Info 64MB V-Cache on 5XXX Zen3 Average +15% in Games

[Kedas]

Well we know now how they will bridge the long wait to Zen4 on AM5 Q4 2022.
Production start for V-cache is end this year so too early for Zen4 so this is certainly coming to AM4.
+15% Lisa said is "like an entire architectural generation"

 

[Joe NYC]

it doesn't contradict what was said by Lisa, only peoples interpretation.

It contradicts interpretation of the most simple minded thinkers, who can't see the big picture.

There were pages and pages of replies berating me for suggesting that AMD and TSMC will make it happen, will get Zen 3D into the market place by the time it needs to be in market place - by Q4 2021

 

[jpiniero]

It contradicts interpretation of the most simple minded thinkers, who can't see the big picture.

There were pages and pages of replies berating me for suggesting that AMD and TSMC will make it happen, will get Zen 3D into the market place by the time it needs to be in market place - by Q4 2021

Christmas is technically Q4 2021 but not really. The Q4 people were saying roughly when Alder Lake "launches".

 

[Joe NYC]

Christmas is technically Q4 2021 but not really. The Q4 people were saying roughly when Alder Lake "launches".

ADL launches in Q4. Launching within a month of ADL would be great timing for Zen3D.

 

[jpiniero]

ADL launches in Q4. Launching within a month of ADL would be great timing for Zen3D.

Christmas would be two months.

 

[Joe NYC]

In fairness to @lobz , it certainly looks like AMD has bumped production of Zen3d forward unexpectedly. I wonder what other optimizations have been made to the B2 stepping?

Since it is new silicon, I think it will incorporate whatever feedback AMD got from testing Zen 3D chips, possibly some additional optimizations for that. Because, it is not likely that AMD hit perfection on a new unknown technology so far back, when original Zen 3 was released.

As far as any wishlist for B2 stepping, by far #1 would be N6 node, but that could add to validation cycle...

Other than N6, a small bump in clock speed would not hurt.

BTW, since the first B2 stepping have just been spotted in the wild, here is another beauty from @lobz seemingly agreeing with @Vattila, apparently without reading @Vattila 's post, because he was contradicting him:

That's correct. I just wanted to point out, how unfeasible it is, expecting AMD to be producing those chips right now.

So if B2 is trickling out now, it must have been in production when @lobz assured it could not have been in production.

 

[Joe NYC]

Christmas would be two months.

haha, do you really want to go there? Arguing definition of Christmas?

 

[jpiniero]

haha, do you really want to go there? Arguing definition of Christmas?

Alder Lake's "launch" is October 27th.

 

[Joe NYC]

TBH the 'X' version of Milan started to be mentioned at many places. Even a SKU list is out. This should mean a 2021 launch at least for the server parts.

On Milan X, I am only guessing, but I would go the other way on that, and say that it is more likely that the official launch will be early 2022.

 

[Joe NYC]

I don’t think you realize the full implications of ADL-S. The 12900k is a $500-$600 chip, not an $800 chip. It’s competition is the 5900X, not the 5950X.

Also, those settings are bone stock with JEDEC timings and all. Compare that to a bone stock 5950x and Intel is faster.

Anandtech scored 1655/15769 in their testing. They run stock with JEDEC timings.

EDIT: If these scores hold up, that means ADL-S is around 10% faster in GB5 than the 5950x, despite being hundreds of dollars cheaper. Zen3D, of course, will reverse things again.

On ADL pricing, I don't think it is 100% assured that 12900k price will be equal to 11900k price.

Intel was already lapped by NVidia, and could possibly be lapped by AMD in not too distant future in Market Cap. Intel hasn't delivered their long suffering shareholders anything. for years.

I don't think Intel is in position of leaving money on the table, if ADL is really worth the same as, for example, 5950x

 

[Joe NYC]

Well, that is preety cheep if you take into account the colossal amount of Cache.

- if AMD integrates 64mb 3D-V-Cache per CCD

- 64 core Epyc has 8 CCD-s=512mb 3D-V-cache

And around 768mb including internal L3.

But, that may not be the full extend of it. There may be more. Some people, in the know, think we are going to be entering GigaByte era.

 

[Joe NYC]

Alder Lake's "launch" is October 27th.

And as far as Christmas Shopping, the Christmas season starts on either Black Friday, November 26 or Cyber Monday, November 29.

 

[jpiniero]

And as far as Christmas Shopping, the Christmas season starts on either Black Friday, November 26 or Cyber Monday, November 29.

Twitter user didn't say Christmas Shopping, said Christmas. December 25th.

 

[Joe NYC]

Twitter user didn't say Christmas Shopping, said Christmas. December 25th.

[facepalm]

 

[jpiniero]

[facepalm]

I will say I am assuming the user is saying just after Christmas, like CES as I mentioned earlier. Not on Christmas itself. If it was Black Friday or whatever the user would have said that.

 

[Thibsie]

It contradicts interpretation of the most simple minded thinkers, who can't see the big picture.

There were pages and pages of replies berating me for suggesting that AMD and TSMC will make it happen, will get Zen 3D into the market place by the time it needs to be in market place - by Q4 2021

That's not what you suggested.
And since it seems you can't even discuss thing appropriately, as far as I'm concerned, you'll be happily ignored from now on.

 

[Joe NYC]

That's not what you suggested.
And since it seems you can't even discuss thing appropriately, as far as I'm concerned, you'll be happily ignored from now on.

Well, yes I did, I pointed out that AMD and TSMC would prioritize it, because it is the highest priority product:

It's called analysis.

That both TSMC and AMD have superb management team, and that these management teams recognize market potential of this breakthrough technology and will prioritize it.

I specified why this is a priority for TSMC as much as AMD:

Suppose you have a technology on the drawing board that can:
- double or more demand for your product
- leaps a generation ahead in technology that has been mired in expensive trench warfare
- sends your most dangerous competitor almost a generation back in the market place

Would you just wait for this to happen when it happens, or push it to top of the priorities, expand resources to expedite it from lab to full production?

And that AMD should launch shortly after ADL:

Regardless, AMD should launch Zen 3D shortly after Alder Lake launch, whether they have 100, 100,000 or 1,000,000 stockpiled.

Ignore at your own risk 🙂

 

[Hougy]

The Zen 2 in consoles and APU has 1/4 of the L3 cache. It doesn't seem to make a big difference. Why would doubling the Zen 3 L3 cache give +15% gaming performance?

 

[DrMrLordX]

it doesn't contradict what was said by Lisa, only peoples interpretation.

ummm

They wouldn't launch at Christmas. That would be silly. They'd wait until CES.

. . . that was my impression of what people expected initially? Basically, "by Christmas" would be at least a month before CES. What Dr. Su said seemed to line up with your (jpiniero) interpretation: by CES.

It contradicts interpretation of the most simple minded thinkers, who can't see the big picture.

Careful with the ad hominem. And read above.

 

[Vattila]

Why would doubling the Zen 3 L3 cache give +15% gaming performance?

See post #423.

By the way, one layer of V-Cache is tripling the amount of L3 cache total in each CCX (and more layers are likely possible and planned).

Hardware Unboxed has now followed up with more cores-vs-cache tests, including 4-core in particular:

PS. Note that the tested SKUs, except for the 4-core, are clocked at 4.5 GHz to isolate the effect of cache and cores. From the written article:

"To put this test together we used the Gigabyte Z590 Aorus Xtreme motherboard, clocking the three Intel K-SKU CPUs at 4.5 GHz with a 45x multiplier for the ring bus and used DDR4-3200 CL14 dual-rank, dual-channel memory with all primary, secondary and tertiary timings manually configured. The Core i3-10105F used the same spec memory. [...] Because the Core i3 models are locked, we're unable to use a 4.5 GHz clock frequency, instead the [4-core] 10105F ran at 4.2 GHz which is the spec all-core frequency, so it’s going to be running at a 7% lower frequency than the K-SKU parts."

Making a Fast Quad-Core Gaming CPU

How much difference can L3 cache make with just 4 CPU cores active? We've gone back to test a variety of CPUs checking for frequencies, number of...

www.techspot.com

 

[moinmoin]

The Zen 2 in consoles and APU has 1/4 of the L3 cache. It doesn't seem to make a big difference.

One is not like the other.

APUs definitely suffer on the smaller cache if you look for direct comparisons. No gaming benchmark will suggest you to use an APU instead a CPU if you use a dedicated GPU right away. That's because even though performance of APUs is very close to their CPU counterparts, the smaller cache will decrease performance due to increased cache thrashing.

Consoles are not as affected by this due to the more controlled nature of its OS and software. Unlike on PC you don't have multiple independent processes accessing resources in an unplanned manner, allowing developers to partition usage of resources in more efficient ways.

 

[Insert_Nickname]

Consoles are not as affected by this due to the more controlled nature of its OS and software. Unlike on PC you don't have multiple independent processes accessing resources in an unplanned manner, allowing developers to partition usage of resources in more efficient ways.

True. But you can minimise the impact by only running the absolutely necessary when gaming. No other open programs, browsers etc.

It's what we did back in the single core age. There any interrupts mattered the heck of lot more.

 

[lobz]

It contradicts interpretation of the most simple minded thinkers, who can't see the big picture.

There were pages and pages of replies berating me for suggesting that AMD and TSMC will make it happen, will get Zen 3D into the market place by the time it needs to be in market place - by Q4 2021

Dear God... just how high your horse is when you think of yourself as a thinker with a beautifully complicated mind?

 

[igor_kavinski]

Why is the release date of Zen 3D such a contentious issue? I don't see the point of so many posts debating that. I do have to question AMD's motive for announcing it months before it intends to release it. Is it worried that it may lose the performance crown to Intel with the launch of Alder Lake?

 

[Mopetar]

The Zen 2 in consoles and APU has 1/4 of the L3 cache. It doesn't seem to make a big difference. Why would doubling the Zen 3 L3 cache give +15% gaming performance?

The difference in access time for cache and system memory is roughly an order of magnitude greater. In other words it's around ten times quicker if you can retrieve something from cache as opposed to having to go out to memory. Bigger cache just means the likelihood of having something there you need increasing.

Consoles as opposed to PCs are a bit different. Another poster already mentioned the OS is simplified and the amount of other software is minimized. Obviously you can do the same on a PC and improve performance. Anyone who's ever had an anti-virus scanner run during a gaming session can attest to this.

However, the bigger difference is that all consoles of a particular type are exactly the same. You don't have a PS5 with a different CPU than another PS5, which means a game can be tuned to work well within the hardware it has. If you know exactly what resources you're guaranteed to have, the game engine, levels, etc. can be designed around that and you can write code a lot closer to the hardware to control everything more tightly.

Not all games would benefit from adding more cache either since the GPU is typically going to be the bottleneck in console titles, especially later into their lifespans. By the time the next generation consoles that will replace them are ready to ship, the GPU in a PS5 is going to be a low-end or entry level PC gaming card. Even on PC games you really only see a lot of CPU advantages come into play when pairing incredibly powerful cards with low resolutions like 720p or below. At 4K even a cheap $100 CPU can perform as well as a $600 CPU because the GPU is the bottleneck.

 

[Tuna-Fish]

Why is the release date of Zen 3D such a contentious issue? I don't see the point of so many posts debating that. I do have to question AMD's motive for announcing it months before it intends to release it. Is it worried that it may lose the performance crown to Intel with the launch of Alder Lake?

The reason they announced it so early is that credible leaks had started to come out, and they wanted to be in front of them. Because the product involves advanced packaging which involves so many other companies, they could not have kept it under wraps.

As to the ceaseless cache amount debate:

The height of the cache stack gives AMD a tool they can use to trade cost for performance. If Alder Lake releases and would beat a Zen3 with a single layer of cache, but not more layers, AMD absolutely will release a version with however many layers it takes to take that crown. However, if Zen3 with a single layer beats the highest-clocked Alder Lakes, I can see AMD preferring to save on cost and just ship one-layer versions.

The positioning of TR is irrelevant, because the selling point of low-core-count TR systems is IO, not performance.

Other than that, trying to divine more meaning to statements from AMD is pointless. They have no reason to give detail this early, so just because they are not directly talking about something yet, doesn't mean they are not doing it.