Speculation: Ryzen 4000 series/Zen 3

[xilli_fiberbit]

if I was going with a new Intel build I would have definitely waited for Rocket Lake-S this Fall...even though it's still using 14nm, it's a new architecture plus has PCIe 4 support

Rocket lake is still derivated from skylake and it should be coming not before 2021, alder lake is a new architecture but we haven't recently not very much leaks or news about zen 3. AMD didn't said something since last intel announcements

 

[Staples]

The lack of leaks makes me a little worried.
1) it might be later than we think
2) it is less of an improvement over zen 3xxx than we think

I chose the 10700 over the 3700x because I spend most of my time doing things where it exceeds. It was a little more expensive so I think even if I had gone with the AMD part, the price to performance I would have seen would have been equal.

 

[eek2121]

AMD has confirmed numerous times Zen 3 is coming this year.

 

[teejee]

AMD has confirmed numerous times Zen 3 is coming this year.

But not if it will be a complete range or just a few models.

Some rumours say we'll only see the high-end versions this year.

 

[Kenmitch]

But not if it will be a complete range or just a few models.

Some rumours say we'll only see the high-end versions this year.

I'm with the wait and see what launches crowd. I don't think AMD's specifically stated which sku's will be launched.

I have a 3700x and might consider it's replacement sku as long as the cure the memory divider issue. My b-die is worthy and waiting. Not like 3600 CL14 is slow, but I figure my sticks are good around 4266-4333MHz CL17 or so.

 

[Staples]

If I understand correctly, the G series (I think, the Renior destop) were built on the Zen 3 architecture, just low powered versions. Is this correct? Because if it is, we should be able to draw some conclusions of how much of an improvement Zen 3 is over Zen 2+. But I have not seen anyone suggest this. Am I misunderstanding something?

 

[Thunder 57]

If I understand correctly, the G series (I think, the Renior destop) were built on the Zen 3 architecture, just low powered versions. Is this correct? Because if it is, we should be able to draw some conclusions of how much of an improvement Zen 3 is over Zen 2+. But I have not seen anyone suggest this. Am I misunderstanding something?

You are. AMD's marketing is crap. Renior is 4000G and is based on Zen 2. The non G parts are 3000 series and are also Zen 2. Zen 3 will ship later this year. There was no Zen 2+ either.

 

[A///]

The lack of leaks makes me a little worried.
1) it might be later than we think
2) it is less of an improvement over zen 3xxx than we think

I chose the 10700 over the 3700x because I spend most of my time doing things where it exceeds. It was a little more expensive so I think even if I had gone with the AMD part, the price to performance I would have seen would have been equal.

We also haven't heard much if anything about RDNA2 unless you count the silly rumors by certain vloggers and bloggers whose names I won't drop because it's free advertising to junk data that's useless. I guess RDNA 2 must either be amazing or suck, too. I would rather have AMD be radio silent on everything than have them make big bold claims and be disappointed. OTOH, Intel is now the loud mouthed AMD making bold claims.

 

[A///]

You are. AMD's marketing is crap. Renior is 4000G and is based on Zen 2. The non G parts are 3000 series and are also Zen 2. Zen 3 will ship later this year. There was no Zen 2+ either.

If that rumor of AMD moving to 5000 for their product line is true, it would do them a lot of service. It's not unheard of either. Intel never did a 1000 series of processors. They went from 900 to 2000. NVidia skipped generations a few times in the last 20 years, too.

 

[A///]

November??...say it ain't so...I was hoping for mid-September after Intel has their little Tiger Lake event

If by some black magic AMD has Zen 3 based mobile ready to go by December-February as well as Zen 3 desktop, I could absolutely see them announcing everything the day of Intel's announcement after it concludes. The entire press will be hyper focused on AMD and not Intel. 11th gen mobile, Tigerlake, is set to go against Zen 2 based mobile processors. It's a nasty way to shut Intel out, but I wouldn't put it past modern AMD to do just that. This is all theory and it ties into them being radio silent for the entire year apart from confirmations that Zen 3 will come in 2020. The question to that has always been what parts, Epyc, mobile or desktop, and usually focuses on the 1st of the 3.

This theory is a huge stretch of the imagination and I don't think it's even possible given console orders are high as heck. I'd love to be wrong with my assumption and have my theory proven correct. I expect one of the junk vlogs will pick up this post and somehow misdirect their readers and or viewers into believing it's a credible rumor.

 

[A///]

I was just thinking because zen4 will be on a more expensive PCI5 DDR5 platform (at least for the first year until prices drop)
The last CPU Zen3 on AM4 may have a much longer life than Zen2.

Hence even a launch of Zen3 only a half year before AM5 zen4 would still be an acceptable situation, hence if they would like to delay zen3 to squeeze an extra 5% out of the last AM4 CPU it may not be such a bad idea as it sounds, Zen3 will have a long life anyway on the AM4 platform.

For Epyc any delay would hurt a bit but it's also the end of a socket so it fits the lower bracket budget for some time to come.

I'm not saying they will delay it it's just from a business point of view not as bad as it would seem.

That's unless Zen4 on AM5 managed to increase IPC like Zen2 did because they're able to bring forward technology present in future Zen to what would be Zen4 by then. Putting aside the weird rumors we've heard about Zen3 which are incredibly sparse themselves, if Zen3 manages a nice 10-15% IPC increase and Zen4 manages to provide a 20-25% IPC increase, then the adoption rate would be high compared to Zen3 for those moving on from Zen+ or on Zen2, and I'd argue those on Zen3 may as well upgrade, too, since the performance would be there and new standards made available. This is pure speculation of course.

 

[randomhero]

If by some black magic AMD has Zen 3 based mobile ready to go by December-February as well as Zen 3 desktop, I could absolutely see them announcing everything the day of Intel's announcement after it concludes. The entire press will be hyper focused on AMD and not Intel. 11th gen mobile, Tigerlake, is set to go against Zen 2 based mobile processors. It's a nasty way to shut Intel out, but I wouldn't put it past modern AMD to do just that. This is all theory and it ties into them being radio silent for the entire year apart from confirmations that Zen 3 will come in 2020. The question to that has always been what parts, Epyc, mobile or desktop, and usually focuses on the 1st of the 3.

This theory is a huge stretch of the imagination and I don't think it's even possible given console orders are high as heck. I'd love to be wrong with my assumption and have my theory proven correct. I expect one of the junk vlogs will pick up this post and somehow misdirect their readers and or viewers into believing it's a credible rumor.

I do think you are on to something.
I was quite surprised that cezanne has vega as gpu part and not RDNA2. The only reasonable explanation for that is time to market. So I think they will have Cezanne ready around CES time, and first product on shelves by late Q1 2021.
If true, rumours about 5000 naming for Vermeer and Cezanne also have sense.

We will see soon, whatever the outcome.

 

[Mk pt]

If by some black magic AMD has Zen 3 based mobile ready to go by December-February as well as Zen 3 desktop, I could absolutely see them announcing everything the day of Intel's announcement after it concludes. The entire press will be hyper focused on AMD and not Intel. 11th gen mobile, Tigerlake, is set to go against Zen 2 based mobile processors. It's a nasty way to shut Intel out, but I wouldn't put it past modern AMD to do just that. This is all theory and it ties into them being radio silent for the entire year apart from confirmations that Zen 3 will come in 2020. The question to that has always been what parts, Epyc, mobile or desktop, and usually focuses on the 1st of the 3.

This theory is a huge stretch of the imagination and I don't think it's even possible given console orders are high as heck. I'd love to be wrong with my assumption and have my theory proven correct. I expect one of the junk vlogs will pick up this post and somehow misdirect their readers and or viewers into believing it's a credible rumor.

AMD focus is high margins markets. And that's server/Epyc market first.
We only get Ryzen almost sooner because bad yields of Epyc chiplets can be good Ryzen chiplets.

Mobile requires chiplets with very good yields for lower power consumption, and those in first months of production will go to Epyc.

 

[DrMrLordX]

Mobile requires chiplets with very good yields for lower power consumption, and those in first months of production will go to Epyc.

That's not really true. EPYC doesn't share silicon with any of the mobile Ryzen products. AMD has, at least in the past, been financially limited in what risks they want to take with their new Zen-based designs. You are correct in asserting that the highest margins are in the server market, which is why AMD prioritized their entire design effort towards server/workstation products (EPYC). Lower-margin, high-volume efforts (mobile, low-power) get short shrift. That's been the case up through Renoir. Renoir's adoption appears to have been quite enthusiastic, though, and if OEMs are privately asking AMD to bring Zen3 to the mobile market as quickly as possible, then maybe AMD will reprioritize development efforts.

 

[yuri69]

The lack of leaks makes me a little worried.
1) it might be later than we think
2) it is less of an improvement over zen 3xxx than we think

The hard fact is, there are still no testing/beta BIOSes supporting Zen 3. In Zen 2 case, that beta support apperated in March. This means 4 months prior launch.

Given the Zen 2 to Zen 3 transition means an entirely new family number a lot of underlying infra might have changed so one would expect a longer testing period in this case...

We also haven't heard much if anything about RDNA2 unless you count the silly rumors by certain vloggers and bloggers whose names I won't drop because it's free advertising to junk data that's useless. I guess RDNA 2 must either be amazing or suck, too. I would rather have AMD be radio silent on everything than have them make big bold claims and be disappointed. OTOH, Intel is now the loud mouthed AMD making bold claims.

GPU leaks usually starts when the AIB partners starts completing their cards. Since there are one of those might mean there will initially be only AMD-secured channels - no AIBs involved. This was the situation with Radeon VII - nobody knew it was coming since the job was done by AMD themselves.

Both Navi 21 (Sienna Cichlid) and Navi 22 (Navy Flounder) are being rather actively brought up in open source drivers. This means the release is close.

 

[eek2121]

Once AMD gets done with the initial delivery of consoles, they will have capacity to push hard into next-gen. That is why we haven’t heard anything more than a vague timeframe. It depends on how quickly they crank through the initial order from Microsoft and Sony.

 

[TitusTroy]

people here keep saying late 2020 for Zen 3 but I think it'll be September/October...main reason being that AMD is also releasing a new generation of GPU's in late 2020 and they aren't going to release both Big Navi and Zen 3 in the same month...Big Navi is rumored to be coming in October/November...so it only makes sense for Zen 3 to be released in September

 

[JujuFish]

they aren't going to release both Big Navi and Zen 3 in the same month

On the contrary, I expect them to release both at the same time.

 

[TitusTroy]

On the contrary, I expect them to release both at the same time.

no way...no chance...these are huge releases, you're not going to release both together...you need to give each product its own time in the spotlight

 

[JujuFish]

no way...no chance...these are huge releases, you're not going to release both together...you need to give each product its own time in the spotlight

We'll see.

 

[amd6502]

I do think you are on to something.
I was quite surprised that cezanne has vega as gpu part and not RDNA2. The only reasonable explanation for that is time to market. So I think they will have Cezanne ready around CES time, and first product on shelves by late Q1 2021.

I think it might be ready this October.

 

[randomhero]

To quote zen master:
"We will se."

 

[GoodRevrnd]

no way...no chance...these are huge releases, you're not going to release both together...you need to give each product its own time in the spotlight

You wouldn't want to release Zen 3 and trigger system build hype with RTX 3000 out and no alternative of your own. I bet we get everything in October.

 

[Gideon]

You wouldn't want to release Zen 3 and trigger system build hype with RTX 3000 out and no alternative of your own. I bet we get everything in October.

That's a valid point, didn't think of it this way. Besides we have precedent with the 07/07 launch last year.

If anything they might release the GPU earlier (as people can upgrade it first and then replace the rig) but only when the schedules are off more than a month. Considering Console ramp-up is eating up wafers till at least september/october and it takes more to assemble a GPU (from a working die) they can't really be that far apart.

 

[A///]

I do think you are on to something.
I was quite surprised that cezanne has vega as gpu part and not RDNA2. The only reasonable explanation for that is time to market. So I think they will have Cezanne ready around CES time, and first product on shelves by late Q1 2021.
If true, rumours about 5000 naming for Vermeer and Cezanne also have sense.

We will see soon, whatever the outcome.

Yes, that works, too. Instead of an April-July launch for mobile, if the have it around CES 2021 or even a few weeks earlier, it'll kill whatever momentum Intel had. The Vega GPU on Cezanne is supposedly a more modernized variant, but if we're discussing theories I'd find it funny if AMD lied and actually went with an RDNA or RDNA2 based mobile GPU. Under promise, over deliver... Current Tiger Lake IPC claims put it against the 4700U most of the time. By the time Tiger Lake comes out assuming Intel didn't fudge delivery dates, it'll be over half a year old. Intel may have a short period between that period and Cezanne. However, as I said, it would be incredibly underhanded and clever for AMD to let Intel announce their product stack for 11th gen, and then announce the same day or even the following day. I believe this is their exact plan for NVidia's Ampere release, though they'll give NVidia one to three weeks of fame as they won't have deliverables up yet except to reviewers. I expect the September 1st announcement/paper launch to be followed by a very late or early October product release.