Seriously though, at $700 it's competing with the superior RTX 2080, so that 5000 unit stock shouldn't be an issue at all...Sadly...
https://www.tweaktown.com/news/64501/amd-radeon-vii-less-5000-available-custom-cards/index.html
Gamers Nexus and Linus Tech have both noted about how quiet under-load the Radeon VII was at the AMD CES Booth.
As a further note., the Radeon Instinct MI60 uses 290w (64CU / 1800MHz / 32GB HBM2) while the Radeon Instinct MI50 uses 230w (60CU / 1746MHz / 16GB HBM2).
Why they're both listed as 300w TDP., eh who knows.
Now as for the "It's real world performance is likely about 30% faster than the RTX 2060"... well that's quite debatable.
AMD themselves have stated that it ranged from 25 - 42% Faster than the RX Vega 64., and well depending on the title (and Resolution) the RX Vega 64 can range between RTX 2060 to RTX 2070 (Stock Clock) Performance.
It seems somewhat obvious that Vulkan and DirectX 12 are seeing the largest Performance Enhancements still,. which makes sense given GCN Architecture in general definitely favours an Asynchronous Threaded Pipeline.
Still keep in mind in here it has 7% Fewer Compute Units and only clocked 17.2% Faster... in aggregate, really it should only be ~10% Faster than the RX Vega 64.
Yet if as AMD reports., it's +25% Faster in "Worst Case" (i.e. Unreal Engine 4 / Fortnite)., then there's definitely other factors in regards to said Performance Uplift.
And yes, I know it has double the ROP but that will only affect the Triangles / Second Performance; primarily things such-as Tessellation should be a particularly healthy uplift in performance., just as Gameworks heavy titles.
This said Fornite doesn't really make particularly heavy usage of Gameworks Features... and it's one of the titles where the RX Vega 64 is only equal to the GTX 1070Ti / RTX 2060.
Even still, you're looking at said "Worst Case" for the Radeon VII to be 4K Epic Settings: 61-62 FPS... and that's kind of important to keep in mind.
Remember that the RTX 2080 ($700) is only 25 - 42% Faster than the RTX 2060 ($350)., depending on Resolution; with the biggest gap being at 4K (2160p).
So I mean., it's a similar discrepancy in terms of performance and price.
Still the RTX 2060 is a "Weird" Card in many ways.
Given it's basically a GTX 1070Ti at $50 Lower Price and RTX Feature Support... and in fairness the GTX 1070Ti for $400 was (likely be accident on NVIDIA's part)., a really good price point for performance; at least within the GTX 10-Series Product Stack.
Which is a little funny consideration it was released specifically to have something competitive against the RX Vega 56 at the same MSRP., which the RX Vega 56 has only just actually begun to sell at; while likely due to NVIDIA over-estimating Cryptocurrency Demand for their Enthusiast Products... most of them stayed very close to MSRP.
•
Look as a Pure Gaming Card (which let's be honest the entire RTX 20-Series are) … no, the Radeon VII isn't exactly a great value product.
I mean it's not bad, and it is clearly going to be competitive with NVIDIA's Enthusiast Product Stack; and eh... the RTX 2060 for $350, somewhat nudges it outside of the Mainstream into the Enthusiast Sector.
NVIDIA is somewhat reliant on well DXR (RTX) but also DLSS … the RTX Performance is, "Eh" as a Feature; but DLSS is something that could increase the value through increased performance from Upscaling; but the thing is., and both Hardware Unboxed and Gamers Nexus have done some good breakdowns and analysis on the limited support, quality and performance.
On paper it sounds impressive., but in practise... again, eh. likely going to depend from game-to-game., but for now I definitely prefer Native.
Where I think most of the Cost., and frankly Value... of the Radeon VII is in the 16GB HBM2.
From my perspective in terms of having a GPU for both Gaming and Content Creation., for $700 … yeah not going to lie that's REALLY tempting.
The thing is, I'd heard countless times that the Radeon VII was intended as a Limited Run GPU … and will almost certainly be replaced in June - August with a Navi-Based Variant., which I'd wager will feature 4GB HBM2 + 8GB GDDR6 (12GB Total) for $5-600 but otherwise functionally and performance wise identical.
If AMD are making a "Loss" on the Radeon VII however... eh., that's probably a matter of perspective.
In terms of just Materials and Production Costs; I'd wager there is a decent potential margin on them.
Guess it depends if they ship them to Retailers or (and I think they should) sell them exclusively Direct via the AMD Website.
Honestly I'd actually like to see AMD begin to sell Reference Models via their Website., leaving Retail entirely for AIB Partners.
But eh... I've always preferred the idea of "Own Brand" Product Lines where said IHV has complete control over Quality Assurance.
I mean for all of the unsavoury practises of NVIDIA and Intel., that aspect of them has always be the one thing that I think they've done right and it's weird that it's also the one thing that AMD never really echoed.
More evidence of this: https://www.tomshardware.com/news/amd-radeon-vii-graphics-card-limited-supply,38566.html
The UK is getting less than 100 units with most of it going to overclockers.co.uk, France and Spain are getting 20 units each, but over there with the 20% VAT it's going to cost a whopping $840 which likely explains their small supply. Over here though in the USA we can get a RTX 2080 for $680 (ignore the bad rating, all 4 are from non-purchasers). According to reports Softbank just dumped their entire investment in nVidia worth $3.4 billion, and that's not chump change even for them at this stage, so come tomorrow we may see nVidia axe a bit off the 2080 just to be safe since it is supposedly faster in games.
