Speculation: Ryzen 4000 series/Zen 3

[Campy]

Yeah, gonna need a link for that. Because it is big news.

One of those slides shows Ryzen 3000 series not being supported on 300 series motherboards too, so perhaps don't read too much into it.

 

[amrnuke]

One of those slides shows Ryzen 3000 series not being supported on 300 series motherboards too, so perhaps don't read too much into it.

It's official support. I know the mobo manufacturers can work around it as they have for 300 series and Zen2, but it's a bit of a blow to those of us who would like to rely on some warrantied-ish ability to use a 4700X on a B450 for instance.

 

[Markfw]

Living under a rock? AMD said today that x370 and x470 motherboards are not supported.

As was said, its about motherboard support. I have a 3900x running on a x370 Taichi, no problem.

 

[Valantar]

They might be decoupled but there is still a limit to how high IF can clock in 1:1 mode, and 4266 MHz LPDDR4x doesn't actually run the IF at 2133 MHz - it runs at 1:2 mode at that speed.

Decoupled means decoupled, as in "not connected". As such the "ratio" for Renoir only exists in that any two numbers can be made into a ratio. Renoir as such never runs at a 1:1 ratio as there is no such mechanism in place in the system. Renoir can scale its IF clock up/down entirely unrelated to DRAM clock. This is necessary to support IF power saving alongside DDR4 (which doesn't vary its clock speed whatsoever) and LPDDR4X (which varies its clock speed, but goes far too high for IF to follow suit reliably - 1:1 goes to 3600-3800 (1800-1900 FCLOCK) on a good desktop chip, so 4233 (2133 FCLOCK) isn't happening). The issue is that mismatched speeds introduces latency in the data path as a signal coming from one bus moves onto the other, causing performance inconsistency - but obviously not enough for it to seriously hamper Renoir. And who knows, maybe hitting 3:1 around (~4800) with (LP)DDR5 will be better than some weird non-integer ratio?

 

[Shivansps]

Personally i belive that if the CPU is not EOL and there is no technnical issues, support MUST be given, And the fact is APU 3000s and CPU 2000s arent EOL... there is no excuses for having worse compatibility than X570... Ryzen 1000 and APU 2000 are EOL so i can accept that.

What will likely happen is that OEM will support everything just to avoid this confusing mess, but AMD had no reason to drop support like this.

 

[CHADBOGA]

Personally i belive that if the CPU is not EOL and there is no technnical issues, support MUST be given, And the fact is APU 3000s and CPU 2000s arent EOL... there is no excuses for having worse compatibility than X570... Ryzen 1000 and APU 2000 are EOL so i can accept that.

What will likely happen is that OEM will support everything just to avoid this confusing mess, but AMD had no reason to drop support like this.

I wonder if PCI 4 is what may cause the OEM's to not offer the support for Ryzen 4000, as they did for Ryzen 3000?

 

[DrMrLordX]

Some of the older boards have large firmware ROMs. Thoes are the ones most-likely to have support for Zen3.

Personally i belive that if the CPU is not EOL and there is no technnical issues, support MUST be given,

Yes you've been very clear about that. Many, many times.

I wonder if PCI 4 is what may cause the OEM's to not offer the support for Ryzen 4000, as they did for Ryzen 3000?

I think we won't see beta UEFI revisions that try to support it like we did prior to the Matisse launch.

 

[maddie]

Some of the older boards have large firmware ROMs. Thoes are the ones most-likely to have support for Zen3.



Yes you've been very clear about that. Many, many times.



I think we won't see beta UEFI revisions that try to support it like we did prior to the Matisse launch.

I suppose you can always write a bios that removes support for Zen1. Not good for new sales however.

 

[DrMrLordX]

I suppose you can always write a bios that removes support for Zen1. Not good for new sales however.

There were some OEMs that did that already for Zen2. Or at least they removed support for Raven Ridge in some cases.

 

[Avalon]

There were some OEMs that did that already for Zen2. Or at least they removed support for Raven Ridge in some cases.

Yeah, for my Gigabyte X370 Gaming K3 board, they removed support for Bristol Ridge to be able to accept Zen 2 3000 series processors. Gives me a bit of hope that they'll be able to update further to support the Zen 3 4000 series. I'm still rocking my trusty Ryzen 5 1600, so at some point later this year it's going to come down to a decision of....can I drop in a Zen 3, do I just drop in a Zen 2 3000 series chip, or do I get a new motherboard and Zen 3?

 

[DrMrLordX]

@Avalon

I expect them to kill support for Bristol Ridge, Raven Ridge, and Picasso first, depending on the board. Maybe Summit Ridge as well. x570 already doesn't support Raven; Bristol; or Summit Ridge, so they may knock off Picasso and Pinnacle Ridge.

 

[Thibsie]

TUM APISAK leaked what looks like AM4 Renoir.

https://twitter.com/i/web/status/1258371834984058884

Sorry if already posted.

 

[DrMrLordX]

@Thibsie

News to me. Looks like they're holding back AM4 Renoir for the Vermeer launch.

 

[Valantar]

@Thibsie

News to me. Looks like they're holding back AM4 Renoir for the Vermeer launch.

... Or they are going forward exactly like they did last year with no word on desktop APUs until after the summer despite laptop ones being announced at CES. If Vermeer doesn't launch until late Q3 or even Q4 there is little reason to think APUs will be held back until then (unless they are facing very serious production shortages (which there shouldn't be now that Apple has left TSMC's 7nm node) and laptop demand is absolutely overwhelming).

 

[soresu]

@Thibsie

News to me. Looks like they're holding back AM4 Renoir for the Vermeer launch.

Makes sense, they did so for Picasso and Matisse + RDNA1 too.

 

[DrMrLordX]

... Or they are going forward exactly like they did last year with no word on desktop APUs until after the summer despite laptop ones being announced at CES. If Vermeer doesn't launch until late Q3 or even Q4 there is little reason to think APUs will be held back until then (unless they are facing very serious production shortages (which there shouldn't be now that Apple has left TSMC's 7nm node) and laptop demand is absolutely overwhelming).

Vermeer should coincide with the B550 launch. At this point it's looking like July-September, and the pessimist in me is thinking September.

 

[soresu]

... Or they are going forward exactly like they did last year with no word on desktop APUs until after the summer despite laptop ones being announced at CES. If Vermeer doesn't launch until late Q3 or even Q4 there is little reason to think APUs will be held back until then (unless they are facing very serious production shortages (which there shouldn't be now that Apple has left TSMC's 7nm node) and laptop demand is absolutely overwhelming).

It makes sense to hold back on desktop APU's so that they can press the advantage Renoir has at present by concentrating on laptop.

 

[soresu]

Vermeer should coincide with the B550 launch.

Another 4-5 months for an already long overdue product? Ouch.

I'm not sure this is the case though, as they discussed B550 being Zen3 compatible, but not at the moment BIOS wise - which implies to me that it is currently being readied for deploying on cheaper Matisse rigs now that their cheaper Matisse variants have been announced with 3100 and 3300X.

 

[DrMrLordX]

@soresu

Hey if they want to launch B550 early let them, I have no problem with that. I don't think they've done that before except with A320 (which launched maybe 6 months before Ryzen).

 

[Valantar]

@soresu

Hey if they want to launch B550 early let them, I have no problem with that. I don't think they've done that before except with A320 (which launched maybe 6 months before Ryzen).

Given that B550 "technically" launched yesterday (though no boards have been announced yet) it would be utterly unheard of for boards to not arrive until late Q3 or Q4 alongside Vermeer. I would expect B550 boards to trickle out as summer arrives, with the weird launch likely being due to various OEMs being hit by Covid-19 in different ways, necessitating a weird, staggered paper launch.

It makes sense to hold back on desktop APU's so that they can press the advantage Renoir has at present by concentrating on laptop.

I agree that it makes some sense, but considering how small the consumer desktop APU market is the impact of allocating a few thousand dice to the desktop would be minimal. A quick and dirty dice-per-wafer calculation based on Anandtech's measurements gives ~384 total Renoir dice per 300mm wafer (including defects), so they really don't need to set aside many wafers to cover initial global consumer desktop APU demand should they want to appear as if they care about this market. Not to mention that desktop APUs could use chips that are otherwise too power hungry to fit 15W or 45W mobile bins. After all, these chips have been in mass production since late '19 so even if the roll-out in laptops is slow, that doesn't mean AMD doesn't have a significant stock of chips.

 

[jpiniero]

I agree that it makes some sense, but considering how small the consumer desktop APU market is

For DIY it is small, but very much not as a whole. Getting OEMs to use AMD's APUs is another story. And they have shown they are fine with using mobile parts.

 

[Valantar]

For DIY it is small, but very much not as a whole. Getting OEMs to use AMD's APUs is another story. And they have shown they are fine with using mobile parts.

As I said above, while that does happen, it isn't common, and it's more expensive and less flexible for OEMs - so they tend to avoid it if possible in normal form factors. But I was talking specifically DIY consumer sales, not OEM sales (of course OEM SFF like the tiny Dells and Lenovos out there universally use laptop parts, but those also rely exclusively on bespoke motherboards and generally don't have PCIe expansion). I doubt there's enough of a market for consumer facing OEM dekstops based on APUs (or enough attention paid to them) that AMD would prioritize those over the rather loud DIY market for something like this.

 

[DrMrLordX]

@Valantar

Q3 is only 2 months from now (July). If B550 launches in July, you think that would be abnormal? Granted, it was supposed to be out in July of LAST year, but we all know why that didn't happen . . .

 

[coercitiv]

TUM APISAK leaked what looks like AM4 Renoir.

https://twitter.com/i/web/status/1258371834984058884

A bit more apparently from _rogame:

https://twitter.com/i/web/status/1258758712984064002

Two chips are being tested
> 3.5GHz CPU, 1750MHz iGPU
> 3.0GHz CPU, 1750MHz iGPU (likely the 8C Eng Sample: 100-000000149-40_40/30_Y from Userbenchmark)

I believe both are 8C

 

[jpiniero]

Hmm, was thinking about it, AMD's probably raising prices on Zen 3 at same core counts. Could just slide the Renior AM4 APUs into the same price points as Matisse and release it right around when Vermeer launches.