Never ceases to amaze me how dedicated you can be to finding the very worst rumors about AMD products and posting them here under the false pretense of "reporting the news".
The new Vega is the same speed as a Fury X and 980 Ti. Nice try.
What clue is there that this is Vega? I don't see anything at all that would indicate that.
Doesn't Vega have tons of HPC dedicated transistors and features not present with Polaris? That would account for the less efficient die usage factor. And when you say "doesn't clock as high as Polaris," which iteration of Polaris are you referring? Clearly we see that clocking up Polaris with the GTX 580 had pronounced negative effects on perf/w. But what if Vega is 25% more efficient that RX 480 (which is more efficient than the "tuned up" RX 580)? That'll translate into still being about 25% less efficient than GTX 1080 TI, which means that at the same performance the average power consumption will be ~290 watts. With spikes and peaks, power consumption would be well north of 300 watts and would be spec-breaking.
People on here or rant and complain about other people making a big deal about efficiency fail to see what happens when dies and core clocks are scaled up. Unless AMD wants something hotter and louder than GTX 480 while simultaneously breaking the PCIe spec for a single card, they need to play within the confines of their efficiency limitations.
https://forum.beyond3d.com/posts/1979293/What clue is there that this is Vega? I don't see anything at all that would indicate that.
I know. A little bit scary that we have an entry this close to the actual launch not outperforming 1 year old GP104, with rumors of Volta this year. But it could be a lower clocked ES with pre-release drivers, if that makes you feel better.
If they are aiming to do a hard launch in June, then the hardware really needs to be finalized now. Of course this could be an old ES and there is certainly time to tweak the drivers.
However if a Vega @ 1200MHz performs at ~1070ish levels, then a 1500MHz Vega (frequency implied by Mi25) would be about at 1080ish levels.
If they are aiming to do a hard launch in June, then the hardware really needs to be finalized now. Of course this could be an old ES and there is certainly time to tweak the drivers.
However if a Vega @ 1200MHz performs at ~1070ish levels, then a 1500MHz Vega (frequency implied by Mi25) would be about at 1080ish levels.
That's the oddest thing about it for me. If this is really what we're getting, then AMD could have done a similar, but cheaper and quicker job just by shrinkig Fiji. And it probably wouldn't be as big a die as Vega.That level of performance would be in line with a simple die shrunk 16nm Fiji with a 200mhz bump on core. Really doesn't make sense to me it would take this long for that level of performance.
If these rumors are correct it seems to suggest AMD is either sandbagging or incompetent.
I guess we shall wait and see.
That level of performance would be in line with a simple die shrunk 16nm Fiji with a 200mhz bump on core. Really doesn't make sense to me it would take this long for that level of performance.
If these rumors are correct it seems to suggest AMD is either sandbagging or incompetent.
I guess we shall wait and see.
That's the oddest thing about it for me. If this is really what we're getting, then AMD could have done a similar, but cheaper and quicker job just by shrinkig Fiji. And it probably wouldn't be as big a die as Vega.
There is packed math and HBCC - whatever it does exactly. But I'm not sure they'd make Vega worth it over a theoretical die shrunk Fiji. Especially one given Polaris' improvements.
We already know from Fiji that a 600mm^2, 275W TBP, HBM, water cooled GPU is profitable at $649.
This is mine GTX1070 oc result
What makes you think Fiji was profitable, at least in the Fury branding? Just because a GPU is released as a product doesn't mean it's profitable.
I suspect the only profit they made on Fiji was in it's professional market incarnations, not the Fury branding. The 980ti rained on their parade and cause them to launch it cheaper than intended, IMO.
The similarly sized 980TI was also priced at the same level.
While there is a cost difference between 6GB of GDDR5 and 4GB of HBM, the shorter Fury board vs the larger 980TI, and differences in R&D, NV still made plenty of profits on it. Which suggests that boards larger than 500mm^2 make business sense at prices below $700.
You're forgetting all the R&D invested into HBM. It's also not just the memory modules, it's the TSV's and the interposer. Raja admitted in several interviews that HBM is very expensive. He also referenced the expense of getting the whole supply chain up and running, since graphics card manufacturers had not yet standardized HBM rollout/implementation etc.
It's unwise to assume AMD made much profit on Fury pro or FuryX cards. Regardless, they sold extremely poorly, as backed up by the limited data we have available, (steam hardware survey, whilst only a small %, is all we have).
And this is the purported Vega compared to my Fury X @1100 accompanied by a Core2Quad Q9450 @3.52Ghz. If you look at the Graphics score (5721 vs 5467), that's less than direct clock scaling even when ignoring the wildly outdated system of my PC.This is mine GTX1070 oc result
http://www.3dmark.com/spy/888033
GTX1070 oc is around 20% faster(GPU score)
This is all reminding me of lead up to the Ryzen launch. AMD, having learnt from previous hype-train derailments, releases insufficient information to extrapolate the performance of the final product. With little to go on but tenuous leaks, the usual suspects try to convince us that it will be a flop, perhaps even a regression in performance. The stage is set for those of us without a vested interest to be pleasantly surprised.
HBM2. 700MHz (1400MHz effective) and 2048bit bus.
This chip definitely isn't GDDR based from that clockspeed.
AMD also showed Ryzen winning vs Intel many times. Where did they show Vega beating Nvidia?