AMD Likely To Announce Vega 20 with 32 GB HBM2 Radeon Instinct at Computex
From the looks of it, AMD will be announcing Vega 20 on Computex. Vega20 is a die shrink of the current model, and should be a 7nm part. It's mentioned that the new Radeon Instinct SKU (machine- deep-learning / enterprise GPU) would get a whopping 32 GB of HBM2 memory.
AMD itself has mentioned the existence of VEGA0 multiple ties away and earlier on mentioning that it would be launching in 218, according to their plans. According to a source of tweak town who spread this information, Vega 20 initially would just be a refresh of Vega 10 for the professional market. It is expected to get the same 64CUs as in Vega 10 and is fabbed at 7nm. Early this year it was already spotted in driver entries as well as a benchmark entry back in April, which also shows 32GB of graphics memory.
There's no word on anything for the consumer (gaming) side. For consumers, 7nm video cards base on Navi seem more viable. AMD has a press-conference this Wednesday.
source: TweakTown
Senior Member
Posts: 2393
Joined: 2016-08-01
When you take into account the fact that 14nmLPP was designed for low clocks and mobile parts, produced low yields (Which contributed to higher voltages) and Vega in general had 40% more compute units than Pascal, you easily start to see why people were 'disappointed' with Vega.
As an architecture, Vega is actually ok... and achieves parity with Pascal on lower frequencies... but the manuf. process it was made on was not its friend. When clocked to 1200MhZ on core, Vega is quite power efficient.
The 14nmLPP is also what prevented Ryzen from clocking much higher than 4Ghz.
Had Vega been made on 16nm TSMC process, it would likely be able to clock much higher on core and HBM with much lower power draw and effectively surpass Pascal in performance.
Considering that 7nm TSMC process allows 35-40% increase in performance over 16nm TSMC ,you have to bear in mind that for Vega, this would translate to much higher performance gains and far lower power draw.
I still expect Vega to draw a bit more power due to having more compute units... but nowhere near the power draw as it did on 14nmLPP even with increased clocks.
You need to pay more attention to manuf. processes and how they affect clocks, efficiency, yields, etc.
Also, hype has nothing to do with this, because leaked benchmarks of Vega 20 on 7nm suggests 3% higher performance on 7nm while running on much lower clocks in comparison to Vega 64.
a) yes indeed when undervolted vega is not tragic power efficiency wise
b) cuda cores and gcn cores are not comparable 1 to 1 in any case
c)traditionally in raw compute performance amd seems to be always packing more compute perfomance ...quite few years now
but that never gave advantage to amd since the cards on paper where going toe to toe with much lower "in raw compute perfomance with nvidia"
d) you do not have 7nm chips in your hand ... i don't none here does ...we do not know in the end how they will perform and how much better .... stop presenting speculations as something that's guaranty
e)the 1000 mhz benchmark that came out and put it close to the vega that is on the market now ... says nothing because when a card is not supported on the benchmark the frequency read is wrong .... so it can be running at 500 ...1000 or 5000 mhz we do not know ...
f) you said ( Considering that 7nm TSMC process allows 35-40% increase in performance over 16nm TSMC ) over what ? performance per watt ? still no products out .... but mind you performance /watt ratio is not raw performance increase .
and to finish ... i really wish every single thing you said comes true ,hell i hope too i look forward to see the gpu arena on fire once again ! but you have to keep in mind over hyping hurts amd ... most of what you said are speculation and linger on the best case scenario up to wishful thinking
Member
Posts: 23
Joined: 2018-02-05
When you take into account the fact that 14nmLPP was designed for low clocks and mobile parts, produced low yields (Which contributed to higher voltages) and Vega in general had 40% more compute units than Pascal, you easily start to see why people were 'disappointed' with Vega.
As an architecture, Vega is actually ok... and achieves parity with Pascal on lower frequencies... but the manuf. process it was made on was not its friend. When clocked to 1200MhZ on core, Vega is quite power efficient.
The 14nmLPP is also what prevented Ryzen from clocking much higher than 4Ghz.
Had Vega been made on 16nm TSMC process, it would likely be able to clock much higher on core and HBM with much lower power draw and effectively surpass Pascal in performance.
Considering that 7nm TSMC process allows 35-40% increase in performance over 16nm TSMC ,you have to bear in mind that for Vega, this would translate to much higher performance gains and far lower power draw.
I still expect Vega to draw a bit more power due to having more compute units... but nowhere near the power draw as it did on 14nmLPP even with increased clocks.
You need to pay more attention to manuf. processes and how they affect clocks, efficiency, yields, etc.
Also, hype has nothing to do with this, because leaked benchmarks of Vega 20 on 7nm suggests 3% higher performance on 7nm while running on much lower clocks in comparison to Vega 64.