• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

AMD Zen 2 (Ryzen 3000) - *** NO COMPETITOR HINTING ***

So why is the Radeon 7 £799 more expensive than a 2080 ?

It’s not faster...

AMD is not offering value here...

I was waiting for that one to pop up. For a start £799 is the mug's price for those who are too impatient to wait for Sapphire stock to show up, so let's keep the argument contained to £650 where it should be.

Radeon VII is a PR stunt to get the world's first 7nm gaming card out of the door, a PR stunt only made possible by Nvidia's price gouging of the RTX line, the woeful implementation and performance of RTX technologies and stagnant raster performance. If any one of these things was different then AMD would not have had an opportunity to repurpose those failed (we suspect) Instinct MI50 packages that were otherwise destined for the trash. As it happens, with a bit of brute force the Radeon VII can match the RTX 2080 in traditional raster performance (once those stupid press drivers are sorted out at least), with the bragging rights of having 16GB worth of 1TB/s VRAM. Out of the blue, the long maligned AMD graphics division has released a card that can finally keep up with an Nvidia product only a few months old.

That's big PR for you. It matters not that those of us watching the actual technology know it's a brute force approach to match 2 year old performance levels, but also it matters not that Nvidia haven't really moved their raster performance either. To the unwashed masses, AMD look like they're back in the mix.

Now you could argue that if AMD were to just dispose those failed packages anyway, they'd still be recouping something if they undercut Nvidia by £100. That's a valid point, but it would (in my belief anyway) cause perception problems further down the road, specifically Navi in the midrange bracket: if AMD can drop 16GB of expensive HBM2 onto a card that retails "as cheaply" as £550, then why would a midrange Navi card with only 8GB of cheap GDDR6 retail "as expensively" as £350**? I think that £650 price point gives Navi some breathing space in terms of specs, but also will paint it in a very positive light when it performs between the outgoing RX Vega 64 and the new Radeon VII for half the price.

But you can't use Radeon VII's price as any gauge or metric for the CPUs, it's irrelevant. The CPUs are cutting edge technology born from a carefully crafted and implemented plan, which includes significantly reduced manufacturing costs compared to what's come before. Radeon VII is a PR stunt cobbled together from failed stock, a crazy man's unfeasible plan and launched because of Nvidia's greed. Ryzen can be priced sensibly, Radeon VII can't.

** I'm throwing £100 onto the leaked Navi price points because they're not going to happen, however much they should happen.
 
I don't know anymore.
After deeper consideration of the facts, I begin to wonder if the first and second generation Ryzen won't be more successful products than the first generation Ryzen with a 7nm chiplet.
The chiplet is 7nm, but why is the IO die on the older 14nm process and not on the newer revised 12nm process?
Or why didn't they make one full chip on 7nm, with the cores and the uncore?
 
The chiplet is 7nm, but why is the IO die on the older 14nm process and not on the newer revised 12nm process?
14nm not being used anymore
12nm used to make APU
7nm for chiplets

as 14nm is sunk cost (ie already have fabs) makes sense commercially to use these fabs to make the IO die

Or why didn't they make one full chip on 7nm, with the cores and the uncore?
scalability
ie: threadripper, epyc
 
Indeed, the chiplet design gives them flexibility and higher yields, which ultimately leads to a lower cost of production across their product line. They have a contract to use GlobalFoundries (they need to use it for a certain percentage of wafers or face penalties if I understand correctly) so using their 14nm process to create the I/O die makes a lot of sense since it apparently doesn't scale well to 7nm anyway (relative to CPU cores and cache, which'll be on the chiplets).
 
After deeper consideration of the facts, I begin to wonder if the first and second generation Ryzen won't be more successful products than the first generation Ryzen with a 7nm chiplet.
The chiplet is 7nm, but why is the IO die on the older 14nm process and not on the newer revised 12nm process?
Or why didn't they make one full chip on 7nm, with the cores and the uncore?

Some things don't scale well, logic circuitry, as found in CPU cores, does. I/O and memory not so much.
 
Indeed, the chiplet design gives them flexibility and higher yields, which ultimately leads to a lower cost of production across their product line. They have a contract to use GlobalFoundries (they need to use it for a certain percentage of wafers or face penalties if I understand correctly) so using their 14nm process to create the I/O die makes a lot of sense since it apparently doesn't scale well to 7nm anyway (relative to CPU cores and cache, which'll be on the chiplets).

For reduction of manufacturing costs it's good but is it equally good for the performance?! We shall see in July.
 
For reduction of manufacturing costs it's good but is it equally good for the performance?! We shall see in July.

Well yeah, if the I/O stuff can't scale down to 7nm then it won't work, will it. So working on a very mature 14nm node vs not working on an immature and overloaded 7nm node?
 
Well yeah, if the I/O stuff can't scale down to 7nm then it won't work, will it. So working on a very mature 14nm node vs not working on an immature and overloaded 7nm node?
It can scale down but just not as well. It's about choosing your battles, i.e. where to focus engineer effort for maximum gain.
 
If the Radeon VII is an £800 card then the 2080 is a £900 card when arbitrarily picking the most expensive model.

Back on topic though, hopefully there will be more information very soon on Zen 2 because I want to know if it's worth upgrading!
 
AMD's 50'th anniversary on the 1'st of May, i wonder if they will have something ready for us by then?
probably the reveal of epyc 2, i guess?
seeing as epyc 2 is rumoured to be released in june, ahead of ryzen 3000...
(according to wccftech, so super mega large pinch of salt needed)
 
Back
Top Bottom