Is this a new leak? I can't even keep them all in order anymore.
>>60252711
Yeah the article was posted only 4 hours ago. Seems to be a new leak. They've done a separate article for every leak so far.
bump for those waiting for vega
>Radeon RX Vega (Vega 10) is set to be a GTX 1080 contender
So it's is slower than GTX 1080ti? The 687F:C3 is faster than the 687F:C1?
Three versions of Vega.
I think AMD is playing shenanigans, Vega 11 still exists and is known to be below Vega 10. If Vega 10 was a 1080 competitor there would be no market for a card between the Vega 10 cutdown and Polaris 20.
Vega 11 vs 1070?
Vega 10 vs 1080 non ti?
Dual Vega 10 vs 1080ti?
Doubt, theres always cutdowns for yields. What everyone thought going in was Vega 11 cutdown below 1070 above Polaris 10/20, Vega 11 between 1070 and 1080; Vega 10 cutdown between 1080 and 1080 Ti, and Vega 10 full between 1080 Ti and Titan Xp.
If Vega 10 does turn out to just be where people think Vega 11 ends up, then what in the fuck is Vega 11.
Vega 10 will be 2*4GB HBM2 and 4096 ALUs, with at least 2 binning variants. Lower ALU count parts might use cheaper, slightly lower clocked HBM.
Vega 11 could be a lot of things, but probably will be 2048-2560 ALUs and GDDR5/GDDR5X.
HBM+interposer is simply too expensive for ~250GB/s parts.
Dude, get this straight
full vega, what is in the pro gpu, at pro mhz, at the very worst amd has ever made a gpu would fall right on top of a 1080, at polaris efficiency it would fall half way between a 1080 and 1080ti being closer to the ti then not.
Amd has had fairly substantial issues with gpu saturation, along with using a hardware scheduler that while it allows for less overhead, incompetent devs just decided lets make games for 1/2 cores and have nvidia sort our shit out.
If amd alleviated its bottlenecks, its going to be a no holds barred rape train of a gpu, and apparently they have been working almost exclusively on bottleneck removing since fury came out, so its possible, but I am going with real world examples, and I assume amd is going to be halfway between 1080 and 1080ti till they come out and show us something, either a higher than pro clock rate or a comparison between fury and vega with raw numbers, not necessarily framerates, just showing us what bottlenecks got alleviated
do you have a figure on the cost?
This gets annoying when people keep saying cost figures without ever giving a cost.
>Remember, the MI25 accelerator is a passively cooled device.
No it's not you curry eating donkey retards
Is there any actual publicly available information either confirming or denying this?
Everyone seems to say it's passively cooled and assumes that means it can run at 100% use constantly without any concern for heat issues but I don't recall AMD every stating it could do so and finding information beyond the original announcement where they just showed clockspeeds is pretty hard.
MI25 just tech demo woodscrew
They're designed to go into a rack with Delta fans blasting air at 3000RPM through them, just like NVIDIA's Tesla accelerators. Retards just think they're 'passive' because they don't have fans.
The Firepro S9300X2 is also 'passive' in this sense but I dare you to try run that dual Fiji (aka Radeon Pro Duo) behemoth passively.
>MI25 just tech demo
I googled that and this was the first result. I'm not entirely sure if it's related to my question though.
tomsguide.com
Alright. That makes a lot more sense than thinking a high end GPU can run at those kinds of clocks with no cooling.
That's not to say workstation chips aren't better binned.
The WX7100 is essentially an RX 480 with a TDP of 130W and a higher base clock.
This is the same card from the same chinese guy that 'leaked' the timespy a few days ago.
It's been in use by liquidsky servers since March.
wish they would release these fucking cards already, i just got a freesync monitor but polaris is still going to choke my 4.4ghz i5