Speculation: Ryzen 4000 series/Zen 3

[exquisitechar]

It would not surprise me if at some point in the lifecycle of Zen 3 that top CPU does 5ghz. I say this because of the Zen 2 XT CPU's. I think the XT max boost clocks will be the starting point of Zen 3.

Ever since Igor leaked the 4.9GHz ES, I've been certain that they will make a 5.0GHz boost part. How often we will see it run at that frequency, though, I'm not sure.

 

[mikk]

That's what you hope 😉

This is what these guys said in the video - if you like it or not.

 

[inf64]

This is what these guys said in the video - if you like it or not.

I can tell you right away that Zen3 will absolutely pummel in the ground RocketLake. Higher IPC, roughly similar clocks, 2x the core count, lower power draw. It will be a sad year for intel.

 

[mikk]

I can tell you right away that Zen3 will absolutely pummel in the ground RocketLake. Higher IPC, roughly similar clocks, 2x the core count, lower power draw. It will be a sad year for intel.

Completely out of context post but I guess the hype must go on and corrections like this are undesired in this topic. But speaking about next year, you are shouting at Rocketlake but deliberately ignored Alder Lake-S which is planned for H2 2021.

 

[reqq]

150w tdp?? Sheit i have one of those bad x570 vrm boards.. i wonder if it can cope with that?

 

[inf64]

Completely out of context post but I guess the hype must go on and corrections like this are undesired in this topic. But speaking about next year, you are shouting at Rocketlake but deliberately ignored Alder Lake-S which is planned for H2 2021.

I said a year (from now). Alderlake, if it launches in 2021 (big IF) will be going against 16C/32T Zen3 in 2021 and 16C+ Zen4 in H1 2022. It's screwed anyway since it's a hybrid chip (8+8 big small) which will need all the help it can get to even compete with Zen3, let alone Zen4. It will probably win by a bit in ST versus Zen3 but it will be yet another year of sadness for intel when Zen4 comes soon after AL.

 

[mikk]

I said a year (from now). Alderlake, if it launches in 2021 (big IF) will be going against 16C/32T Zen3 in 2021 and 16C+ Zen4 in H1 2022. It's screwed anyway since it's a hybrid chip (8+8 big small) which will need all the help it can get to even compete with Zen3, let alone Zen4. It will probably win by a bit in ST versus Zen3 but it will be yet another year of sadness for intel when Zen4 comes soon after AL.

Nope, this is what you said:

It will be a sad year for intel.

Basically everything what you are saying is based on speculations from your side. AMD related you are overly optimistic and Intel related you are overly pessimistic.

 

[inf64]

Nope, this is what you said:



Basically everything what you are saying is based on speculations from your side. AMD related you are overly optimistic and Intel related you are overly pessimistic.

Alright, we'll see then how it goes for intel. We have these posts for reference purposes. I look forward to how next year unfolds, the time is ripe for giants to fall (deservedly).

 

[Asterox]

150w tdp?? Sheit i have one of those bad x570 vrm boards.. i wonder if it can cope with that?

No it will burn fast, or same scenario as this cheep Gigabyte motherboard.

 

[Nereus77]

Nope, this is what you said:



Basically everything what you are saying is based on speculations from your side. AMD related you are overly optimistic and Intel related you are overly pessimistic.

The writing is on the wall for Intel there, my friend. 2021 will not fare well for them. 2022 might be worse. They are going to be in the back seat for the next 3 years at least, until AMD gets complacent.

Intel is welcome to meet the challenge though, but it will be an uphill battle being two steps behind AMD.

 

[rtxtwt]

This legit?

This channel just upload a new video about R9-5800 vs R9-3800X, IMO this result looks 'realistic' but I don't know if this channel make a fake/clickbait or not.

I've already get used to it that there would be a lot so-call 'leak' when AMD about to release a new product......

these leaker/channel should hand out something like cinebench result other than gaming cuz gaming result is not pure CPU workload that could easy to be fake and have no way to figure out whether they fake it or not.......

 

[uzzi38]

The writing is on the wall for Intel there, my friend. 2021 will not fare well for them. 2022 might be worse. They are going to be in the back seat for the next 3 years at least, until AMD gets complacent.

Intel is welcome to meet the challenge though, but it will be an uphill battle being two steps behind AMD.

Going to disagree on this tbh. Intel have pretty strong offerings to come, the question is whether or not they can volume produce enough to satisfy the entire market. But the product timelines are basically locked in now, products will launch as they're supposed to, even if they end up slightly worse or in lower volume than first intended.

 

[yuri69]

The writing is on the wall for Intel there, my friend. 2021 will not fare well for them. 2022 might be worse. They are going to be in the back seat for the next 3 years at least, until AMD gets complacent.

Intel is welcome to meet the challenge though, but it will be an uphill battle being two steps behind AMD.

So, let's recap Intel's offerings:
* Q4 2020 - Tiger Lake U - better performance than Renoir in graphics and classic workloads; AMD wins only in >4T workloads
* H1 2021 - Tiger Lake H - 8c mobile; this one will remove AMD's advantage in higher TDP notebook segments
* H1 2021 - Rocket Lake-S - over 5GHz Turboing Willow Cove with 8c; it should be faster than Zen3 in most of the workloads - games are unclear due to slow Willow cache; AMD wins in MT due to 16c
* H2 2021 - Alder Lake-S - Golden Cove cores with another 15-20% IPC over Willow Cove; 8c+8c "Atom cores" mixed with an advanced power mngmnt; AMD might win in MT due 16c, although the IPC advantage might cancel this

To me this seems pretty solid for Intel given they can produce 10nm in quantity.

 

[Nereus77]

Going to disagree on this tbh. Intel have pretty strong offerings to come, the question is whether or not they can volume produce enough to satisfy the entire market. But the product timelines are basically locked in now, products will launch as they're supposed to, even if they end up slightly worse or in lower volume than first intended.

Maybe you aren't seeing what I'm seeing...

Chiplet advantage => AMD
Process advantage => AMD
IPC advantage => AMD
Frequency advantage => Intel (for now)
Financial advantage => Intel
Marketshare advantage => Intel (for now)
Market growth advantage => AMD
Current new builder preference => AMD

With Zen 3 having a rumored 15 - 20% IPC increase over Zen 2, Intel is looking like they need to pull a rabbit out of a hat for Rocket Lake and Alder Lake to remain competitive, let alone overtake AMD. Looking at previous generational IPC increases (maybe 5%, sometimes it went backwards) and the reaction to Tiger Lake, I don't think Intel is going to spring any surprises on us anytime soon.

 

[exquisitechar]

So, let's recap Intel's offerings:
* Q4 2020 - Tiger Lake U - better performance than Renoir in graphics and classic workloads; AMD wins only in >4T workloads
* H1 2021 - Tiger Lake H - 8c mobile; this one will remove AMD's advantage in higher TDP notebook segments
* H1 2021 - Rocket Lake-S - over 5GHz Turboing Willow Cove with 8c; it should be faster than Zen3 in most of the workloads - games are unclear due to slow Willow cache; AMD wins in MT due to 16c
* H2 2021 - Alder Lake-S - Golden Cove cores with another 15-20% IPC over Willow Cove; 8c+8c "Atom cores" mixed with an advanced power mngmnt; AMD might win in MT due 16c, although the IPC advantage might cancel this

To me this seems pretty solid for Intel given they can produce 10nm in quantity.

Rocket Lake isn't Willow Cove.

That said, I agree that Intel's offerings are looking very solid for 2021. Alder in particular will be strong in both the desktop and laptop market. AMD needs to get Zen 4 out as soon as possible to compete, Rembrandt won't be good enough.

 

[leoneazzurro]

So, let's recap Intel's offerings:
* Q4 2020 - Tiger Lake U - better performance than Renoir in graphics and classic workloads; AMD wins only in >4T workloads
* H1 2021 - Tiger Lake H - 8c mobile; this one will remove AMD's advantage in higher TDP notebook segments
* H1 2021 - Rocket Lake-S - over 5GHz Turboing Willow Cove with 8c; it should be faster than Zen3 in most of the workloads - games are unclear due to slow Willow cache; AMD wins in MT due to 16c
* H2 2021 - Alder Lake-S - Golden Cove cores with another 15-20% IPC over Willow Cove; 8c+8c "Atom cores" mixed with an advanced power mngmnt; AMD might win in MT due 16c, although the IPC advantage might cancel this

To me this seems pretty solid for Intel given they can produce 10nm in quantity.

You forgot that in 2021 AMD is readying Cezanne and Van Gogh mobile CPus, based on Zen3 and the second with RDNA2 GPU, and even possibly a Renoir refresh for the low end.
How Rocket Lake and Zen3 will position regarding each other is not known, but so far even Willow cove has not demostrated a big IPC advantage relative to Comet Lake, most of Tiger Lake gains came from higher sustained boost speeds - which are less an issue in a desktop environment. Alder Lake - we'll see. But it will also come near the Zen 4 launch window and probably after a Zen3 referesh. So yes, Intel will have goos offerings but it must be seen, if they will vbe really better than competition.

 

[yuri69]

Chiplet advantage => AMD
Process advantage => AMD
IPC advantage => AMD
Frequency advantage => Intel (for now)
Financial advantage => Intel
Marketshare advantage => Intel (for now)
Market growth advantage => AMD
Current new builder preference => AMD

Chiplet advantage => this one is dubious - AMD can scale to high core counts, but the energy efficiency gets notably worse (see Matisse vs lowend Comet Lake); Intel migh jump to multiple dies with Sapphire, who knows
Process => AMD but the new Tiger Lake's 10nm+++++++ is nice too
IPC => It's been Intel since Ice Lake, and now it's still Intel with Willow. Zen 3 might best the Willow but will be beaten by Golden
Frequency => Intel, Rocket ES already boosts to 5.4GHz - this is probably out of reach of Zen 3
Financial => Intel is 10x AMD
Market share => Intel
Growth => AMD, since it's way smaller
Builder preference => completely irrelevant in the big picture since the vast majority is mobile & data center

 

[yuri69]

You forgot that in 2021 AMD is readying Cezanne and Van Gogh mobile CPus, based on Zen3 and the second with RDNA2 GPU, and even possibly a Renoir refresh for the low end.

There is a problem with the three APUs you mentioned since we don't know if one of those will be an exclusive for Apple/Google/M$.

Cezanne should compete well with Tiger in CPU workloads on IPC basis, although we don't know the frequency/power for Zen3 compared to Zen2. Also Cezanne features the same GPU as Renoir, so Tiger beats that.

Lucienne aka Renoir-refresh will gain like 100MHz and probably better firmware for power mngmnt but that's it.

Van Gogh will finally feature a great GPU but we don't know a thing about its Zen 3 Zen 2 CPU config.

 

[soresu]

Van Gogh will finally feature a great GPU but we don't know a thing about its Zen 3 CPU config.

From all current leaks and rumours Van Gogh is Zen2 based.

 

[yuri69]

From all current leaks and rumours Van Gogh is Zen2 based.

Sorry, a stupid typo

 

[Antey]

renoir gpu is already great. the only problem is it has only 8 Compute units. only 20-25 mm² at best. 96 EU Xe takes close to 45 mm². 16 Vega CUs would be amazing and if bandwidth is not a problem it would be much better than intel 96 EU gpu.

 

[soresu]

renoir gpu is already great. the only problem is it has only 8 Compute units. only 20-25 mm² at best. 96 EU Xe takes close to 45 mm². 16 Vega CUs would be amazing and if bandwidth is not a problem it would be much better than intel 96 EU gpu.

I think AMD are concentrating on price/area as much as power here - they seem to have invested far more in getting Renoir to actually get design wins than any previous APU generation, and it seems to have paid off from what I have seen so far.

If Raphael is indeed going APU and leaving pure CPU Zen4 to TR5 then I think we can expect Rembrandt to be more like Renoir in terms of how anaemic the GPU is - ie relatively small, power efficient and cheerful, leaving Raphael to have more oomph in the top end of APU's, like AMD's own more compact Vega M finally.

 

[yuri69]

TBH back in days when AMD's CPU perf, power mngmt, and absolute power were horrid, AMD simply relied on pushing a rather large GPU. GPU was the only viable part of the whole SoC.

Starting with Renoir, the CPU perf, power, and power mngmnt are actually excellent, compared to Picasso and older APUs. So AMD decided they did not need to "play the GPU card" anymore.

 

[jamescox]

Chiplet advantage => this one is dubious - AMD can scale to high core counts, but the energy efficiency gets notably worse (see Matisse vs lowend Comet Lake); Intel migh jump to multiple dies with Sapphire, who knows
Process => AMD but the new Tiger Lake's 10nm+++++++ is nice too
IPC => It's been Intel since Ice Lake, and now it's still Intel with Willow. Zen 3 might best the Willow but will be beaten by Golden
Frequency => Intel, Rocket ES already boosts to 5.4GHz - this is probably out of reach of Zen 3
Financial => Intel is 10x AMD
Market share => Intel
Growth => AMD, since it's way smaller
Builder preference => completely irrelevant in the big picture since the vast majority is mobile & data center

I haven’t paid much attention to what Intel is doing sine it doesn’t seem relevant to my interest. With their process tech so delayed, they are going to have serious difficulty competing. I think Zen 3 is going to perform very well and probably be the top performer across the board on the desktop and Milan is going to extend their lead in almost everything for server/workstation/HPC. Intel may still have AVX512. I don’t think Intel will be able to compete by pushing the clock speed.

So, if Intel manages to get 10 nm actually yielding well, that will be about the time that I expect AMD 5 nm APUs to hit. Is Intel going to be able to compete against a 5 nm Zen 3 + RDNA2 based device with a 10 nm device? Seems unlikely. I think AMD may push their APUs into the desktop market, so we may actually see an AM5 based APU before we get a chiplet based AM5 part. Chiplets are necessary for scaling up to 64 cores, but there is nothing stopping AMD from switching back to monolithic parts for low and mid range desktop parts (maybe up to 8 cores). They could still do a 5 nm chiplet refresh, but that seems kind of unlikely with Zen 4 time frame.

 

[yuri69]

So, if Intel manages to get 10 nm actually yielding well, that will be about the time that I expect AMD 5 nm APUs to hit. Is Intel going to be able to compete against a 5 nm Zen 3 + RDNA2 based device with a 10 nm device? Seems unlikely.

AMD's Zen 4 is scheduled to (late?) 2022. It'd be pretty bad for Intel 10nm to start yielding in 2022, given they plan to release a 7nm-based client CPU in late '22/early '23.