I can't believe it will premere with less than the 2990wx. I would say 48core tops, 24 core min. Then 64 core later.
I can't believe it will premere with less than the 2990wx. I would say 48core tops, 24 core min. Then 64 core later.
I can't help but imagine that this is a function of AMD's counter-intuitive binning strategy of more cores == higher clock-speeds. The 3900X still not being anywhere close to meeting demand (even though there are 2 cores disabled per chiplet!), the 3950X delay, the boost behaviour thing, and Thread Ripper 3 starting at only 24 cores all together just makes it sound like AMD grossly mis-predicted yields for high end parts. They can get a lot of technically usable dies, but not many with the characteristics they're looking at for their high end parts.
Does Samsung even have a node that can perform well enough for Zen2 right now?
Pretty sure their 7nm contains some EUV.
But Lisa did say core count would go up and the *80 chipset and platform seems a bit wasted if it stops at 24 cores.
These parts will start at 24 cores.
There is no way AMD will disable SMT on anything besides the low end Ryzens. This isn't low end, its just low power, and no removal of the X has ever meant a feature difference so far.
The X is just market setting name. It doesn't have anything to do with xfr and Zen 2 stuff doesn't really have XFR. Chances are the x has more to do with "extreme" then anything else.The X on all Ryzen CPUs has meant that it has full XFR instead of the half that the non-X versions have, nothing should change.
In the 2000 series, the X did mean better XFR, and PBO. In the 3000 series it just means a little better bin, and a higher stock TDP.The X is just market setting name. It doesn't have anything to do with xfr and Zen 2 stuff doesn't really have XFR. Chances are the x has more to do with "extreme" then anything else.
No, it meant high performance. Always did just that high performance also included (with XFR). But never specifying how much XFR, and non X CPUs included XFR. The if anything the roles are reversed XFR gets its name from being associated with X series chips.In the 2000 series, the X did mean better XFR, and PBO. In the 3000 series it just means a little better bin, and a higher stock TDP.
That being said, GF should have just kept licensing nodes from Samsung. They did good work adapting 14nm LPP to 12nm. If they had licensed 7nm LPP, they could iterate upon it and serve as a second-source for AMD products instead of being the also-ran they are today.
Samsung continuing to license to GF is not beneficial to Samsung especially as it is a much more established player now, certainly not for any reasonable (or realistic) costs that GF would or could pay.
IMO this is a bit over the top and unfair to Intel for the time being, as Ice Lake surely still outperforms Picasso. This would be a whole lot more true when talking about Renoir.The 15-inch Surface Laptop 3 comes with an AMD Ryzen Surface Edition processor. Panay called it the fastest processor that AMD has ever created, “bar none.” It’s also the fastest processor in its class, Panay claimed.
BTW, Microsoft announced new Surface laptops, also having AMD options.
Dispite these currently using Picasso (12nm APU), this announcement has Renoir (AMD's 7nm laptop APU) written all over it, once it becomes available. The reasons to think so are numerous (and even Ryan hints at some):
- Even Intel versions have no Thunderbolt (just USB-C)
- Rumoured upcoming 6 and 8 core versions (Ice Lake has no equivalent, why not Comet Lake? - see below)
- The thin chassis with no GPU and only choosing Ice-Lake for Intel (as there is no room provisioned for a separate GPU)
- Most importantly: They took all the time design and to validate these for AMD processors, despite the current-gen Ryzen being slightly slower than Ice Lake in both multithreaded and single-threaded workloads and only having marginally faster GPU than Ice-Lake with LPDDR4X-3733 memory. Yet, selling at the same price (if this is all they planned, AMDs models would surely be the cheaper option).
- Microsoft clearly draws attention to the Ryzen chips (IMO even being unfairly to Intel for now, as 12nm certainly isnt better than 10nm Ice lake):
IMO this is a bit over the top and unfair to Intel for the time being, as Ice Lake surely still outperforms Picasso. This would be a whole lot more true when talking about Renoir.
This in turn only feeds my suspicions, that Renoir was the product initially planned for GloFo 7nm process for the end of this year (AMD mentioned, they had one for that timeframe that got transferred to TSMC) and these announced Picasso chips are stop-gap replacements. After all, if Renoir was planned for TSMC 7nm all along 2020Q1, there is no reason why it couldn't have been Navi from the start (instead of a 20CU Vega).
TL;DR:
Really excited for a LPDDR4X-3733 equipped 20CU and 6-8 core Renoir Surface laptop. It would crush Ice Lake in both MT and GPU workloads and be reaaaallly close in ST.
Ooh so tasty! If 3500X is $150 or less, @ Newegg , on Oct. 5, I'm going to try to snag one... maybe two. We'll see.If the information is correct, lounch date for Ryzen 5 3500/Ryzen 5 3500X is October 5.
Would price-performance, for workloads like PrimeGrid (that may actually benefit throughput by disabling SMT or running only 6 threads, such that each thread can take full advantage of each core's pipelines, being highly-optimized code, be such that a 3500X w/32MB L3 wouldn't lose much to a 3600, and thus could be a better deal @ $150?