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!

 

[JM Popaleetus]

Just quit now.

In my very first post, I stated "I imagine the 13900KS will be 5-10% better than the K and again about equal in gaming due to brute force."

Probably closer to the 5%. Nonetheless, what part of brute force, and my offhand remark of "when cranked up to 11 with fast RAM" are you not understanding?

roughly 10-15% on average faster than the 13900K in 1080p

That statement was clearly in reference to the X3D chips. I did not intend the latter half of that sentence to imply the KS will ALSO be 10-15% overall faster. I clearly said otherwise prior. I just foresee it being able to trade blows in select circumstances.

You basically said nothing.

Heavily tuned processor is faster than stock processor, news at 11

I'd argue that's what the KS is. A factory heavily tuned processor.

 

[Hail The Brain Slug]

In my very first post, I stated "I imagine the 13900KS will be 5-10% better than the K and again about equal in gaming due to brute force."

Probably closer to the 5%. Nonetheless, what part of brute force, and my offhand remark of "when cranked up to 11 with fast RAM" are you not understanding?



That statement was clearly in reference to the X3D chips. I did not intend the latter half of that sentence to imply the KS will ALSO be 10-15% overall faster. I clearly said otherwise prior. I just foresee it being able to trade blows in select circumstances.

You said both will be 10-15% faster. Nowhere did you say 5-10% for 13900KS.

Also what part of "making a claim about how well a heavily overclocked/tuned product can perform relative to stock products" is a totally meaningless statement don't you understand?

 

[Kocicak]

So they don't seem to have solved the V/F problem for 3D stacked CCD, basically the same 400MHz deficit as with the 5800x3d.
And given the most of the Zen3->Zen4 performance uplift came from higher clock speeds, this doesn't bode well for 7800x3d. Let's see what they squeeze out of the "hybrid approach" in 2CCD models.
All in all, dissapointing so far.

I think that is mostly thermal issue. The dies are just slapped together and the electrical connections are touching each other, there is no solder interface. I think that avoiding too much heat production in the CPU die is there to avoid too much thermal expansion/contraction and corrupting the connections.

 

[Hail The Brain Slug]

I think that it is mostly thermal issue. The dies are just slapped together and the electrical connections are touching each other, there is no solder interface. I think that avoiding too much heat production in the CPU die is there to avoid too much thermal expansion/contraction and corrupting the connections.

I saw somewhere that the stacked die can do up to 1.4V which is an improvement from Zen 3D, which was limited to 1.35V. Not a big improvement, but an improvement nonetheless.

I do see single core boost VID going as high as 1.465v on my 7950X, so it makes sense at 1.4V there is some boost headroom lost. Less headroom lost when compared to 5950X 1.5 VID for 4.9 GHz vs. the 1.35V max for the 5800X3D, so there's probably a fair amount of extra performance to squeeze out with PBO and CO for those willing to put forth that effort.

 

[JM Popaleetus]

Nowhere did you say 5-10% for 13900KS.

Incorrect.

https://forums.anandtech.com/thread...ryzen-7000-etc.2571425/page-462#post-40924362

I imagine the 13900KS will be 5-10% better than the K

Good day.

 

[Hail The Brain Slug]

Incorrect.

https://forums.anandtech.com/thread...ryzen-7000-etc.2571425/page-462#post-40924362



Good day.

I guess if you just want to spam posts with differing info, that's your perogative. I was directly quoting you explicitly saying 10-15%. Denying you said that is on you.

 

[Joe NYC]

So they don't seem to have solved the V/F problem for 3D stacked CCD, basically the same 400MHz deficit as with the 5800x3d.
And given the most of the Zen3->Zen4 performance uplift came from higher clock speeds, this doesn't bode well for 7800x3d. Let's see what they squeeze out of the "hybrid approach" in 2CCD models.
All in all, dissapointing so far.

Tough crowd here. The best gaming CPU is "disappointing".

 

[Kocicak]

Tough crowd here. The best gaming CPU is "disappointing".

What do you mean? 7950X3D is the best. 7800X3D is fully cushioned by the cash, which robs both ST and MT performance from it. Dual chiplet CPUs can still run that naked chiplet quickly not losing any ST or low thread count non gaming performance.

 

[Dave3000]

If the 7900X3D and 7950X3D do end up getting one normal CCD and a 3D V-cached CCD with lower frequency, would a program such as Process Lasso be beneficial for gaming in that situation? In Process Lasso assign the cores of the 3D V-cached CCD to games that benefit more from the larger cache than the higher frequency normal CCD and assign the cores of the normal CCD to games that benefit more from the higher frequency than the larger cache? Still you would have to know much each game on a individual basis benefits from larger L3 cache vs higher frequency for the optimal gaming performance on these two CPU's?

 

[JM Popaleetus]

I guess if you just want to spam posts with differing info, that's your perogative. I was directly quoting you explicitly saying 10-15%. Denying you said that is on you.

You clearly don't want to have an actual discussion, but I made four predictions across two posts.

1. The 7800X3D and 7950X3D will be roughly equal in gaming.
2. The 7800 X3D and 7950X3D will average 10-15% faster than the 13900K, again in gaming. And obviously when not GPU bound.
3. The 13900KS will be 5-10% faster than the K.
4. The KS may trade blows with the X3D when "when cranked up to 11 with fast RAM".

The third I agree will be closer to the 5%. The fourth, I thought it was clear I meant at 720P in certain games, with DDR5-8000 and TDP be damned.

I very well may have worded my posts initial poorly, but you're the one continuing to harp on the latter-half of that one post as if it was gospel.

So again, good day. This was a fun way to occupy the waste of time that was my Zoom meeting.

 

[Exist50]

Tough crowd here. The best gaming CPU is "disappointing".

I think, as usual, people got over-inflated expectations from wild rumors. Remember when 3D V-Cache was supposed to make it 37% faster, with the frequency penalty solved and even better scaling than Zen 3? #11,092

You can find comments from literally yesterday along the same vein. No surprise there's some whiplash once reality sets in.

 

[PJVol]

Tough crowd here. The best gaming CPU is "disappointing".

1. Let's see the benches first (rdna3 slides didn't add much trust to today's nice pictures)
2. I don't give a damn **** which CPU is best for gaming or how much it will cost and whatever.
The only thing I interested in is technology challenges and how good are engineers at solving them. So, yes... a tough crowd here in a TECH forum from the average consumer point of view.

Remember when 3D V-Cache was supposed to make it 37% faster, with the frequency penalty solved

Aside from an obvious 37% exaggeration, maybe you've got some insider info pointing to unreasonable expectations of “frequency penalty” improvement?

 

[Kaluan]

Here's a 7950X3D productivity performance v 13900K slide that (to my knowledge) wasn't shown at CES:

 

[Kocicak]

Slide showing only positive numbers is suspitious.

 

[Hail The Brain Slug]

You clearly don't want to have an actual discussion, but I made four predictions across two posts.

1. The 7800X3D and 7950X3D will be roughly equal in gaming.
2. The 7800 X3D and 7950X3D will average 10-15% faster than the 13900K, again in gaming. And obviously when not GPU bound.
3. The 13900KS will be 5-10% faster than the K.
4. The KS may trade blows with the X3D when "when cranked up to 11 with fast RAM".

The third I agree will be closer to the 5%. The fourth, I thought it was clear I meant at 720P in certain games, with DDR5-8000 and TDP be damned.

I very well may have worded my posts initial poorly, but you're the one continuing to harp on the latter-half of that one post as if it was gospel.

So again, good day. This was a fun way to occupy the waste of time that was my Zoom meeting.

Alright, I see how I misread what you had written in that post to think you were indicating that "both" meant Zen4 3D + 13900k, instead of the 7800X3D and 7950X3D that you indicated. My apologies.

 

[Kaluan]

Interestingly, 7000X3D SKUs have a Tjmax value of 89C

 

[Hans Gruber]

So what you guys are saying. They didn't fix the 3d v-cache temperature limit with the Zen 4 3D version. So clock speeds will be 400-500mhz lower than standard 7700x/7800x CPU's?

 

[Hail The Brain Slug]

So what you guys are saying. They didn't fix the 3d v-cache temperature limit with the Zen 4 3D version. So clock speeds will be 400-500mhz lower than standard 7700x/7800x CPU's?

Unsure if it's thermals, voltage, or a combination of both. I mentioned a few posts ago I heard the Zen 4 3D CCD is limited to 1.4V, which is higher than Zen 3D 1.35 VMax. There were rumors they refined the entire setup to improve thermals, which will be tested when we see reviews and samples in the wild. The fact they are willing to allow PBO and CO is a promising sign, as is the increased voltage. I think it seems like it's been meaningfully improved over Zen 3D, albeit not fully "fixed", as in made equivalent to a non-stacked die. I don't know if stacking can ever be made fully equivalent to a non stacked die for thermals and voltage tolerance, though.

 

[Markfw]

Slide showing only positive numbers is suspitious.

Why, because your favorite CPU isn't winning ? Without all the cache, the 7950x is a monster at multi-threaded apps, but at single threaded apps fell behind once in a while. This fixes that.

 

[Kocicak]

My favourite CPU at the moment is AMD 5700G...

 

[Hail The Brain Slug]

Why, because your favorite CPU isn't winning ? Without all the cache, the 7950x is a monster at multi-threaded apps, but at single threaded apps fell behind once in a while. This fixes that.

I think we all got used to AMD demonstrating a reasonably accurate suite of performance benchmarks, which frequently included ties or losses, until the RDNA3 announcement where it appears the vast majority of what they demonstrated was false.

I can understand the breach of trust at their last announcement might make some people more skeptical of their claims for the near future. That is, unless they can build that trust again.

 

[Harry_Wild]

Realistically, the "120%" of PC users are actually waiting for even cheaper AM5 motherboards.

Current situation, $160 is too much for the cheapest AM5 board if you want to build a "cheap but very good PC".So another month of waiting, and only then it will be much easier when you have AM5 motherboards for $80-100$.

Maybe true but then one who have to wait still again for it to back in stock and plentiful to buy! Maybe another 6 months of waiting it out? I just buy what is available that a brand new chipset at any price for my 7600 that can do 24 lanes and DDR5 RAM at 6600! Remember, am planning to keep the motherboard to 2025+, that alone is worth to me, $300 for a decent motherboard!

 

[nicalandia]

CapframeX being a Biased Intel fan yet again..

 

[Carfax83]

I'm curious about the performance of Zen 4 3D in games that use ray tracing. Raptor Lake has a huge performance lead over Zen 4 in heavy RT games like Spider-Man Miles Morales that use the CPU for BVH building and maintenance. Tests have it at nearly 50%.

The 3D cache should make a nice dent in that gap, but I doubt it will be enough to fully close it.

 

[Kaluan]

CapframeX being a Biased Intel fan yet again..
View attachment 74019
View attachment 74021

Ah, where's that (Intel branded) copium huffing device meme when you need it...


AMD Zen4X3D is "only this" and "only that"... my boy knows 13900K is "only" 6-7% faster than 5800X3D at 1080p, right? While also consuming more than twice the wattage in those gaming tasks... But eh, efficiency doesn't matter unless the brand you hate is bad at it right? (pretty sure I'll find him blasting RDNA3 power efficiency in his tweeter feed)

Going by AMD's numbers, which they didn't mislead with back in the 5000X3D drum-up period (and I hope they don't do now), 7800X3D is ~21% faster (8 game average) than 5800X3D and 7950X3D is ~16% faster (6 game average) than 13900K. That fits pretty great into the "5800X3D 6% slower than 13900K, 7000X3D 16% faster than 13900K... so ~21% faster than 5000X3D" triangle.
(If we asume 7800, 7900 and 7950 X3Ds all perform roughly the same in games)