Speculation: Ryzen 4000 series/Zen 3

[Hitman928]

Fiiiiiiiiiiiiiiiiiiiiine

I still think AMD is slacking. Just because the competition is stumbling doesn't mean they can afford to be lazy. Maybe they pulled in more features (like TAGE for Zen2) for Zen3 so it won't sting so much once it shows up.

AMD is iterating (so far) at a much faster pace than both intel and ARM competitors (where they compete). I don't know how long they can sustain this pace, but if they keep it up, their competitors are going to have an extremely difficult time ever catching up. We'll see what happens.

 

[uzzi38]

Fiiiiiiiiiiiiiiiiiiiiine

I still think AMD is slacking. Just because the competition is stumbling doesn't mean they can afford to be lazy. Maybe they pulled in more features (like TAGE for Zen2) for Zen3 so it won't sting so much once it shows up.

They're not slacking out of choice. I'm not certain of this, but supposedly they've had to easen up on some of the roadmaps, mostly due to trying to speed up future cadence between uArchs.

Yeah, even AMD felt they were going too slow. In the last few days, I've started feeling they might be as well the way things are going. The next couple of years will be interesting ones as the entire industry really picks up the pace.

 

[moinmoin]

They're not slacking out of choice. I'm not certain of this, but supposedly they've had to easen up on some of the roadmaps, mostly due to trying to speed up future cadence between uArchs.

Yeah, even AMD felt they were going too slow. In the last few days, I've started feeling they might be as well the way things are going. The next couple of years will be interesting ones as the entire industry really picks up the pace.

Their partners not keeping up the pace (like GloFo dropping out of the 7nm race, ASMedia failing to deliver a PCIe 4 chipset on time, board partners taking longer to deliver the platform for the chips etc.) likely slows down AMD as well. In fact that should be an advantage Intel should clearly have over AMD, alas...

 

[DisEnchantment]

View attachment 21433

The top row of paintings are drawn by Vermeer, Cezanne and Rembrandt in that order.

Not sure why Gigabyte would include these paintings in the press kit, but they did.

I have some difficulty wrapping my head around this, as to why this would be even in Press deck from Gigabyte.
It is also not well done and polished, but OK could pass.

However, it is not too out of this world to wonder if something else is at play here
We got price cuts across the board for Ryzen 3K , at least in Mindfactory

Taxes inclusive @19%
3600 - 169 EUR
3600X - 195 EUR
3700X - 288 EUR
3800X - 319 EUR
3900X - 424 EUR. (This is 125 EUR less than what I paid for)

Loads of new rumors like the last time around.

AMD 10-core Ryzen 2800X in response to Intel Core i9-9900K?

Do you remember when Ryzen 2000 (Zen+) launched with the flagship processor being the Ryzen 7 2700X? We all felt that was a little weird, as the previous generation had a Ryzen 7 1800X. In my 2700X re...

www.guru3d.com

 

[uzzi38]

I have some difficulty wrapping my head around this, as to why this would be even in Press deck from Gigabyte.
It is also not well done and polished, but OK could pass.

You'll have to trust me on that one, it's in the slide deck. It's weird AF, but it's there for sure.

 

[DisEnchantment]

You'll have to trust me on that one, it's in the slide deck. It's weird AF, but it's there for sure.

I don't doubt that. That is why I find it strange.
R3 3300X/3100 came out of the blue but we got leaks of 3850X/3750X is what makes me wonder.

 

[.vodka]

I have some difficulty wrapping my head around this, as to why this would be even in Press deck from Gigabyte.
It is also not well done and polished, but OK could pass.

It's not the first time we've had some weird press decks that looked like they were done in MS paint by someone whose blood has turned into coffee racing against the clock.

So they've confirmed Cezanne as the rumored name for the Zen3+RDNA2 APU... now what the hell could Rembrandt be?

 

[Ajay]

You'll have to trust me on that one, it's in the slide deck. It's weird AF, but it's there for sure.

That is not a professional slide deck. No way, no how. "Checks calendar" - nope, not April 1st.

 

[uzzi38]

That is not a professional slide deck. No way, no how. "Checks calendar" - nope, not April 1st.

I'm not sure how else to put it. I have a copy, it's real.

 

[DisEnchantment]

I'm not sure how else to put it. I have a copy, it's real.

I believe you.
Its just that the fact that the slide is in the press deck is baffling.

Somehow the 3850X reminds me of the old 2800X rumor when a lot of people were waiting for Zen2 while 9900K launched is what I am wondering.

 

[amrnuke]

I'm not sure how else to put it. I have a copy, it's real.

It looks like it was created in Google Slides on a Chromebook, exported to pptx, imported to Keynote on a Mac, exported to ods, then imported to PowerPoint on a Windows machine that only has Times New Roman font installed.

 

[jpiniero]

BIOSTAR leaks AMD Ryzen 4000G "Renoir" APUs - VideoCardz.com

BIOSTAR has just confirmed Ryzen 4000G series. AMD Ryzen 4000G clock speeds and OPN codes leaked In a surprising leak for the late Friday, Biostar leaked AMD’s upcoming Renoir APU series codenamed Renoir AM4. The series feature Ryzen 4000G and GE processors, which are 65W and 35W respectively...

videocardz.com

Videocardz has (some of?) the Renoir desktop SKU list. Same core counts as Matisse at the same branding level.

 

[soresu]

It looks like it was created in Google Slides on a Chromebook, exported to pptx, imported to Keynote on a Mac, exported to ods, then imported to PowerPoint on a Windows machine that only has Times New Roman font installed.

Is that anything like the Star Wars prequels translated to Chinese and back again through Google translate?

 

[IntelUser2000]

I still think AMD is slacking. Just because the competition is stumbling doesn't mean they can afford to be lazy. Maybe they pulled in more features (like TAGE for Zen2) for Zen3 so it won't sting so much once it shows up.

Intel/AMD wants the cadence to be a year, but in reality its closer to 13 months. Any little setback adds to that. 15 months is not bad at all for today's complex CPUs.

 

[Valantar]

Fiiiiiiiiiiiiiiiiiiiiine

I still think AMD is slacking. Just because the competition is stumbling doesn't mean they can afford to be lazy. Maybe they pulled in more features (like TAGE for Zen2) for Zen3 so it won't sting so much once it shows up.

Whatever the case may be, I can guarantee you that "slacking" is not a reason for the launch cycle being more than 12 months. That would be a gross underestimation of the complexity of designing a CPU. Also, I would much rather have a slightly slower launch cycle with a fully realized design than some rushed, half-baked design with features cut or disabled just to hit a deadline; wouldn't you? Why do we need a perfect 12-month cycle? Does the world end if we can't upgrade our CPUs every year?

 

[DrMrLordX]

Their partners not keeping up the pace (like GloFo dropping out of the 7nm race, ASMedia failing to deliver a PCIe 4 chipset on time, board partners taking longer to deliver the platform for the chips etc.) likely slows down AMD as well. In fact that should be an advantage Intel should clearly have over AMD, alas...

Plus now AMD and the board partners have to maintain separate UEFI betas for old boards to include support for Vermeer.

now what the hell could Rembrandt be?

Dali and Van Gogh's roommate.

Does the world end if we can't upgrade our CPUs every year?

Wrong perspective. Does AMD's world end if they can't maintain their cadence? Look at what has happened to Intel. Anyone can fail, at any time.

 

[moinmoin]

Does AMD's world end if they can't maintain their cadence? Look at what has happened to Intel. Anyone can fail, at any time.

Intel still does maintain a cadence. Just the wrong kind, minor updates to conceal that the actual intended improvements didn't pan out as planned.

 

[DrMrLordX]

Intel still does maintain a cadence.

Actually, they botched Comet Lake-S. They were off by 5-6 months. The intended release was Dec. 2019, not May 2020.

 

[Valantar]

Actually, they botched Comet Lake-S. They were off by 5-6 months. The intended release was Dec. 2019, not May 2020.

Yep, and now Rocket Lake-S is rumored for release around 6 months after CML-S (with recent leaks not being very promising in terms of performance). They have the cadence of a drunk trying to play the maracas.

 

[Ajay]

Yep, and now Rocket Lake-S is rumored for release around 6 months after CML-S (with recent leaks not being very promising in terms of performance). They have the cadence of a drunk trying to play the maracas.

I only saw one leak on RKL-S, that was pretty lame - but one sample doesn't mean much in the grand scheme of things.

 

[DrMrLordX]

Yeah we don't know what Zen3's competition will be like. Not entirely. Outside of Comet of course.

 

[eek2121]

Step back and look at the efficiency gains going from Matisse to Renoir. Yes the chips have less cache (among other small differences), but they added a 2.1GHz GPU without increasing TDP or decreasing clocks.

Why is this important? I mentioned in previous posts that I strongly believed that AMD would be focused on efficiency improvements. Zen 3 will have even higher clocks than Zen 2. Zen 3 will also perform better at gaming and certain other workloads thanks to a number of architectural improvements. People claiming Zen 3 won’t have higher clocks are nuts. The only know ES leaked already has higher clocks. The max boost clock appears to be 50MHz lower, but the boost curve has shifted upwards. That is on an early ES!

 

[Valantar]

Step back and look at the efficiency gains going from Matisse to Renoir. Yes the chips have less cache (among other small differences), but they added a 2.1GHz GPU without increasing TDP or decreasing clocks.

Why is this important? I mentioned in previous posts that I strongly believed that AMD would be focused on efficiency improvements. Zen 3 will have even higher clocks than Zen 2. Zen 3 will also perform better at gaming and certain other workloads thanks to a number of architectural improvements. People claiming Zen 3 won’t have higher clocks are nuts. The only know ES leaked already has higher clocks. The max boost clock appears to be 50MHz lower, but the boost curve has shifted upwards. That is on an early ES!

Base clock typically doesn't apply when the iGPU is under load in a TDP-limited scenario, so that argument doesn't stand up to scrutiny. We have no idea what CPU clocks Renoir on the desktop can maintain when the iGPU is being stressed.

 

[moinmoin]

Step back and look at the efficiency gains going from Matisse to Renoir.

Those are efficiency gains in the uncore, not the core (aside smaller cache). The improvements can't be applied verbatim to the IOD since it's cut down compared to the others.

 

[ksec]

Fiiiiiiiiiiiiiiiiiiiiine

I still think AMD is slacking. Just because the competition is stumbling doesn't mean they can afford to be lazy. Maybe they pulled in more features (like TAGE for Zen2) for Zen3 so it won't sting so much once it shows up.

I have a similar feeling on AMD not working hard enough on their sales. ( As discussed in some of my previous post how Intel is not dead and shown by both Intel and AMD's quarterly report )

But I have a self serving reason that might explain this.

AMD has always planned Zen 3 to be their big year. Dr Lisa Su has said it many times where Zen 1 and Zen 2 were announced and prove to the market they can do it. Zen 3 will be the key. So they are taking their time to test it out rather than rushing it.

They have also learned their lessons with regards to Enterprise and HyperScaler. AWS as of this moment is *still* testing Zen 2 and not rolling out yet. ( No idea why all the engineers refuse to talk more about it )

So there are many lessons learned in Zen 2 in terms of execution. Hopefully they will do a lot better with Zen 3, along with much more 7nm capacity. The only problem is ASML has stopped shipping EUV machine for a few months now. That is basically the whole industry capacity set back by at least a quarter if not more.