AMD Ready with Zen 4 3DV Cache Chiplet, Expects to Repeat 5800X3D Magic Versus Raptor Lake

[btarunr]

AMD is allegedly ready with a working "Zen 4" chiplet that has stacked 3D Vertical Cache (3DV cache) memory, which supplements the on-die L3 cache, and is found to massively improve gaming performance. "Moore's Law is Dead" reports that the Zen 4 + 3DV Cache chiplet will be used with various Ryzen 7000X3D SKUs, as well as special EPYC "Genoa" SKUs.

The 3DV Cache deployed with the "Zen 4" chiplet is a second-generation to the one on the "Zen 3 + 3DV cache" chiplet, and AMD has worked on a number of bandwidth and latency improvements, so it performs in-sync with the generationally-faster on-die L3 cache of the "Zen 4" chiplet. Unlike the CCD below it that's built on TSMC N5 (5 nm EUV), the L3D (the stacked die with the 3DV cache) is possibly be built on an older node, such as N6 (6 nm), since it only contains a slab of memory and doesn't warrant N5. "Moore's Law is Dead" reports that AMD expects to repeat the magic of the 5800X3D when it comes to gaming performance, and expects Ryzen 7000X3D processors to dominate Intel's 13th Gen "Raptor Lake" processors. This was echoed by another reliable source, greymon55.



View at TechPowerUp Main Site | Source

 

[Crackong]

I think 99% of us expected an 7800X3D when we saw there is no 7800x on the leaked info..

 

[maxfly]

Right, pretty anti-climactic. Altho this may not help early sales to gamers.

 

[Dirt Chip]

Very good news if verified.
Will (keep on) squeeze Intel.
The KS is sure already cooking as well.

3DV is peurly for gaming, KS is way way to much power and heat so none of them is in my to-buy list but anyway, high end stuff is sexy nonetheless.

 

[ModEl4]

I think 99% of us expected an 7800X3D when we saw there is no 7800x on the leaked info..

I guess it was the easy option.
8core 7700X will have the smaller increase vs Zen3 in relation with all the other Zen4 models due to TDP differences.
They can increase TDP if they there is a need for 7800X3D.
By naming the 8core Zen4 7700X, now they can price 7800X3D higher and get away with it more easily:

7700X->7800X->7800X3D
-----------------------------------------
Another stupid assessment from Moore's law is Dead:

He's spreading the rumor that while Zen3 V cache was 10-15% faster (essentially the known average difference) than Zen3 in the report, Zen4 V cache will be 30% faster than Zen4 (by using essentially the on average difference between Zen3 V cache/Zen3 that we already know (without saying that is the average difference) he is essentially implying that the Zen4 V cache/Zen4 difference is the average also) You may find 1-2 games that it will be 30% faster (like 5800X3D) but on average it won't be more than 14% faster vs 7900X.
Another meaningless prediction that will fail miserably!

 

[AM4isGOD]

Well this is sure to get some chops slavering.

 

[openbox1980]

I

I guess it was the easy option.
8core 7700X will have the smaller increase vs Zen3 in relation with all the other Zen4 models due to TDP differences.
They can increase TDP if they there is a need for 7800X3D.
By naming the 8core Zen4 7700X, now they can price 7800X3D higher and get away with it more easily:

7700X->7800X->7800X3D
-----------------------------------------
Another stupid assessment from Moore's law is Dead:

He's spreading the rumor that while Zen3 V cache was 10-15% faster (essentially the known average difference) than Zen3 in the report, Zen4 V cache will be 30% faster than Zen4 (by using essentially the on average difference between Zen3 V cache/Zen3 that we already know (without saying that is the average difference) he is essentially implying that the Zen4 V cache/Zen4 difference is the average also) You may find 1-2 games that it will be 30% faster (like 5800X3D) but on average it won't be more than 14% faster vs 7900X.
Another meaningless prediction that will fail miserably!

I dont think they will release a 7800x.

 

[ratirt]

He's spreading the rumor that while Zen3 V cache was 10-15% faster (essentially the known average difference) than Zen3 in the report, Zen4 V cache will be 30% faster than Zen4 (by using essentially the on average difference between Zen3 V cache/Zen3 that we already know (without saying that is the average difference) he is essentially implying that the Zen4 V cache/Zen4 difference is the average also) You may find 1-2 games that it will be 30% faster (like 5800X3D) but on average it won't be more than 14% faster vs 7900X.

Maybe he is accounting for the fact that the IF will run at 3000Mhz (almost double) and this additional cache (not sure about the capacity) maybe will give higher boosts in games than zZen3.
Saying 'wont be more than 14%' may also not be correct. Zen4 is different than Zen3 in so many aspects and yet you base your prediction on Zen3's performance boost with 3d-vcache.

 

[ModEl4]

I

I dont think they will release a 7800x.

I agree, they probably won't.
This doesn't change the fact that if AMD wanted, they could release a higher TDP 8core variant (7800X) so the possibility despite small is there and being there along with the $450 5800X earliest history, it leaves the impression that a big difference between 8core and 8core V cache is easier not to attract so negativity by naming it 7700X imo.

Maybe he is accounting for the fact that the IF will run at 3000Mhz (almost double) and this additional cache (not sure about the capacity) maybe will give higher boosts in games than zZen3.
Saying 'wont be more than 14%' may also not be correct. Zen4 is different than Zen3 in so many aspects and yet you base your prediction on Zen3's performance boost with 3d-vcache.

I doubt it will be 2X in speed and the capacity will be the same as 5800's V-cache.
The upper limit that i used (14%) vs 7900X is already higher than the 5800X3D vs 5900X difference (10.6% in 720p in TPU results) and on top of that 7900X/7700X difference will be greater than what 5900X/5800X 720p difference was (1.1%) due to frequency and TDP difference.
So i also assumed an increase already, I just don't see how it will reach 30% on average.
In any case, time will tell who's right and who's wrong.

 

[Mussels]

Let's hope the new IHS helps with the heat, and that they unlock more PBO settings

I think 99% of us expected an 7800X3D when we saw there is no 7800x on the leaked info..

Yeah, 7700x for the masses, 7800x3D for the people with money to burn

Eventually (even if just OEM) i'm sure we'll see a 7800x, as well as 7700 and 7800 non-X chips.

 

[ratirt]

I doubt it will be 2X in speed and the capacity will be the same as 5800's V-cache.
The upper limit that i used (14%) vs 7900X is already higher than the 5800X3D vs 5900X difference (10.6% in 720p in TPU results) and on top of that 7900X/7700X difference will be greater than what 5900X/5800X 720p difference was (1.1%) due to frequency and TDP difference.
So i also assumed an increase already, I just don't see how it will reach 30% on average.
In any case, time will tell who's right and who's wrong.

It will be 2x since the DDR5 for Zen4 will run at 6000Mhz so 3ghz is almost 2x what Zen3 had (1600Mhz for 3200Mhz ram)
The 5800X3d when I looked over Hardware Unboxed at 1080p with 41 games benchmarks was 15% faster from a 5800x.
I'm waiting for reviews and then it will be confirmed but 30% uplift in gaming performance zen4 over zen3 (i think that is what we are talking about here) is plausible. Zen4 does bring a lot new stuff in the equation. I think the result 5800x to 7700x will be more than 15% uplift. Will it be 30%? Like I said. I think it is very plausible especially if you consider 7800x3d.

 

[Assimilator]

Disappointing that V-Cache is not part of the entire lineup and still considered a premium feature that AMD will charge an arm and leg for.

 

[ModEl4]

It will be 2x since the DDR5 for Zen4 will run at 6000Mhz so 3ghz is almost 2x what Zen3 had (1600Mhz for 3200Mhz ram)
The 5800X3d when I looked over Hardware Unboxed at 1080p with 41 games benchmarks was 15% faster from a 5800x.
I'm waiting for reviews and then it will be confirmed but 30% uplift in gaming performance zen4 over zen3 (i think that is what we are talking about here) is plausible. Zen4 does bring a lot new stuff in the equation. I think the result 5800x to 7700x will be more than 15% uplift. Will it be 30%? Like I said. I think it is very plausible especially if you consider 7800x3d.

6000MHz isn't the sweet spot like many site reported, is the preferred memory frequency if your sample can hit IF at 3GHz, essentially the max IF frequency (Like 2000MHz was for Zen3 (e.g. TPU testbed) but possibly slightly more frequent regarding achievability) 3.2GHz will be extremely rarely achievable but maybe technically possible, that's my understanding.
Check TPU's sources:
Wccftech
1usmus
«So DDR5-6000 for AMD Ryzen 7000 "Zen 4" CPUs already sounds great for AM5 and that will be DDR5-5600 by default. Higher frequency DIMMs are supported but as soon as you go above the DDR5-6000 limit, you will drop down to a 1:2 IFC. We are told that DDR5-6400 running at 1:2 will produce poor results and is not recommended if you are looking for better gaming performance»
So it's 1.5X not 2X regarding IF limits!
And the 30% uplift MLID assumption in gaming performance is zen4 V-cache over standard zen4 not zen4 V-cache over Zen3 like you say (did you even read the MLID pic that i had in my original post?) If the assumption was on average 30% from standard Zen3 it would be OK as a prediction but that was not what MLID said...

 

[ratirt]

Disappointing that V-Cache is not part of the entire lineup and still considered a premium feature that AMD will charge an arm and leg for.

Not sure about arm and leg especially if you look over to the 5800x3d variant, its cost was exactly what the 5800x was at launch.
Not sure what the price for this one will be but would it make sense to release 7950x (high core CPU) with 3d Vcache if that one is meant for crunching data and as we know Vcache does not give much if anything in that department? Also, clocks might be lower than a 7950x.
The other thing is, these CPUs would take capacity which I don't think would have been a great idea. 7800x3d for games is perfect solution 8c16t (safe bet) and vcache to speed things up. Having all of the lineup with 3dvcache.
I'm surprised you actually say that. First you are disappointing entire line-up not having it and then you AMD charges extra for a premium feature? I dont think it is a premium feature. this cpu has a different focus which is games.

6000MHz isn't the sweet spot like many site reported, is the preferred memory frequency if your sample can hit IF at 3GHz, essentially the max IF frequency (Like 2000MHz was for Zen3 (e.g. TPU testbed) but possibly slightly more frequent regarding achievability) 3.2GHz will be extremely rarely achievable but technically possible, that's my understanding.
Check TPU's sources:
Wccftech
1usmus
«So DDR5-6000 for AMD Ryzen 7000 "Zen 4" CPUs already sounds great for AM5 and that will be DDR5-5600 by default. Higher frequency DIMMs are supported but as soon as you go above the DDR5-6000 limit, you will drop down to a 1:2 IFC. We are told that DDR5-6400 running at 1:2 will produce poor results and is not recommended if you are looking for better gaming performance»
So it's 1.5X not 2X regarding IF limits!
And the 30% uplift MLID assumption in gaming performance is zen4 V-cache over standard zen4 not zen4 V-cache over Zen3 like you say (did you even read the MLID pic that i had in my original post?) If the assumption was on average 30% from standard Zen3 it would be OK as a prediction but that was not what MLID said...

Sweet spot for Zen3 was not 2000Mhz considering the graps AMD used and 'sweet spot' in my understanding was 3800Mhz for zen3
I'm sure, if the sweet spot is 6000Mhz for AM5 IF and DDR5, it will be possible to go higher than that.
Also, wasn't the sweet spot aka 3000 mhz IF or 6000Mhz perceived as the AMD official supported speed for the AM5 and Zen4? Just like the 3200Mhz was for Zen3?

 

[Aquinus]

is found to massively improve gaming performance.

I still find it weird how Phoronix's review was quite literally the exact opposite. There were minimal gaming improvements in Linux, but some non-gaming workloads were accelerated significantly with some workloads like zstd compression outpacing even the 5950x. I guess the CPU scheduler on Linux vs Windows could be to blame.

 

[btk2k2]

I guess it was the easy option.
8core 7700X will have the smaller increase vs Zen3 in relation with all the other Zen4 models due to TDP differences.
They can increase TDP if they there is a need for 7800X3D.
By naming the 8core Zen4 7700X, now they can price 7800X3D higher and get away with it more easily:

7700X->7800X->7800X3D
-----------------------------------------
Another stupid assessment from Moore's law is Dead:

He's spreading the rumor that while Zen3 V cache was 10-15% faster (essentially the known average difference) than Zen3 in the report, Zen4 V cache will be 30% faster than Zen4 (by using essentially the on average difference between Zen3 V cache/Zen3 that we already know (without saying that is the average difference) he is essentially implying that the Zen4 V cache/Zen4 difference is the average also) You may find 1-2 games that it will be 30% faster (like 5800X3D) but on average it won't be more than 14% faster vs 7900X.
Another meaningless prediction that will fail miserably!

The 5800X3D clocks closer to a 5700X than a 5800X and despite the regression was still 15% faster on average in a standard gaming suite. If you check a suite with stuff like Stellaris, ACC, MSFS etc the gains are much greater.

For Zen 4 if AMD can clock it higher relative to other parts than the 5800X3D is they can gain a few more % boost just from not regressing in that regard. Also if AMD have tuned it a bit better and managed to further improve latency and or bandwidth from the stacked L3 that will also improve performance. Still even with all of that combined I see it in maybe the 20% above standard Zen 4 region rather than 30%.

7950x (high core CPU) with 3d Vcache if that one is meant for crunching data and as we know Vcache does not give much if anything in that department? Also, clocks might be lower than a 7950x.

A V-Cache 7950X if it can achieve similar clocks to the vanilla version is going to be the ultimate do anything desktop CPU. One thing the 12900K (and probably 13900K) has going for it is that it is a monster in gaming and can keep up with a 5950X in MT workloads give or take for when you need it so it can service users who want / need top tier performance in both arenas.

So I can easily see a stack such as
7950X3D - Top tier performance everywhere
7950X - Top tier MT performance - Good Gaming performance
7900X - Good MT performance - Good Gaming performance
7800X3D - Top tier gaming performance - okay MT performance
7700X - Good gaming performance - okay MT performance
7600X - Good gaming performance - meh MT performance

That kind of stack gives almost everybody who can afford the platform costs an option based on wants, needs and budget.

 

[HenrySomeone]

"Moore's Law is Dead" reports that AMD expects to repeat the magic of the 5800X3D when it comes to gaming performance, and expects Ryzen 7000X3D processors to dominate Intel's 13th Gen "Raptor Lake" processors.

So, if we translate the ridiculously unapologetic AMD fanboy MLiD into potential actual truths, it means that at best 7000x3D will be roughly on par with Raptor Lake, perhaps taking the lead in a couple titles (just like 5800x3d vs Alder Lake), but will take several more months to get to the market and will probably start at close to $600, meanwhile you'll be able to get pretty much the same gaming performance for under $300 with 13600k (and even less, taking bclk OCing into account). What a deal from team red!!!

 

[Space Lynx]

I think I am going to do my next build with a 7600X, then upgrade it to a 7800X3D whenever that comes out. This is all dependent if I can get my hands on a RDNA3 top tier gpu though, which I doubt I will, graphics cards are so hard to get the first few months they come out. Especially with scalper bots these days, fucking atrocius nightmare its going to be for a solid year, unless you just get really lucky.

Fuck the scalper bot third party seller bitches.

@the54thvoid If it's too much cursing here, I completely understand if you need to delete it, but such topics, imo require lots of curses.

 

[ModEl4]

Sweet spot for Zen3 was not 2000Mhz considering the graps AMD used and 'sweet spot' in my understanding was 3800Mhz for zen3

Correct regarding Zen3 memory sweet spot (i never said 2000MHz was the sweet-spot but the IF cache limit)

I'm sure, if the sweet spot is 6000Mhz for AM5 IF and DDR5, it will be possible to go higher than that.

3000MHz IF cache according to TPU sources (that i quoted) is not the sweet-spot but the MAX IF frequency

Also, wasn't the sweet spot aka 3000 mhz IF or 6000Mhz perceived as the AMD official supported speed for the AM5 and Zen4? Just like the 3200Mhz was for Zen3?

My understanding is that the official IF frequency will probably be 2800MHz (there is a small possibility for 2400MHz also, but since 5600MHz memory is officially supported 2800MHz makes more sense, on the other hand 1600MHz is -20% from the 2GHz IF cache limit regarding Zen3, if the ratio regarding standard IF frequency vs IF limit is the same on Zen4, with 3000MHz IF cache limit the standard should be 2400MHz following this logic)

The 5800X3D clocks closer to a 5700X than a 5800X and despite the regression was still 15% faster on average in a standard gaming suite. If you check a suite with stuff like Stellaris, ACC, MSFS etc the gains are much greater.

According to TPU testbed +12% vs 5800X. I think TPU is indicative enough (sure there is stuff like the one you mention that the gains are higher)
And we are talking about 720p, in 1080p or higher the difference goes smaller and smaller!

For Zen 4 if AMD can clock it higher relative to other parts than the 5800X3D is they can gain a few more % boost just from not regressing in that regard. Also if AMD have tuned it a bit better and managed to further improve latency and or bandwidth from the stacked L3 that will also improve performance. Still even with all of that combined I see it in maybe the 20% above standard Zen 4 region rather than 30%.

My prediction takes account IF and clock frequency increase (i said that there is possibly if AMD wants to increase the TDP for 7800X3D implying different frequency ratio possibility for 7800X3D/7700X vs 5800X3D/5800X.
5900X was 1.1% on average faster than 5800X in 720p, 7900X is going to be at least 2.5%, probably 3.5% faster than 7700X, so my 14% upper limit vs 7900X translates up to 18% vs 7700X

 

[LemmingOverlord]

Please stop parroting "Moore's Law is Dead". It's a fucking joke.

 

[Assimilator]

"Moore's Law is Dead" reports that AMD expects to repeat the magic of the 5800X3D when it comes to gaming performance, and expects Ryzen 7000X3D processors to dominate Intel's 13th Gen "Raptor Lake" processors.

So, if we translate the ridiculously unapologetic AMD fanboy MLiD into potential actual truths, it means that at best 7000x3D will be roughly on par with Raptor Lake, perhaps taking the lead in a couple titles (just like 5800x3d vs Alder Lake), but will take several more months to get to the market and will probably start at close to $600, meanwhile you'll be able to get pretty much the same gaming performance for under $300 with 13600k (and even less, taking bclk OCing into account). What a deal from team red!!!

Pot, kettle, black.

 

[btk2k2]

Correct regarding Zen3 memory sweet spot (i never said 2000MHz was the sweet-spot but the IF cache limit)

3000MHz IF cache according to TPU sources (that i quoted) is not the sweet-spot but the MAX IF frequency

My understanding is that the official IF frequency will probably be 2800MHz (there is a small possibility for 2400MHz also, but since 5600MHz memory is officially supported 2800MHz makes more sense, on the other hand 1600MHz is -20% from the 2GHz IF cache limit regarding Zen3, if the ratio regarding standard IF frequency vs IF limit is the same on Zen4, with 3000MHz IF cache limit the standard should be 2400MHz following this logic)


According to TPU testbed +12% vs 5800X. I think TPU is indicative enough (sure there is stuff like the one you mention that the gains are higher)
And we are talking about 720p, in 1080p or higher the difference goes smaller and smaller!



My prediction takes account IF and clock frequency increase (i said that there is possibly if AMD wants to increase the TDP for 7800X3D implying different frequency ratio possibility for 7800X3D/7700X vs 5800X3D/5800X.
5900X was 1.1% on average faster than 5800X in 720p, 7900X is going to be at least 2.5%, probably 3.5% faster than 7700X, so my 14% upper limit vs 7900X translates up to 18% vs 7700X

Techspot / HUB had a 15% advantage for the 5800X3D in a 40 game suite at 1080p (not quite max settings either) when tested with a 3090Ti.

Computerbase across their 2 tests also saw a 17% increase at 720p. In fact when you look at the results TPU are actually a bit of an outlier with one of the smallest perf increases over the 5800X vs other outlets so relying on a single number, and a low ball one at that, will give you a lower bound. If you take the 20% average wins you get an upper bound.

Given that I expect Zen 4 3D to be around 18%-25% faster in games than vanilla Zen 4. This sort of ball park seems perfectly achievable because we know AMD left performance on the table regarding clock speeds and with some work they can lower the latency cost of triple the cache or maybe even remove the latency penalty all together and those two things combined will help achieve a greater performance uplift for the Zen 4 3D parts than the 5800X3D managed over the 5800X.

 

[ThrashZone]

Hi,
AM5 is just getting started and lots more series chips to follow still supported on the same original boards
Intel will be two chip series and board is ewaste.

 

[ModEl4]

Techspot / HUB had a 15% advantage for the 5800X3D in a 40 game suite at 1080p (not quite max settings either) when tested with a 3090Ti.

Computerbase across their 2 tests also saw a 17% increase at 720p. In fact when you look at the results TPU are actually a bit of an outlier with one of the smallest perf increases over the 5800X vs other outlets so relying on a single number, and a low ball one at that, will give you a lower bound. If you take the 20% average wins you get an upper bound.

Given that I expect Zen 4 3D to be around 18%-25% faster in games than vanilla Zen 4. This sort of ball park seems perfectly achievable because we know AMD left performance on the table regarding clock speeds and with some work they can lower the latency cost of triple the cache or maybe even remove the latency penalty all together and those two things combined will help achieve a greater performance uplift for the Zen 4 3D parts than the 5800X3D managed over the 5800X.

Yes, sure, if you think that other publications has more indicative difference than TPU, 20% seems just fine.

 

[stimpy88]

So AMD cache starved Zen 4 deliberately for marketing... These kind of shenanigans will get you beaten AMD! You need to come back to fighting with much higher IPC!