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"

 

[naukkis]

It'd be two 75-ish mm2 bonded plus the IOD versus 189 mm2 of 7 nm. The desktop Cezanne margins have to be non ideal too, they must only do it because the chips are so leaky that they would be so unusable in mobile. Epyc is likely binned a lot tighter in comparison.

Cache chiplet is 36mm2.

 

[Timorous]

It'd be two 75-ish mm2 bonded plus the IOD versus 189 mm2 of 7 nm. The desktop Cezanne margins have to be non ideal too, they must only do it because the chips are so leaky that they would be so unusable in mobile. Epyc is likely binned a lot tighter in comparison.

80mm^2 for zen 3 ccd + 36mm^2 for cache chiplet is 116mm^2 of N7. That means for every 3 cezanne parts AMD can make 5 stacked CCDs.

I also expect AMD are making more margin on DIY cezanne than OEM and the rumours are v-cache *edit: Ryzen, obviously Milan-X is OEM* will be DIY only.

 

[jpiniero]

Cache chiplet is 36mm2.

Oh you're right. Must have missed that part, thought it was bigger. Does make the cost increase compared to no vcache quite a bit less. I'd still expect AMD to charge at least $50 more if not $100 for any vcache models compared to Vermeer's MSRP.

The 12900K is (sort of, it's not really available) going for $650. Maybe that's the MSRP for the 12 core with vcache? That'd be $100 more than the 5900X's MSRP.

 

[leoneazzurro]

Oh you're right. Must have missed that part, thought it was bigger. Does make the cost increase compared to no vcache quite a bit less. I'd still expect AMD to charge at least $50 more if not $100 for any vcache models compared to Vermeer's MSRP.

The 12900K is (sort of, it's not really available) going for $650. Maybe that's the MSRP for the 12 core with vcache? That'd be $100 more than the 5900X's MSRP.

I don't think so, BOM is only a part of the cost of a CPU. Yes, it will cost more to AMD but other costs, i.e. testing, binning are practically the same and packaging is moderately higher. Other than this, these CPU must be positioned competitively respect to ADL, and pricing them 100$ higher will simply damage sales.

 

[moinmoin]

Zen3D can be sold as a loss leader just to maintain (or grow) desktop market share.

Aside the fact that currently AMD is losing desktop market share while focusing on and increasing laptop and server market share, why would AMD want to sell a loss leader while everything they make is getting sold anyway?

 

[Kedas]

Assuming what AMD claimed is correct.
Since 12900k is less than 15% faster on average in games this would mean that AMD will take the lead again in a few months or at least the same for less heat/noise.
This is with CPU limited systems that most people do not have though.

I assume the new V-cache versions will take the current prices and the non V-cache prices will drop in price.

 

[leoneazzurro]

Aside the fact that currently AMD is losing desktop market share while focusing on and increasing laptop and server market share, why would AMD want to sell a loss leader while everything they make is getting sold anyway?

Except that this not being true?

AMD took almost a quarter of the x86 CPU market in Q3, its largest share since 2006

The Q3 2021 x86 CPU Market Share report from analyst firm Mercury Research (via HardwareTimes) will be welcome reading for AMD. The biggest reveal is that the...

www.techspot.com

probably now that ADL is out, AMD will lose desktop market share, but until now it did not happen, quite the opposite.

 

[epsilon84]

Assuming what AMD claimed is correct.
Since 12900k is less than 15% faster on average in games this would mean that AMD will take the lead again in a few months or at least the same for less heat/noise.
This is with CPU limited systems that most people do not have though.

I assume the new V-cache versions will take the current prices and the non V-cache prices will drop in price.

If the 15% figure holds true, that would put AMD back on top by roughly 5% in CPU bound gaming scenarios. I hope AMD prices these competitively, if they pull another 5600X/5800X move again then we are basically back to square one, just this time it would be slightly faster Zen 3D vs ADL chips rather than slightly faster Zen 3 vs RKL chips.

 

[Thibsie]

If the 15% figure holds true, that would put AMD back on top by roughly 5% in CPU bound gaming scenarios. I hope AMD prices these competitively, if they pull another 5600X/5800X move again then we are basically back to square one, just this time it would be slightly faster Zen 3D vs ADL chips rather than slightly faster Zen 3 vs RKL chips.

Without the need to change platform for AMD.
Would make adl quite expensive for what it is.
We shall see.

 

[moinmoin]

Except that this not being true?

According to Mercury Research that's the current state indeed:

https://twitter.com/x/status/1425999100227510274

 

[leoneazzurro]

According to Mercury Research that's the current state indeed:

https://twitter.com/x/status/1425999100227510274

That's Q2, and it may be conditioned by supply and contracts, I was referring to Q3.

 

[DrMrLordX]

Aside the fact that currently AMD is losing desktop market share while focusing on and increasing laptop and server market share, why would AMD want to sell a loss leader while everything they make is getting sold anyway?

Because if they don't, they may not have Zen4 on desktop until Q42022. That would cause a huge drop in market share if they just sat on the same product for two years.

It's like this:

Sell low margin Zen3D to maintain market share, sell all N5 wafers as Genoa/Bergamo, $$$

Rush Raphael to market to expand desktop market share, lose out on some Genoa sales, $$

Going low-margin on a boutique DiY N7 product may make them more money in the long term. Yeah they can sell those N7 wafers as Milan et al. instead for higher short-term profits, but still. Personally I just want Raphael already, bring it on! Their ideas may differ from mine.

 

[jpiniero]

That's Q2, and it may be conditioned by supply and contracts, I was referring to Q3.

Still should be the same. Intel said desktop volume was up 16% compared to Q3 2020. That's more because of a return of the Corporate market buying desktops again which is a market AMD isn't in right now.

if they just sat on the same product for two years.

It's not the same product.

 

[leoneazzurro]

Still should be the same. Intel said desktop volume was up 16% compared to Q3 2020. That's more because of a return of the Corporate market buying desktops again which is a market AMD isn't in right now.

It's not the same product.

Volume <> market share. Even in Q2, when AMD posted a decrease on the desktop market share, sales went up quite a bit. Btw, if total market share is almost 25% in Q3, and notebook is 22%, and server 20%, desktop must be >25% for the number to sum up.

 

[jpiniero]

Volume <> market share. Even in Q2, when AMD posted a decrease on the desktop market share, sales went up quite a bit. Btw, if total market share is almost 25% in Q3, and notebook is 22%, and server 20%, desktop must be >25% for the number to sum up.

AMD's Q3 financial statement says that CCG unit shipments were down 23%. ASPs OTOH were up an amazing 83%. If anything that shows that AMD is prioritizing pricer products and is not in the mood to go cheap or do any sort of loss leader stuff.

CCG does include all of Radeon so keep that in mind.

 

[leoneazzurro]

Then something does not add up. How AMD could have >24% X86 market share in Q3 when notebook is 22% and server less than 20%? The third part of the equation is desktop.

 

[eek2121]

Oh you're right. Must have missed that part, thought it was bigger. Does make the cost increase compared to no vcache quite a bit less. I'd still expect AMD to charge at least $50 more if not $100 for any vcache models compared to Vermeer's MSRP.

The 12900K is (sort of, it's not really available) going for $650. Maybe that's the MSRP for the 12 core with vcache? That'd be $100 more than the 5900X's MSRP.

Minor gripe here: People need to stop getting their prices from Newegg. If you paid $650 for the 12900k you paid (at least) $30 too much.

Microcenter gave a $20 discount for the motherboard/CPU combio. Best Buy had them for $620. Some retailers had them for $599.

Newegg is not ghe amazing site it once was. They charge too much, the shipping is sliow, and they have poor customer service.

 

[eek2121]

Then something does not add up. How AMD could have >24% X86 market share in Q3 when notebook is 22% and server 20%? The third part of the equation is desktop.

Sounds like you are comparing percentages of segments to percentage of the whole market and getting confused.

 

[leoneazzurro]

Sounds like you are comparing percentages of segments to percentage of the whole market and getting confused.

Well, whole market is made up by the sum of all sub-markets. So if notebooks ship A number of units, servers ships B number of units and desktop is C number if units, then the whole shipment for X86 market is A+B+C and total market share is the weighted average of this, weights being A, B, and C. Now, if you use basic mathematics, there is no possibility that 22%*A+20%*B+X%*C=24,6% with X% being less than 24,6% (which is higher than previous quarter and previous year share). Consoles shipment are not counted as "X86 market" IIRC.

 

[DrMrLordX]

It's not the same product.

I'm talking about Vermeer. If they can Zen3D and delay Raphael to Q4 2022, then they'll be on Vermeer B0 + B2 stepping for two years.

Then something does not add up. How AMD could have >24% X86 market share in Q3 when notebook is 22% and server less than 20%? The third part of the equation is desktop.

Some analysts include console APU sales as part of the x86 market. There are also x86 market segments that AMD isn't in at all where Intel has 100% share, like network appliances.

 

[jpiniero]

I'm talking about Vermeer. If they can Zen3D and delay Raphael to Q4 2022, then they'll be on Vermeer B0 + B2 stepping for two years.[

I doubt they can Zen 3D. OTOH I assume there will be Vermeer rebrands to fill out the stack.

 

[moinmoin]

That's Q2, and it may be conditioned by supply and contracts, I was referring to Q3.

Still true for Q3.

https://twitter.com/x/status/1456505432114180097

Because if they don't, they may not have Zen4 on desktop until Q42022. That would cause a huge drop in market share if they just sat on the same product for two years.

It's like this:

Sell low margin Zen3D to maintain market share, sell all N5 wafers as Genoa/Bergamo, $$$

Rush Raphael to market to expand desktop market share, lose out on some Genoa sales, $$

Going low-margin on a boutique DiY N7 product may make them more money in the long term. Yeah they can sell those N7 wafers as Milan et al. instead for higher short-term profits, but still. Personally I just want Raphael already, bring it on! Their ideas may differ from mine.

You give a great reason why AMD is launching Zen 3D. You give a bad reason for why they should launch it as a loss leader.

 

[maddie]

Still true for Q3.

https://twitter.com/x/status/1456505432114180097

You give a great reason why AMD is launching Zen 3D. You give a bad reason for why they should launch it as a loss leader.

How could overall market share be greater than every subgroup market share? That seems impossible.

 

[jpiniero]

How could overall market share be greater than every subgroup market share? That seems impossible.

The overall is including Consoles.

 

[DrMrLordX]

You give a great reason why AMD is launching Zen 3D. You give a bad reason for why they should launch it as a loss leader.

If you see it on the market, you know it'll have a low-ish margin compared to whatever else they could do with the same silicon. They would have to price it outside of its own market otherwise.