Discussion Speculation: Zen 4 (EPYC 4 "Genoa", Ryzen 7000, etc.)

[Vattila]

Except for the details about the improvements in the microarchitecture, we now know pretty well what to expect with Zen 3.

The leaked presentation by AMD Senior Manager Martin Hilgeman shows that EPYC 3 "Milan" will, as promised and expected, reuse the current platform (SP3), and the system architecture and packaging looks to be the same, with the same 9-die chiplet design and the same maximum core and thread-count (no SMT-4, contrary to rumour). The biggest change revealed so far is the enlargement of the compute complex from 4 cores to 8 cores, all sharing a larger L3 cache ("32+ MB", likely to double to 64 MB, I think).

Hilgeman's slides did also show that EPYC 4 "Genoa" is in the definition phase (or was at the time of the presentation in September, at least), and will come with a new platform (SP5), with new memory support (likely DDR5).

What else do you think we will see with Zen 4? PCI-Express 5 support? Increased core-count? 4-way SMT? New packaging (interposer, 2.5D, 3D)? Integrated memory on package (HBM)?

Vote in the poll and share your thoughts!

 

[eek2121]

Neither of those nodes are available out in the wild and in the case of Intel, def not in the hands of any 3rd party.

How you can say such thing is beyond me. Nodes are more than just about density and Intel can parade 2-2,4x as much they want, we know Intel 4 will be garbage. Why else would they need to follow it up/replace it with Intel 3 immediately after?

I also think some people need to look up when was the last time a Intel node was used in a low power or power efficient design (or worse, in a successful/innovative one)... I can wait. Even Samsung probably beats them here.

Last but not least, they seem more expensive than TSMC too. Even tho capacity may not be a problem for them. Intel Alchemist anyone?


Also, what's up with focusing on what Pat's saying? I don't trust 8 out of 10 things Lisa says, despite her great track record and non-petty attitude. Let alone those 2 other clowns.

The specs for both were leaked a while ago.

 

[eek2121]

My 5600 is manually set (1.35v) voltage. I run it at 4.7ghz locked in 24x7 and it uses a maximum 75watts running Cinebench or gaming. In Ryzen master the PPT is either 18-19% of 395w.

I cannot see how the 7600x goes above 80-85w power consumption. It's only a 6 core 12 thread CPU. It does not take as much juice as the 8 core 16 thread parts.

Out of curiosity, has a “safe” voltage ever been established for Zen 3? IIRC Zen 1 max recommended was 1.35V.

 

[yuri69]

Also..

32 Core ThreadRipper 5975WX vs Fastest 32 Core Genoa so far

PEGATRON Lightning vs LENOVO 30E0CTO1WW - Geekbench

browser.geekbench.com

...this is a Linux vs Windows situation. So not really comparable results.

This Genoa sample was frequency-locked at 3.427GHz. The TR score was obtained at 4.2-4.3GHz.

 

[Hans Gruber]

Out of curiosity, has a “safe” voltage ever been established for Zen 3? IIRC Zen 1 max recommended was 1.35V.

I think the safe voltage is anything that keeps your temps below 95C. My Zen 2 3600 needed 1.4v for 4200mhz locked speeds. When people run their stuff on auto everything. The performance is good but significantly less. Opinions differ on this way of doing things. For me under-volting the CPU never never resulted in better performance.

If you run Zen CPU's at stock settings. You will see voltages go as high as 1.5v under heavy load on Zen 2 and probably Zen 3. The voltages bounce around a lot running stock. The higher the clock speed the higher the voltage.

 

[igor_kavinski]

If you run Zen CPU's at stock settings. You will see voltages go as high as 1.5v under heavy load on Zen 2 and probably Zen 3.

You have the 5600. What voltage is it running at, the temps and how high have you clocked the RAM?

 

[Hans Gruber]

You have the 5600. What voltage is it running at, the temps and how high have you clocked the RAM?

I have my 5600 running @ 1.35v in bios with a 47 multiplier. VRM stuff (LLC) is set to medium. CPU will always read the CPU speed at 4700mhz. When you run stock CPU-Z, Zen Master will have voltages as low as .80v to 1.55v when gaming. It bounces around a lot. With a locked multiplier, there is no boost, it's just 4700mhz and with a locked vcore, it runs at 1.35v all the time. When running all core Cinebench, it will go up to 1.4v. So even using manual settings, AMD overrides those settings when extra power is needed.

47.5 or 48 multiplier will crash running Cinebench all core but will run single core just fine. In gaming, 4800mhz will crash on games like BF5 but after many hours of gaming.

The soft wall is 4700mhz on the 5600 pretty much like the 5600x. So the B2 stepping is pretty much identical to the BO steppings on the 5600x. I guess it could be better but the silicon for the 5600x is a better bin than the 5600. The CPU's identical when you OC the 5600.

My memory is a Micron E-die 3000mhz kit 2x 16GB (32GB) dual rank. I run it 3800mhz with 16-19-16-16-36-58 with a 1T command rate @1.41v. Rock solid memory setup, never crashes.

AMD has created a very reliable Zen platform over the years.

 

[PJVol]

Out of curiosity, has a “safe” voltage ever been established for Zen 3?

How you define safe then?
Amd cpus voltage is a function of various operating metrics, such as temp, power, current, number of active cores, etc.
If we talking about reliability, then the max safe voltage at the certain point defined by keeping the actual FIT value below the FIT limit, where FIT itself is a function of voltage and temperature with coefficients derived from process node characteristics.
So the higher t° the less max reliable voltage.

 

[DrMrLordX]

I think the safe voltage is anything that keeps your temps below 95C.

Not necessarily the case. It has more to do with the current draw. While current draw and reported temp will be related, it's probably safer to know how much current you plan to draw with your chosen workloads and apply voltage accordingly.

 

[igor_kavinski]

it's probably safer to know how much current you plan to draw with your chosen workloads and apply voltage accordingly.

How do you do it with your Ryzen? Safest current and voltage for you in your chosen workload?

 

[DrMrLordX]

How do you do it with your Ryzen? Safest current and voltage for you in your chosen workload?

According to The_Stilt and the FIT tables he referenced on his 3900X sample, he found a maximum safe voltage of ~1.325v. Which if I recall was the max safe voltage for high-current workloads involving all cores. The same CPU could reach voltages as high as 1.5v in ST workloads. Those voltages varied from chip to chip so it was just an approximation.

After that you need to pay attention to the TDC during any given workload.

I haven't heard of anyone dumping FIT table data on a 5900X or 5950X (or 3950X) so I don't know "safe" voltage approximations created on that basis. But if you can hunt down The_Stilt on OCN or someplace like that, he might tell you if you ask nicely. Or others there may already have the information.

 

[Just Benching]

With official Raptor Lake power limits being raised to 253w, he may want to start focusing on the products and stop talking crap about the competition. I know the quote is old. It is just ridiculous that every new Intel chip consumes more and more power and perf/watt goes lower and lower.

Anyways, has AMD released any details related to overclocking or PBO improvements?

And that is not true at all. The 13900k will be way more efficient than the 12900k. It wont even be close actually. It wasnt true last gen either, 12900k had the same power limit as the 11900k and the 10900k,yet performance was 70+% higher. So what are you talking about?

Also there will be a 13900t version at 35w, it will be the most efficient cpu on planet earth, by far.

The only one that actually increased powet consumption this time around was amd.

 

[Hans Gruber]

And that is not true at all. The 13900k will be way more efficient than the 12900k. It wont even be close actually. It wasnt true last gen either, 12900k had the same power limit as the 11900k and the 10900k,yet performance was 70+% higher. So what are you talking about?

Also there will be a 13900t version at 35w, it will be the most efficient cpu on planet earth, by far.

The only one that actually increased powet consumption this time around was amd.

There is no way Intel can match AMD's power efficiency on Zen 4. That is using the new 5nm process. Intel is improving efficiency with Raptor Lake but they need the 7 nm process to make significant gains in efficiency. That will come next year or in 2024. Remember, Alder Lake is the 1st 10nm processor. Everything before that was 14nm.

 

[Just Benching]

There is no way Intel can match AMD's power efficiency on Zen 4. That is using the new 5nm process. Intel is improving efficiency with Raptor Lake but they need the 7 nm process to make significant gains in efficiency. That will come next year or in 2024. Remember, Alder Lake is the 1st 10nm processor. Everything before that was 14nm.

Really? I think only the 7950x will be decent in efficiency compared to Raptorlake, the rest of the lineup will get beaten - significantly.

Same wattage the 7600x has no chance to win in efficiency against the 13600k or the 13400, and the same goes for the rest of lineup.


Also both the 13700T and the 13900T will be the most efficient cpus by far out of the box. There will be no contest.

 

[LightningZ71]

There is no way Intel can match AMD's power efficiency on Zen 4. That is using the new 5nm process. Intel is improving efficiency with Raptor Lake but they need the 7 nm process to make significant gains in efficiency. That will come next year or in 2024. Remember, Alder Lake is the 1st 10nm processor. Everything before that was 14nm.

Cannon lake, Ice Lake, Tiger Lake and Jasper Lake beg to differ with this statement...

 

[Asterox]

There is no way Intel can match AMD's power efficiency on Zen 4. That is using the new 5nm process. Intel is improving efficiency with Raptor Lake but they need the 7 nm process to make significant gains in efficiency. That will come next year or in 2024. Remember, Alder Lake is the 1st 10nm processor. Everything before that was 14nm.

There is much higher chance, that the Klingons will attack Australia vs Intel Raptor Lake can match Zen 4 in power efficiency.

 

[Just Benching]

There is much higher chance, that the Klingons will attack Australia vs Intel Raptor Lake can match Zen 4 in power efficiency.

You willing to bet money on it? 100$ says 13900t will wipe the 7950x in efficiency out of the box




Take it to PMs.


esquared
Anandtech Forum Director

 

[igor_kavinski]

You willing to bet money on it? 100$ says 13900t will wipe the 7950x in efficiency out of the box

We'll be waiting quite a bit to see that battle. i9-13900T isn't going to be high on Intel's list of priorities. We might see it maybe six months after launch.

You also need to specify efficiency in ST or MT or both. MT, there is a good chance due to more cores but ST, that's gonna take a miracle. A T-series CPU won't be able to boost as high as a K series one.

 

[nicalandia]

You willing to bet money on it? 100$ says 13900t will wipe the 7950x in efficiency out of the box

At what cost though? It's not going to beat it on performance metric.

 

[Just Benching]

We'll be waiting quite a bit to see that battle. i9-13900T isn't going to be high on Intel's list of priorities. We might see it maybe six months after launch.

Okay, we can compare the 7600x with the 13600k in efficiency, it's not going to be pretty. AMD needs 3nm to compete with Intel's aging 10 apparently

 

[Just Benching]

At what cost though? It's not going to beat it on performance metric.

But the zen 4 will beat raptor lake in both performance and efficiency?

I mean we already know the 13600k will be 80%+ faster than the 7600x in MT, and unless its going to consume twice as much power (which most probably won't), it will be both much faster and more efficient. So I guess klingons are attacking australia soon

 

[nicalandia]

But the zen 4 will beat raptor lake in both performance and efficiency?

Absolutely on both counts at ISO Threads

 

[Just Benching]

Absolutely on both counts at ISO Threads

Uh, iso threads, yes yes. Very interresting. So basically you want to compare the 200$ intel model to the 400$ amd model. I see, very fair comparison, you must not be an amd fanboy, no way.

Anyways, the 13900T will be the most efficiency cpu on planet, followed by the 13700t, out of the box, hands down. Now we have to wait for the klingons

User insults and inflammatory statements (e.g. AMD fanboy) are not allowed.

AT Mod Usandthem

 

[Saylick]

Uh, iso threads, yes yes. Very interresting. So basically you want to compare the 200$ intel model to the 400$ amd model. I see, very fair comparison, you must not be an amd fanboy, no way.

Anyways, the 13900T will be the most efficiency cpu on planet, followed by the 13700t, out of the box, hands down. Now we have to wait for the klingons

Bold words.

 

[nicalandia]

Uh, iso threads, yes yes. Very interresting. So basically you want to compare the 200$ intel model to the 400$ amd model. I see, very fair comparison, you must not be an amd fanboy, no way.

AMD 32T CPU will beat/Match the 32T 13900K at gaming/ST/MT while being more efficient at stock(Both CPUs at stock)

Also no name calling(Would not like to report you again)

 

[Saylick]

AMD 32T CPU will beat/Match the 32T 13900K at gaming/ST/MT while being more efficient at stock(Both CPUs at stock)

Also no name calling(Would not like to report you again)

lol @ non-descript AMD 32T model. Trying to leave the V-cache variant as the ace up your sleeve, eh?

But I agree. I think it's going to be hard for Intel to match AMD on a perf/W basis given the node advantage that Zen 4 enjoys.