AMD Releases Zen1 and ryzen, which aim to raise the standards for processors...

>AMD Releases Zen1 and ryzen, which aim to raise the standards for processors, featuring much more cores at lower price points
>Intel: Zen is just a glued™ together CPU, lmao
Later...
>AMD Releases Radeon VII, which aim to raise the standards for GPUs, effectively doubling the amount of VRAM for regular GPUs, and helping content creators alot.
>Nvidia: ayy, your new GPU's performance is shit, and our gimmicks can crush it anyday, lmao
DEJA VU
I Have been in this time before
>Also AMD still didn't release NAVI

Attached: images (22).jpg (329x550, 56K)

The moment AMD started kicking Intel in the teeth they went full retard with moar housefires moar chillers and dead on arrival 10nm shit

This will also be the future of Nvidia
Tell me, how is Nvidia gonna cramp in more RT cores into their already monolithic, housefires die?

they'll glue™ the RT cores on. They've been looking into glueing chips together for some while

Wouldn't this introduce latency to the raytracing process?
Or Nvidia going MCM before AMD?

probably, the question is will it matter
you'd need to ask the leather jacket man

Nah, I don't want to deal with more Graphworks™
But in theory, would the raytracing part of GPU, on a separate die, introduce latency to frame rendering with RTX on?

Depends on how it would be connected to the main core I guess. If they'd use an interposer like with HBM memory and a wide-ass bus the latency might be negligible.

The only AMD GPU anywhere close to Turing in efficiency is the VII which isn't even out yet and it's still going to be 30% behind in efficiency. The 590 literally uses 2x the power of its closest competitor, the 1060. AMD tards are legitimately retarded just because Ryzen did well again a fat company that stopped innovating on desktop and data center because they had no competition. Meanwhile in mobile where they were competing against ARM Intel chips still use half the idle power of Ryzen APUs, have better boost behavior, better driver support, and Intel co-developed exclusive tech like the 1W display while AMD will be getting the typical OEM junk using 4W on a 54WHr battery because they're not working with OEMs at all, which is why the Thinkpads and Envy x360 with Ryzen use last gen chassis, hand me downs from Intel

Housefire my ass. The 2080ti uses less power than the VII while delivering 30% higher performance, and that's WITH a process disadvantage.

trips confirm

The Radeon VII is not a big deal, not with that price.
Now Zen 2 is a big deal, it's a quite big deal, specially with intel giving AMD literally 6-8 months of no competition whatsoever.

with 14 times FP64 performance at a fraction the cost, the entire Quadro RTX lineup is going to get fucking murdered on Feb 7.

(vega 2)
Radeon VII - $699 - FP64 - 6912 GFLOPS

(turing)
Quadro RTX 8000 - $10000 - FP64 - 509.8 GFLOPS
Quadro RTX 6000 - $6300 - FP64 - 509.8 GFLOPS
Quadro RTX 5000 - $2300 - FP64 - 348.5 GFLOPS
Quadro RTX 4000 - $900 - FP64 - 246.2 GFLOPS
TITAN RTX - $2500 - FP64 - 509.8 GFLOPS

(volta)
Quadro GV100 - $6299 - FP64 - 8330 GFLOPS
TITAN V - $2999 - FP64 - 7450 GFLOPS

Quadro pays for those leather jackets dummies, not GeForce

Interesting.
I was aware it was nibbling the 5000, but didn't expected the FP64 bomb.

Apparently the current Vega 64 also offers 786.4 GFLOPS of FP64 performance which blows out most of those Turings on your list too. I wonder why Vega isn't discussed as much for a workstation card with that in mind.

Because almost no one needs FP64 outside of like academic scientific computing and those guys are all using CUDA

Guys will begin to write/port to OpenCL if they can get FP64 at such low cost

Maybe Lisa Su and co are taking a loss for every card sold, but I imagine this is necessary to obtain a toehold in scientific computing space

At the end of the the day it's up to the software being optimized for which brand. And the most unfortunate is these software company can't stop sucking Leatherman's dick.

Vega architecture is basically the modern-day GPU rendition of the Bulldozer, design mindset wise since the entire thing is basically stream processors everywhere (like Bulldozers with moar cores except Vega is actually more relevant). If you guys think about it, the higher TDP, more stream processors / core count and a lower clock speed / IPC seems characteristic of the old Bulldozer too, although Vega marketers noted its markedly higher clock speed compared to its predecessors. While we saw first hand that a smaller number of cores on high speed more than just compete well with higher cores at lower speeds, we missed a point that user mentioned.

Vega did start out as workstation too. Its just the RX branding and modification added later on that basically defined it as a gaming GPU. It is better than Fiji and Polaris. Hell, the RX590 on 12nm / 4th gen GCN with 225W tdp gets its ass BTFO by V56 on 14nm / 5th gen GCN with 210W tdp. So on that note, since Vega uses DSBR, I have reasons to believe its energy efficiency is on par with Maxwell arch.

I just want that juicy Radeon VII FP64 in my gaming rig to mine Gridcoin to find the cure for cancer thus ending this board for good

OpenCL is deprecated, if you are writing new apps you should use ROCM, which has source compatibility with CUDA.

>Vega architecture is basically the modern-day GPU rendition of the Bulldozer

You mean GCN. Not much has changed for GCN