Question Zen 6 Speculation Thread

[adroc_thurston]

Perf/thread is what matters it you put a ceiling on max number of threads, and it'll be higher even for E-cores than Cinememe Thread spam.

Not a real metric, all that matters is socket throughput@rated power.

 

[inquiss]

But the overall performance gain of AM4 Socket was 124% over the life of the socket makes a CAGR of 17.3%

That information was available to Steve when he made the comments that "platform longevity means nothing".

He just succumbed to Zen 5% mass hysteria, and he was not thinking clearly.

He knew AM4 had lots of gains but didn't know whether a measly 5% would be all you got via AM5 from the upgrade to zen 5 from zen 4. It wasn't known at the time that AM5 would continue to zen6 and zen 7 and the public doesn't know what the performance of those cups will be.

You're implying that since AM4 was good, that AM5 would also be good for platform longevity. If 5% is all you got from the first upgrade gen, and if AMD hasn't yet confirmed zen 6 and zen7 support for AM5 then what he says is totally valid. Not sure what you're missing.

 

[Thunder 57]

Yes. OUCH! . So do you really think 52C NVL-S will cost $2499-$4999 like 32-64C TR which you compared it to? If not, those TR models will be in trouble.

I don't understand why people fail to get that the price will be what the market will tolerate. If performance is high but it is a low volume part the price will be high. Past price is nothing more than a suggestion.

I still want to know what you are going to do with all those cores.

 

[Fjodor2001]

Not a real metric, all that matters is socket throughput@rated power.

Perf/thread does not matter? Then whaddabout Amdahl's law and all that which usually is being thrown around by some.

 

[adroc_thurston]

Then whaddabout Amdahl's law and all that which usually is being thrown around by some.

It's a GPU workload (tiled renderer ffs), you're far and away from hitting Amdahl limits in cinememe.

 

[adroc_thurston]

Amdahl's law only applies when it suits your arguments

Workloads tend to be different.

Noted. So I take it that if Ahmdahl's is N/A even for 32C/64C to 64C/128T TR workloads, then it certainly should be N/A for NVL-S 52C/52T workloads too.

Does that really matter when 24c Z6 will win?

 

[adroc_thurston]

No technical arguments/evidence needed

100% more of suck + a single shrink versus 50% more of awesome with a double shrink (and a packaging/uncore change) are self-explanatory.
Modeling perf for OMR is relatively easy given 5800X, 7900X and the others exist.

 

[reaperrr3]

Yes. OUCH! . So do you really think 52C NVL-S will cost $2499-$4999 like 32-64C TR which you compared it to? If not, those TR models will be in trouble.

So you think AMD will drop the mid-low range TR price down to DT level?

- 52C NVL-S will be hampered heavily by 2ch memory for many workloads you'd want a TR for.
- Nothing is stopping AMD from dropping TR 9K prices, if needed.
- next-gen TRs will likely up core count around those prices by 50%, with higher IPC and clockspeeds on top.

 

[adroc_thurston]

next-gen TRs will likely up core count around those prices by 50%, with higher IPC and clockspeeds on top.

They don't, SP8 caps at 96c.
Much higher perf per core and mem speeds, true.

 

[jpiniero]

- 52C NVL-S will be hampered heavily by 2ch memory for many workloads you'd want a TR for.

You would think that would be obvious.

 

[adroc_thurston]

You would think that would be obvious.

It's not gonna hurt in cinememe much.

 

[jpiniero]

It's not gonna hurt in cinememe much.

It would be a very Intel thing to design a processor solely to win Cinebench... At a time when if you were going to make a meme processor, it should be focused on AI.

 

[adroc_thurston]

It would be a very Intel thing to design a processor solely to win Cinebench... At a time when if you were going to make a meme processor, it should be focused on AI.

NVL is theoretically leadership everything.
Alas, the reality of it.

 

[Joe NYC]

He knew AM4 had lots of gains but didn't know whether a measly 5% would be all you got via AM5 from the upgrade to zen 5 from zen 4. It wasn't known at the time that AM5 would continue to zen6 and zen 7 and the public doesn't know what the performance of those cups will be.

Isn't the point of shopper advice YouTube channel to give advice to shoppers, which sometimes means making rational assumptions about the future?

Steve made horrible assumptions about the future and gave his audience bad advice that (AM5) platform longevity is irrelevant.

Just from 7700x to 9800x3d, the performance uplift is 33%, while dead end Alder Lake had low single digits and another dead end Arrow Lake had a performance regression. Zen 6? Could easily be +25%, already getting to +66% combined.

By discounting platform longevity, he made those 2 horrible platforms (dead end Alder and dead end Arrow) equally valid to long life AM5 platform.

You're implying that since AM4 was good, that AM5 would also be good for platform longevity. If 5% is all you got from the first upgrade gen, and if AMD hasn't yet confirmed zen 6 and zen7 support for AM5 then what he says is totally valid. Not sure what you're missing.

In discussion about longevity, AM5 _HAS_ longevity, even more so than AM4.

As far as gen-on-gen performance improvement, you can take average of past improvements, or you can take a single outlier.

Steve did the dumb thing of taking a single outlier and made a projection on a single outlier rather than taking a broader average, such as that of the generations of AM4 platform from first to last.

Steve also, effectively, equalized the insanity of Intel platform changes to highly consumer friendly AMD approach of extreme platform longevity, by implying (based on faulty assumption) it would be pointless to upgrade any more frequently than longevity of a socket, which in case of AM4 was 5 years (even though the uplift was 124%) and in case of AM5, > 8 years between upgrades.

Further, he implied there would be no CPU only upgrade, only full system upgrades (by discounting platform longevity as irrelevant).

 

[MoistOintment]

Isn't the point of shopper advice YouTube channel to give advice to shoppers, which sometimes means making rational assumptions about the future?

Steve made horrible assumptions about the future and gave his audience bad advice that (AM5) platform longevity is irrelevant.

Just from 7700x to 9800x3d, the performance uplift is 33%, while dead end Alder Lake had low single digits and another dead end Arrow Lake had a performance regression. Zen 6? Could easily be +25%, already getting to +66% combined.

By discounting platform longevity, he made those 2 horrible platforms (dead end Alder and dead end Arrow) equally valid to long life AM5 platform.




In discussion about longevity, AM5 _HAS_ longevity, even more so than AM4.

As far as gen-on-gen performance improvement, you can take average of past improvements, or you can take a single outlier.

Steve did the dumb thing of taking a single outlier and made a projection on a single outlier rather than taking a broader average, such as that of the generations of AM4 platform from first to last.

Steve also, effectively, equalized the insanity of Intel platform changes to highly consumer friendly AMD approach of extreme platform longevity, by implying (based on faulty assumption) it would be pointless to upgrade any more frequently than longevity of a socket, which in case of AM4 was 5 years (even though the uplift was 124%) and in case of AM5, > 8 years between upgrades.

Further, he implied there would be no CPU only upgrade, only full system upgrades (by discounting platform longevity as irrelevant).

What in the world are you talking about.

HUB constantly talks about the value of platform longevity. He often cites it as one of the main reasons to go AMD.

He was speaking in hypotheticals - The smaller the performance delta between the first generation on a socket vs the last generation on the socket, the less value that platform longevity provides.

That is 100% a true statement. He was not making a definitive prediction / statement about the future of AM5. Just that if AM5 generational improvements are as small as the jumps between 7800X3D -> 9800X3D, then the value that socket longevity provides becomes less compared to, say, AM4.

The context of the conversation surrounds 1) Arrow Lake's performance being unknown, 2) Zen 6 being an unknown and that AMD should officially confirm Zen 6 on AM5, 3) Total uncertainty on whether Zen 7 would be on AM5, 4) A complete unknown around Zen 6's (rumored) performance.

 

[dr1337]

I still want to know what you are going to do with all those cores.

use them? just because you don't know how to use a computer beyond email and gaming doesn't mean the rest of us need to suffer for your ignorance/incompetence.

If AMD leaves MT behind, (which was one of ryzen's original and absolute KEY SELLLING POINT), its a mistake.

 

[Thunder 57]

What in the world are you talking about.

HUB constantly talks about the value of platform longevity. He often cites it as one of the main reasons to go AMD.

He was speaking in hypotheticals - The smaller the performance delta between the first generation on a socket vs the last generation on the socket, the less value that platform longevity provides.

That is 100% a true statement. He was not making a definitive prediction / statement about the future of AM5. Just that if AM5 generational improvements are as small as the jumps between 7800X3D -> 9800X3D, then the value that socket longevity provides becomes less compared to, say, AM4.

The context of the conversation surrounds 1) Arrow Lake's performance being unknown, 2) Zen 6 being an unknown and that AMD should officially confirm Zen 6 on AM5, 3) Total uncertainty on whether Zen 7 would be on AM5, 4) A complete unknown around Zen 6's (rumored) performance.

Zen 5 was server first. Zen 6 should benefit client much more with a new IOD and higher clocks. Some say Zen 7 will be on AM5, I am not so sure. With RAM being what it is these days I am more inclined to believe it will be.

Also remember AM4 got that massive boost because of the 5800X3D. AM5 started with that so it is unlikely to see gains that match AM4, at least in gaming.

 

[adroc_thurston]

Some say Zen 7 will be on AM5, I am not so sure

It is AM5.

With RAM being what it is these days I am more inclined to believe it will be.

Has nothing to do with DRAM pricing and everything to do with "there's nothing really better than normal DDR5 for desktop/luggables in Q3'28".

 

[Thunder 57]

It is AM5.

Has nothing to do with DRAM pricing and everything to do with "there's nothing really better than normal DDR5 for desktop/luggables in Q3'28".

I guess a better phrase than pricing would be lack of a future DDR that fits the bill. I don't think anyone really wants LPDDR.

 

[adroc_thurston]

don't think anyone really wants LPDDR.

Well it won't be a choice.
Also LPCAMMs are fine.

 

[Joe NYC]

What in the world are you talking about.

HUB constantly talks about the value of platform longevity. He often cites it as one of the main reasons to go AMD.

Except, in the case I show, Steve said the opposite.

He was speaking in hypotheticals - The smaller the performance delta between the first generation on a socket vs the last generation on the socket, the less value that platform longevity provides.

That is 100% a true statement. He was not making a definitive prediction / statement about the future of AM5. Just that if AM5 generational improvements are as small as the jumps between 7800X3D -> 9800X3D, then the value that socket longevity provides becomes less compared to, say, AM4.

He did make a prediction when he said that for AM5, platform longevity means nothing. Which means, he made a prediction that all of the future AM5 CPUs will be worthless upgrades.

The context of the conversation surrounds 1) Arrow Lake's performance being unknown, 2) Zen 6 being an unknown and that AMD should officially confirm Zen 6 on AM5, 3) Total uncertainty on whether Zen 7 would be on AM5, 4) A complete unknown around Zen 6's (rumored) performance.

It's like if a stock analyst / advisor tells you: "All stocks are equal, because we don't know today what their price will be next year. The companies and the Stock Exchange did not confirm to me what the stock prices will be next year.

How much is that advice worth?

 

[inquiss]

What in the world are you talking about.

HUB constantly talks about the value of platform longevity. He often cites it as one of the main reasons to go AMD.

He was speaking in hypotheticals - The smaller the performance delta between the first generation on a socket vs the last generation on the socket, the less value that platform longevity provides.

That is 100% a true statement. He was not making a definitive prediction / statement about the future of AM5. Just that if AM5 generational improvements are as small as the jumps between 7800X3D -> 9800X3D, then the value that socket longevity provides becomes less compared to, say, AM4.

The context of the conversation surrounds 1) Arrow Lake's performance being unknown, 2) Zen 6 being an unknown and that AMD should officially confirm Zen 6 on AM5, 3) Total uncertainty on whether Zen 7 would be on AM5, 4) A complete unknown around Zen 6's (rumored) performance.

Sanity. The previous poster is forgetting that known information now was not known at the time.

 

[inquiss]

use them? just because you don't know how to use a computer beyond email and gaming doesn't mean the rest of us need to suffer for your ignorance/incompetence.

If AMD leaves MT behind, (which was one of ryzen's original and absolute KEY SELLLING POINT), its a mistake.

What are they going to do with them, not you.

 

[MoistOintment]

Except, in the case I show, Steve said the opposite.



He did make a prediction when he said that for AM5, platform longevity means nothing. Which means, he made a prediction that all of the future AM5 CPUs will be worthless upgrades.



It's like if a stock analyst / advisor tells you: "All stocks are equal, because we don't know today what their price will be next year. The companies and the Stock Exchange did not confirm to me what the stock prices will be next year.

How much is that advice worth?

No, I think this argument is crazy.

The context is incredible obvious (1:00:29-1:05:45)

His off-script statements reflect the sentiment at the time, which was over a year ago - that Zen 6 (Edit: *on AM5) is still (at the time of the video) unconfirmed from AMD. That LGA1851 longevity was a complete unknown (at the time of the video), that Zen 7 was completely unknown, and so that he was not going to use Platform Longevity in his upcoming Zen 5 vs ARL video (which, btw, he hadn't even tested ARL at time of recording).

Anybody who watches HUB regularly knows that he constantly brings up the value of platform longevity in his videos, despite you finding an out of context sentence from a year and a half old video, where he was speaking from a place of unknowns.

 

[adroc_thurston]

use them? just because you don't know how to use a computer beyond email and gaming doesn't mean the rest of us need to suffer for your ignorance/incompetence.

If AMD leaves MT behind, (which was one of ryzen's original and absolute KEY SELLLING POINT), its a mistake.

The usage model for your average DIY PC is gaming.
And sometimes emails.