Speculation: Ryzen 4000 series/Zen 3

[Thunder 57]

New information on the Vermeer engineering sample has been published by Igor'sLAB

Translation by VideoCardz

The new information is for the clock speeds, which were decoded from an OPN code Igor was given by the source:

100-000000059-52_ 48/35 _ Y

The 48 stands for 4.8 GHz boost clock, while 35 means 3.5 GHz base clock. It is worth noting that this is still not the final sample, but definitely a faster stepping from the previous report. In June Igor covered first data on the B0 samples which were featuring a boost clock of 4.6 GHz and a base clock of 3.7 GHz.

100-000000059-14_46/37_Y
100-000000059-15_46/37_N

AMD Vermeer to feature Per core voltage adjustment
Intel recently introduced a new feature with the 10th Gen Core series. This feature allowed per-core voltage adjustment for increased optimization and overclocking. This feature is also coming to AMD platform, as the changelog for the ComboAM4v2PI 1.0.6.0 (or 1006) reveals:

View attachment 27613

Very interesting. I'm sure AMD would to get to that magical 5GHz number. But maybe we won't see that until AM5 with DDR5, PCIe 5, Ryzen 5000 series, baked in 5nm, and launching on 5/5 naturally.

 

[Nereus77]

Very interesting. I'm sure AMD would to get to that magical 5GHz number. But maybe we won't see that until AM5 with DDR5, PCIe 5, Ryzen 5000 series, baked in 5nm, and launching on 5/5 naturally.

Put me down for the Ryzen 5 5500 Pentacore

 

[Kenmitch]

Last I checked Walmart doesn't make anything. They dictate cost and pricing from their vendors. What is your point? The term rollback pricing is something AMD should consider.

....And you ask what's my point?

Maybe you should voice your opinion towards those whom make the motherboards. AMD has no control of the end price of a motherboard.

 

[Kenmitch]

Very interesting. I'm sure AMD would to get to that magical 5GHz number. But maybe we won't see that until AM5 with DDR5, PCIe 5, Ryzen 5000 series, baked in 5nm, and launching on 5/5 naturally.

So they'll launch at 5:55AM or PM?

 

[Asterox]

Very interesting. I'm sure AMD would to get to that magical 5GHz number. But maybe we won't see that until AM5 with DDR5, PCIe 5, Ryzen 5000 series, baked in 5nm, and launching on 5/5 naturally.

4.8ghz Singlecore boost is expected.5ghz mantra is blah, "only Intel needs 5ghz to stay relevant in 1080p gaming on 1200$ GPU".

Weird people with 1200$ GPU slap games in 4K, well it could be much worse.

 

[Thunder 57]

So they'll launch at 5:55AM or PM?

That's where the run out of luck. Because the best they could do would be 05:55. Though maybe they combine the 5th day with the 0 in 05:55 to make 50, and say it's AMD's 50th anniversary. Except that was last year .

4.8ghz Singlecore boost is expected.5ghz mantra is blah, "only Intel needs 5ghz to stay relevant in 1080p gaming on 1200$ GPU".

Weird people with 1200$ GPU slap games in 4K, well it could be much worse.

I know that, we probably all do. But at least AMD could bring back that 5GHz tachometer/speedometer video and leave the 5 in it. Great for marketing. Don't forget when it comes to numbers people are stupid. $99? That sounds way better and cheaper than $100! What a deal!

 

[Ajay]

4.8 GHz boost sounds nice, if true. The point that matters most to me is typical boost, as opposed to peak boost. But, yeah, marketing...

 

[Asterox]

4.8 GHz boost sounds nice, if true. The point that matters most to me is typical boost, as opposed to peak boost. But, yeah, marketing...

It is no doubt for one Core. All Core Boost will be hm, around 4.2-4.3ghz no higher for 16/32 CPU.What else do you need, 8 Core CCX+other CPU enhancements or again 15% higher IPC.

 

[Ajay]

It is no doubt for one Core. All Core Boost will be hm, around 4.2-4.3ghz no higher for 16/32 CPU.What else do you need, 8 Core CCX+other CPU enhancements or again 15% higher IPC.

I won't be upgrading, the improvements won't be enough for me to bother (and I'd rather use the same system, sans GPU, for ~5 years to get some value out of it).
If AMD does pull off around 15% higher performance per clock, that would pretty clearly end any claim to Intel ST performance superiority.

 

[Makaveli]

I won't be upgrading, the improvements won't be enough for me to bother (and I'd rather use the same system, sans GPU, for ~5 years to get some value out of it).
If AMD does pull off around 15% higher performance per clock, that would pretty clearly end any claim to Intel ST performance superiority.

If the Zen 3 shows a nice boost paired with a IF of 2000 paired with DDR4000 I would consider it. However I'm only on 8 cores at the moment and would just upgrade my HTPC to the 3800X or sell it. This is also one of the reason's I'm staying at 16GB and not going to 32GB incase I need to buy new memory.

Priority for me right now is GPU upgrade though to RDNA 2 is on the list before cpu upgrade.

 

[Ajay]

Priority for me right now is GPU upgrade though to RDNA 2 is on the list before cpu upgrade.

Same here. Hoping I can pick up something next year, though, unless AMD's drivers are excellent this time around; I'll stick with NV.

 

[BTRY B 529th FA BN]

Ooo, Juicy Rumors

Looking forward to this, have money saved

 

[Asterox]

This will most likely be a top model 16/32 Zen 3 CPU.

Expected price hm, around 800$ for 4.9ghz Singlecore Turbocore boost.

AMD 16-core Zen3 (Ryzen 9 4950X?) engineering sample boosts up to 4.9 GHz - VideoCardz.com

New OPN codes for the upcoming Zen3 architecture have been disclosed. AMD Ryzen 9 4950X could boost up to 4.9 GHz Igor’sLAB has an update to his story from yesterday. Igor has been very actively sharing AMD OPN codes for the upcoming Ryzen 4000 series processors. It all began with 4000G “Renoir...

videocardz.com

 

[Makaveli]

This will most likely be a top model 16/32 Zen 3 CPU.

Expected price hm, around 800$ for 4.9ghz Singlecore Turbocore boost.

AMD 16-core Zen3 (Ryzen 9 4950X?) engineering sample boosts up to 4.9 GHz - VideoCardz.com

New OPN codes for the upcoming Zen3 architecture have been disclosed. AMD Ryzen 9 4950X could boost up to 4.9 GHz Igor’sLAB has an update to his story from yesterday. Igor has been very actively sharing AMD OPN codes for the upcoming Ryzen 4000 series processors. It all began with 4000G “Renoir...

videocardz.com

What was the Launch price for the 3950X around $750 no?

So that price seem reasonable for a part that offers the same core count but is faster.

 

[Kenmitch]

What was the Launch price for the 3950X around $750 no?

So that price seem reasonable for a part that offers the same core count but is faster.

Yep $750 on another note it's currently $670 at MicroCenter

I'm more interested to see how the memory clocks turn out. It would be nice to run my b-die closer to it's true potential without sacrificing latency.

....Waits for reviews or at least some juicy leaks.

 

[Gideon]

These clocks were a nice surprise, i did not expect it to reach 4.9. This might make it really tempting for marketing to do a 5.0GHz Black Edition or something similar.

The added move instructions will help as well, can't wait to see some leaks.

 

[soresu]

So that price seem reasonable for a part that offers the same core count but is faster.

It will be interesting to see what happens to pricing and SKU core counts per segment when the next big shrink comes with Zen 4 and we get another doubling of cores per socket.

I hope 32 cores comes down by at least 1/3, I could live with that and would definitely go for it at £1300-1350.

 

[Ajay]

It will be interesting to see what happens to pricing and SKU core counts per segment when the next big shrink comes with Zen 4 and we get another doubling of cores per socket.

I hope 32 cores comes down by at least 1/3, I could live with that and would definitely go for it at £1300-1350.

Doubt that will happen with Intel sucking so badly.

 

[soresu]

Doubt that will happen with Intel sucking so badly.

Depends how fast they get 10nm more production capable and scale up the core count again, right now they are chugging electrons faster than a Bugati does gas to keep relatively 'even' with AMD on core count.

 

[Markfw]

Depends how fast they get 10nm more production capable and scale up the core count again, right now they are chugging electrons faster than a Bugati does gas to keep relatively 'even' with AMD on core count.

server, 28 vs 64 is relatively even ? Or desptop, 10 is relatively even with 16 ? HEDT, 18 vs 64 ? Where does Intel come close ?

 

[inf64]

If Alder Lake matches Zen 3 16C in ST and MT performance it will be a miracle. I'm not even thinking about Zen4 which will be more like Zen1->Zen2 jump, with 2x the core count and FP throughput (Vs Zen3). It will be brutal 2-3 years for intel.

 

[Ajay]

Depends how fast they get 10nm more production capable and scale up the core count again, right now they are chugging electrons faster than a Bugati does gas to keep relatively 'even' with AMD on core count.

Uh, yeah, when were 10nm Intel products due for release - 2016/2017. Not certain, but, after several years they have only 3 fabs running 10nm. Performance and yields have improved a fair bit for small dice (quad core CPUs). But, for large dice like ICL-SP (supposedly up to 38 cores - which we haven't even seen leaks of yet); seems like Intel will be producing those at a loss for some big customer who needs certain features found in the Ice Lake ISA.

 

[amrnuke]

If Alder Lake matches Zen 3 16C in ST and MT performance it will be a miracle. I'm not even thinking about Zen4 which will be more like Zen1->Zen2 jump, with 2x the core count and FP throughput (Vs Zen3). It will be brutal 2-3 years for intel.

While this is not the thread for Zen4, I have to ask... is Zen4 confirmed to go up to 32 cores / 64 threads on Ryzen line? And 128 cores / 256 threads on the TR side? And 256 cores / 512 threads on EPYC?

 

[Ajay]

While this is not the thread for Zen4, I have to ask... is Zen4 confirmed to go up to 32 cores / 64 threads on Ryzen line? And 128 cores / 256 threads on the TR side? And 256 cores / 512 threads on EPYC?

Not that I've heard. And if I heard that right now, I would be very skeptical that it was anything more than a guess.

 

[soresu]

server, 28 vs 64 is relatively even ?

Eh?

I thought that they had 56 core SKU's?

That's what I meant by relatively even, and even then only in absolute single thread or MT + AVX 512.

I'm not amping up Intel here, actually the opposite - I was meaning that when 10nm matures further that they will scale up beyond 56 to better compete, even if the TDP goes thru the (already burning) roof in the process.