NVIDIA's Next-Generation Ampere GPUs to be 50% Faster than Turing at Half the Power

[CandymanGR]

Trust me, it will not.

 

[InVasMani]

An add-on APU AIC over PCIe would be a terrible idea unless it included modifications to the Windows scheduler that strictly segregated the two chips with no related processes ever crossing between the two. Without that you would have absolutely horrible memory latency issues and other NUMA-related performance issues, just exacerbated by being connected over (for this use) slow PCIe. Remember how 1st and 2nd generation Threadripper struggled to scale due to NUMA issues? It would be that, just multiplied by several orders of magnitude due to the PCIe link latency. It could work as a compute coprocessor or something similar (running its own discrete workloads), but it would be useless for combining with the existing CPU/APU. Scaling would be horrendous.

Technically AMD could build a LPDDR4 and M.2 card right into a infinity fabric bridge. It could actually be quicker for gaming like quad channel LPDDR4 and a pair of raid-0 M.2 slots built into the infinity fabric bridge itself. The socketed CPU could just handle the background OS tasks independent from it running the game off the APU's CPU cores. Which honestly in such a scenario a dual core CPU built into the APU's would suffice that allows it to communicate with the bridge itself perfectly and execute a single core per CPU. Essentially quad core type CPU performance coupled with quad core performance that can cache accelerate the M.2 slots and compress/decompress/encrypt/decrypt on the fly across the bridge. Sort of like they are all networked together with ram disks on each side. The real benifit of course is any the OS stuff going on with the main socketed CPU won't adversely impact game performance. You could pretty much run a CPU stress test off it while gaming on the discrete APU's in theory I think in practice because the games themselves would be run off the APU's. That would of course require some kind of Direct X process aware scheduler segregation though. I believe the other parts with M.2 and compression and encryption though could be performed thru the driver itself and a Linux installation on the M.2 devices themselves that the APU's operate it could be headless for that matter.

 

[Super XP]

I can't imagine that I live in a world where fanbois argue about pre-release specs that come out of the advertising department ... and, on top of that, arguing that their brand's fake specs are all real and the other guys are all fake. Save ya arguing for then the cards are tested. My bet is we just going to see more of the same ...

Mantle was gonna change everything ... it didn't
It did change everything, it forced Microsoft into launching DX12. Mantle also evolved into Vulcan by the Krono's group.

HBM2 was gonna change everything ... it didn't
It did change everything, AMD tested the waters with HBM2, had the balls to do so and learned a great deal on the do's and don'ts. Especially for the enterprise scene. And as another posted pointed out, Nvidia is also utilizing HBM2. Thank You AMD!

7nm was gonna change everything ... it didn't
It did change everything, AMD had the balls to move forward with a new manufacturing process, and learned a great ton of things about it. After years on 7nm, RDNA2 = coming soon. The entire industry is moving to 7nm. lol

What we do know is that the GPU market stopped being competitive with the 7xx versus 2xx series, where nVidia walked away with the top two tiers (all cards overclocked). AMD lost another tier against the 970 and another tier against the 1060. The next generation didn't go well for both sides in some respects ... AMD had to make huge price cuts; nVidia didn't because they didn't have to. The bright shining light to was the 5600 XT, pretty much nothing else got me excited out of AMD ... if they can scale that up into the upper tiers, things may finally get interesting.
AMD was concentrating on its CPU Business, which is where the majority of its revenues come from. As soon as ZEN was launched, AMD was quickly hard at work on RDNA as a stop gap approach, then soon to come RDNA2 as a permanent graphics solution. They can now build on top of RDNA2 moving forward as they've done with ZEN to ZEN+ to ZEN2 and coming soon ZEN3 moving forward.

AMD is far from perfect, and despite its lack to compete in the GPU industry in a higher level they still managed to compete on there level with cost to performance. 2020 & 2021 will be a very crucial growth year for AMD before Nvidia and especially Intel wake up.

Nvidia's Next - Generation Ampere GPUs to be 50% Faster than Turing at Half the Power?
I highly doubt that, more like wishful thinking. I believe RDNA2 forced Nvidia to rush this Ampere, which is probably a slightly tweaked & overclocked Turing design on a new manufacturing process which requires more power.
Its going to be interesting to see Ampere vs. RDNA2 battles coming later this year.