to think of different naming for NV SPs as an easy way to differentiate many architectures they can represent (marketing or not, I don't like to name two things the same when they clearly are not representing the same value [in this case performance]). For example Fermi and Kepler "384SPs" GPUs, would mean two different performance categories of cards (GT 710/740 vs. GTX 560 Ti). Same goes for Maxwell/Pascal/Turing/Ampere, and ALL of them have different "per SP unit" performance from AMD's GCN SPs (which are somewhat cosistent throughout GCN versions range), but throw in Terascale/RDNA GPUs, and everything get's confusing again.