Speculation: Ryzen 4000 series/Zen 3

[Panino Manino]

Hey, maybe this virus can benefit AMD?
Looks like smartphone production will slowdown, this must free the fabs a bit, right?

 

[Veradun]

Maybe:

- Desktop September 2020 (first models)​

- Server November 2020​

- Threadripper January 2021​

- APUs March 2021 (last models)​

?

I would switch TR and APU, they're pushing on the laptop market AND there's really no need of a new TR, they can definitely focus on the other products

 

[moinmoin]

I think it's nVidia that uses a "strange" fiscal year

Heh, February to January fiscal year is indeed an odd one. This also moves every quarter by one month, cheeky!

 

[soresu]

I would switch TR and APU, they're pushing on the laptop market AND there's really no need of a new TR, they can definitely focus on the other products

Depends what Zen3 does - if it moves the mhz/watt needle significantly I would definitely buy TR4 over TR3 for TDP drops alone.

 

[Topweasel]

Depends what Zen3 does - if it moves the mhz/watt needle significantly I would definitely buy TR4 over TR3 for TDP drops alone.

That train has left the building. More performance, maybe more clocks, but doubtful we see a tdp reduction.

 

[soresu]

That train has left the building. More performance, maybe more clocks, but doubtful we see a tdp reduction.

Changing those clocks to the TR3 equivalent will achieve the same effect I am looking for (assuming it can be done in a stable way).

You get more of a TDP drop than a clock increase for every node change, and that suits me just fine.

 

[NostaSeronx]

My speculation:
8K Micro-op/L0i cache
4x ALU-AGU pairs and each with a dual-ported unified scheduler (Reduction of scheduler and PRF networking:Zen 6-schedulers, Zen2 5-schedulers, just examples)
4x 128-bit FMAC + Single PRF array (>50%(or 40%?) smaller area at same nominal perf and lower power)

EPYC and Ryzen Mobile with TDP limited versions will potentially see a 1 GHz boost when 4x FMUL+4x FADD occurs.

With that sub-10 watt Zen3 Cezanne will potentially have same or improved clocks to 15w Zen2 Renoir.

Zen2 => 2x 256b FMAC arrays ported as 2x FMUL + 2x FADD(low 128-bit PRF) and 2x FMUL + 2x FADD(high 128-bit PRF)
Zen3 => 1x 512b FMAC array ported as 4x FMAC(128-bit PRF, tagged to 512-bit)
^-- Not sure on the logicistics, but FPU scheduler might be two-leveled?.. 32?-entry four-ported that feeds into 4x??-entry dual-ported which is tied to a FMUL+Bridge unit and a FADD unit.

Zen2 => 4x16 + 1x28 Schedulers w/ 4x ALU+3x AGU
Zen3 => 4x?? Schedulers that operate on dense fused micro-ops rather than individual micro-ops w/ 4x ALU+4x AGU
^-- reduces load-compute-store instruction to operation complexity in Zen. Potentially, comes with a bandwidth increase of 96B/clk in Zen2 to 128B/clk Zen3.

 

[Topweasel]

Changing those clocks to the TR3 equivalent will achieve the same effect I am looking for (assuming it can be done in a stable way).

You get more of a TDP drop than a clock increase for every node change, and that suits me just fine.

That assumes there is a clock increase at all. Might be all IPC increases.

 

[joesiv]

Maybe:

- Desktop September 2020 (first models)​

- Server November 2020​

- Threadripper January 2021​

- APUs March 2021 (last models)​

?

I would say nov for hpc. Basically not for public sale, but 'shipping'.
First set of Consumer cpus r7s and r9s maybe dec 2020
R5s jan 2021
Apus mid 2021

 

[jamescox]

I would switch TR and APU, they're pushing on the laptop market AND there's really no need of a new TR, they can definitely focus on the other products

There is no need for a new TR but there is also very little work in making a new TR once the cpu chip is done. Making a new APU is a lot of work. They have to do a new layout with cpu cores and IO on the same die. New TR parts are lowest priority, but they will generally come out before new APUs.

 

[lobz]

Oh my God.

Launch slides leaked on Videocardz:

AMD launches Ryzen 4000 Mobile Renoir processors with 7nm Vega iGPU - VideoCardz.com

Today AMD launches its high-performance Ryzen 4000H series and low-power Ryzen 4000U series for laptops. AMD Ryzen 4000 Mobile Laptops based on Ryzen 4000 Mobile will go on preorders today, however, it may take few weeks for them to be widely available. AMD releases more information on the...

videocardz.com

4900H -> 8 CUs, 8 Cores, 16 Threads, 4.4 GHz boost, and..... my God...
3.3 GHz BASE CLOCK???
Good luck Comet Lake.

 

[inf64]

Intel is in a very very bad spot and things will get much worse before they get any better. AMD is rubbing salt in the wound with Renoir and price cuts on desktop 3000 series. Zen 3 for desktop will be brutal blow, even Tigerlake on 10nm on desktop with clock parity (highly unlikely to happen) would just get them about equal per core and per clock to Zen 3 desktop models.

 

[lobz]

Intel is in a very very bad spot and things will get much worse before they get any better. AMD is rubbing salt in the wound with Renoir and price cuts on desktop 3000 series. Zen 3 for desktop will be brutal blow, even Tigerlake on 10nm on desktop with clock parity (highly unlikely to happen) would just get them about equal per core and per clock to Zen 3 desktop models.

It's really very unlikely when you take into account that Tiger Lake is not coming to desktop.

 

[DrMrLordX]

It's really very unlikely when you take into account that Tiger Lake is not coming to desktop.

There is a rumour of a TigerLake-H part. 8c/16t. That would be the uh, desktop part I guess.

 

[lobz]

There is a rumour of a TigerLake-H part. 8c/16t. That would be the uh, desktop part I guess.

So basically AMD just launched a uh, 7nm monolithic desktop APU line. Cool =D

 

[DrMrLordX]

So basically AMD just launched a uh, 7nm monolithic desktop APU line. Cool =D

basically they did. That part looks like it could be a desktop APU.

edit: in any case, if Intel does launch 8c Tigerlake-H this year, I believe it will be to compete with Renoir.

 

[eek2121]

Zen 3 is very likely to have not only higher base/boost clocks, but a lower TDP as well. I'm willing to put money on it at this point. Every single indication we've seen, from Financial Analyst day to Renoir and more, indicates that Zen 3 will focus on IPC, clocks, and power optimizations. Renoir alone should give people a taste of things to come. If AMD can get a 45 watt CPU to 4.4 GHz, they can get a 65/95/105 watt CPU to 4.8 GHz. I expect we'll see stronger power management combined with a better boost function that will allow for higher sustained clocks. Shoot, there are people out there that have managed to get the 3950X up to 4.75 GHz on at least one CCX.

EDIT: A 3950X boosting to 4.7 GHz: https://browser.geekbench.com/v5/cpu/1467584.gb5

 

[inf64]

Zen 3 , according to multiple sources will have 15+% IPC boost vs Zen 2. Since current top Renoir models are 3.3 base and 4.3Ghz boost clock, even if Vermeer/whatever it's called for mobile gets the same clocks, we would essentially have a 8C/16T 95W Zen2 (3000 desktop line) performance in 15-25W package, which is mind blowing. Intel would need a Golden Cove core on 7nm to equal or *maybe* beat that performance provided they have clock parity. AMD is just distancing themselves further and further away from intel parts as time goes by.

 

[DrMrLordX]

Who knows when a Zen3-based APU will be available, though? Earliest I would expect it would be Jan 2021.

 

[inf64]

That's true , I was just thinking aloud.Renoir has already set the bar very high, I think intel know they are screwed. Zen3 is close though, some sources claim early summer, some late summer. Whatever it is, it will be brutal.

 

[Markfw]

That's true , I was just think aloud.Renoir has already set the bar very high, I think intel know they are screwed. Zen3 is close though, some sources claim early summer, some late summer. Whatever it is, it will be brutal.

Its not already brutal ? AMD owns server, hedt, and most of desktop, and now with Renoir, the laptop space. They have been screwed for months.

 

[amrnuke]

Zen 3 , according to multiple sources will have 15+% IPC boost vs Zen 2. Since current top Renoir models are 3.3 base and 4.3Ghz boost clock, even if Vermeer/whatever it's called for mobile gets the same clocks, we would essentially have a 8C/16T 95W Zen2 (3000 desktop line) performance in 15-25W package, which is mind blowing. Intel would need a Golden Cove core on 7nm to equal or *maybe* beat that performance provided they have clock parity. AMD is just distancing themselves further and further away from intel parts as time goes by.

I was considering snagging a 3900X at $399, but I may just wait for the 4900X or whatever chip has an oversized L3$ per core for my DC workload if the IPC really does hit a 15% boost and the power draw is essentially the same. An IPC gain of 15% tells me they're probably not going to N7P, but rather N7+, even though it seems like they're trying to backtrack on their slides.

 

[uzzi38]

Who knows when a Zen3-based APU will be available, though? Earliest I would expect it would be Jan 2021.

I doubt anyone but someone from AMD knows the answer to that question, but H1 2021 is probably a safe bet. I wouldn't bet om any kore just yet, the next year looks a bit uncertain all things considered.

 

[CHADBOGA]

I was considering snagging a 3900X at $399, but I may just wait for the 4900X or whatever chip has an oversized L3$ per core for my DC workload if the IPC really does hit a 15% boost and the power draw is essentially the same. An IPC gain of 15% tells me they're probably not going to N7P, but rather N7+, even though it seems like they're trying to backtrack on their slides.

This has been my recent thinking too, but AMD may well stagger the Zen 3 launch and a 4900x may not be amongst the first release of Zen 3 products.

 

[tamz_msc]

An IPC increase of only 15% would be disappointing, in light of what Forrest Norrod has said about Zen 3.