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"

 

[DrMrLordX]

Run Cinebench, Blender

Honestly, the extra L3 may not make a huge difference in those workloads. Might surprise us, who knows?

 

[Joe NYC]

I get it now. I was totally not factoring in the TSMC aspect. AMD's hands are tied coz they can't force TSMC to do whatever they need. They can't be TSMC's boss. They have to be nice and professional with them and can only just coax them to do their best. Intel has the upper hand here since they have direct control over their foundries and can dedicate the required resources to whatever the CEO/CTO dictate.

Hopefully, the server business boom will allow AMD to buy their own fab, hopefully dedicated for gamers

I think it is a little different. For SoIC stacking AMD is TSMC's partner, to get the technology to mass market. Just like Apple was the partner for N5.

So, it is not about AMD bossing anyone, it is their joint project and both sides will share the fruits of it being successful.

With node shrinks slowing down, advance packaging and stacking in particular is going to be another way for TSMC to manufacture increasingly capable products to customers.

 

[Joe NYC]

It is interesting they show N3 as being available for top die in 2023 which is the first year it will really be available in quantity anyway. Nothing listed for bottom die, does that mean you'd have to use N5 for that?

The timing of the N3 stacking seems quite optimistic to be available on the same year it enters mass production.

 

[Joe NYC]

Okay, at least we know there won't be a latency hit. That's a good sign. More benches please? Is there some embargo on Milan-X benchmarks?

Milan-X is selling retail now (okay technically its still in pre-order retail), why isn't anyone benchmarking it to completion?

AMD (and Microsoft) asked people who have account on Azure Milan-X to not publish any benchmarks.

 

[DrMrLordX]

AMD (and Microsoft) asked people who have account on Azure Milan-X to not publish any benchmarks.

I'm sure they did. And MS made sure it was hard for anyone to buy Milan-X.

 

[deasd]

Finally out. But still... BONIC test?

https://twitter.com/x/status/1482976061986320384

Computer 913263

 

[Ajay]

I'm sure they did. And MS made sure it was hard for anyone to buy Milan-X.

Why would MS do that?

 

[moinmoin]

Why would MS do that?

MS just bought everything it seems. As a result Azure has a competitive edge, at least for the time being.

 

[Doug S]

The timing of the N3 stacking seems quite optimistic to be available on the same year it enters mass production.

Why? The difficulty of stacking is getting the whole procedure down and ironing out the kinks. Doing it on a new process at a somehow higher density is easier once they've gone through the procedure on N7 and then N5. They could easily be working on it already since N3 has entered risk production. From TSMC's perspective what matters is whether they have a customer who wants to use this who will be getting initial allocations of N3 - i.e. if Apple or (perhaps more likely?) Intel is interested.

If they do it would be the same sort of timeline as InFO packaging where initially it trailed mass production but after a few years was available from day one.

 

[Saylick]

Why would MS do that?

See below interview.

 

[Joe NYC]

Why? The difficulty of stacking is getting the whole procedure down and ironing out the kinks. Doing it on a new process at a somehow higher density is easier once they've gone through the procedure on N7 and then N5.

You are right, the process can be accelerating, but still, they have N5 with a 2-year lag from start of the volume production.

Also, some of the roadmaps for SOIC started emerging at the time N3 was assumed to be getting into volume production earlier.

 

[Joe NYC]

See below interview.

One thing from that interview that I found quite interesting is the fact that because the cloud providers are the first ones to get CPUs in highest demand (Milan, Milan-X) and on-premises customers are way behind on the waiting list, it helps the cloud providers sell the cloud solution to the on-premises customers, because that's the only way they can get their hands on the very latest technology...

 

[Ajay]

MS just bought everything it seems. As a result Azure has a competitive edge, at least for the time being.

So AMD, show me the money. MS okay. Profit.

 

[moinmoin]

So AMD, show me the money. MS okay. Profit.

While Amazon was slow on the uptake with Epyc and pushed its exclusive Graviton instances instead, Microsoft Azure likely saw a chance to differentiate itself from other Cloud providers by being early with Epyc instances in each gen which they introduce day and date since Rome. Azure, along Baidu, were the first to announce deployment of Epyc Naples back in 2017, so this sure is some continuity.

I noted before that it can't really be in AMD's interest for Microsoft to hoard everything instead increasing the audience for its products, so there likely is some more substantial cooperation in the backstage.

 

[Zucker2k]

How long before MS buys AMD, at this rate? Risky, I know, with Intel looming large in the background but this doubling down is quite the novelty. Plus, AMD produces other custom chips for MS as well.

 

[DrMrLordX]

Finally out. But still... BONIC test?

Anyone got a 5800X BOINC score for comparison?

 

[Mopetar]

How long before MS buys AMD, at this rate? Risky, I know, with Intel looming large in the background but this doubling down is quite the novelty. Plus, AMD produces other custom chips for MS as well.

They won't because they'd lose the x86 license deals if AMD were acquired by another company. Past behavior on the part of Microsoft would also make it difficult for that kind of deal to get past various regulatory bodies. It would also be fairly expensive since AMD has been doing well. Right now their market cap is about $160 billion, so a buyout would cost Microsoft at least $200 billion. Never mind that it would also damage their reputations with other CPU manufacturers.

 

[Markfw]

Anyone got a 5800X BOINC score for comparison?

Post a thread in DC, I am sure you will get much better response than here.

 

[Zucker2k]

They won't because they'd lose the x86 license deals if AMD were acquired by another company. Past behavior on the part of Microsoft would also make it difficult for that kind of deal to get past various regulatory bodies. It would also be fairly expensive since AMD has been doing well. Right now their market cap is about $160 billion, so a buyout would cost Microsoft at least $200 billion. Never mind that it would also damage their reputations with other CPU manufacturers.

I forgot about the x86 license.

 

[Doug S]

How long before MS buys AMD, at this rate? Risky, I know, with Intel looming large in the background but this doubling down is quite the novelty. Plus, AMD produces other custom chips for MS as well.

If Microsoft wanted to get into the hardware business for real, it would be a lot cheaper to hire a team to build ARM designs that rival the M1. Even if owning AMD was a lot cheaper, they'd still have to go head to head against Intel in the x86 market. If they went all-in on ARM they could slowly phase out support for x86 Windows over a decade or so, and wouldn't have to support anyone else's ARM hardware either.

Not saying they should try to become Apple as there are a long list of reasons why that would never work out for them, but making a twelve digit sized acquisition of AMD could only possibly make sense to me if that was the reason.

 

[Hans Gruber]

Here is what Intel had to say about AMD.

"Alder Lake. All of a sudden...Boom! We are back in the game," exclaims the impish tech CEO. "AMD in the rearview mirror in clients [consumer market]," he adds, "and never again will they be in the windshield; we are just leading the market."

 

[Markfw]

Here is what Intel had to say about AMD.

"Alder Lake. All of a sudden...Boom! We are back in the game," exclaims the impish tech CEO. "AMD in the rearview mirror in clients [consumer market]," he adds, "and never again will they be in the windshield; we are just leading the market."

Yea, it was posted elsewhere here. Intel is delusional. They are not on top, they are fighting for the top.. in CONSUMER and lost in the server world.

 

[Hans Gruber]

Yea, it was posted elsewhere here. Intel is delusional. They are not on top, they are fighting for the top.. in CONSUMER and lost in the server world.

I thought you said I was delusional. You mean to tell me Intel is also delusional?

 

[Markfw]

I thought you said I was delusional. You mean to tell me Intel is also delusional?

Where did I post that you were ? I might have but its not recent that I can find.

 

[Mopetar]

I thought you said I was delusional. You mean to tell me Intel is also delusional?

Where did I post that you were ? I might have but its not recent that I can find.

You didn't, but since he actually is delusional he misread what you wrote and thought you said he was delusional even though you did't.