From the performance figures they've given, Ampere has 98% more flops per watt than Turing but only 21% more performance, on average. That means one needs 1.61 Ampere flops to equal the peformance of 1 Turing flops, and 1.5 Ampere flops to equal 1 RDNA 1.0 flops.

It seems clear to me each shader was effectively cut in half before some architectural improvements, or perhaps it was the increased number of FP32 engines themselves that increased the performance relative to Turing.

With RDNA 2.0 apparently focusing on IPC, it would seem like Nvidia and AMD have more or less switched places concerning what their GPU design philosophy historically used to be. Ampere is very Terascale-like (lots of shaders, lower clocks and performance) while RDNA 2.0 is kind of Fermi-like (higher cloks and IPC but less shaders).

An Ampere CUDA core also has some similarities with Bulldozer modules, in that a second (integer in the case of Bulldozer, floats in the case of Ampere) unit was added to each processing core to increase performance and also make into those PR slides with twice the number of cores.

So, I don't think it's feasible to expect there's more performance left in future drivers (the same way that magical expectation wasn't feasible with Terascale or GCN).

DF has a new video out. It's just them going through and reacting to the stream and giving some info we may have missed.

I doubt that Series X comparison is accurate cause it seems too low for the Series X but if so... Jesus

13 RT Teraflops for the Series X, 34 for Turing, 58 for the 3080 using a similar methodology to calculate them. The main reason I doubt it's accurate is a) It's Nvidia saying it? and b) Series X is a different architecture

There other stuff if you want to watch it. Mostly just the hype and confirming the accuracy of the performance uplift in both Rasterization and Ray Tracing

Earlier I was looking for a good comparison for the new CUDA cores. Bulldozer is a great one.

Thanks for the detailed analysis.

Unless you know something the rest of us don't, isn't it a bit early to compare Ampere to RDNA 2.0 given how little we know about the latter? Sure, we know some stuff from mostly the XSX deep dive, but it that enough to compare both architectures before AMD reveals the specs of Big Navi and the rest of the line?

You aren't wrong, and that's why I'd preferred to refer to RDNA 1.0 first and then extrapolate from the info we have (from AMD itself) without claiming exactly how they will stack against each other in terms of performance, let alone ray tracing etc.

Something is fishy about those shaders, so eagerly awaiting for Anandtech's architectural deep dive.

Same. Just with GN as well.

Ok.

We should start putting pressure into AMD until they reveal their new cards .

If we can take a break from the GPU talk for a second....

Zen 3 beta bios seems to be rolling out. Won't be long now!

https://www.gigabyte.com/au/Motherboard/X570-AORUS-MASTER-rev-10/support#support-dl-bios

"Update AGESA ComboV2 1.0.8.1 for New Gen AMD Ryzen processors support"

Keep em coming. The launch of Zen 3 is the only thing I care about right now because it's the one thing that keeps me from building a PC.