Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)

[DisEnchantment]

Speculate at will

 

[Fjodor2001]

It doesn't matter you do.
All that matters is what ships fused out of the fab.
And that is 170W GNR parts.

It matters, because a lot of people don't want to run the CPU at higher TDP than necessary. On Zen4, you would not get much lower performance if running at 120W compared to 170W.

So it would be interesting to know if the same applies to Zen5.

If that is not the case for Zen5, then it means that if running at 120W TDP the perf/watt difference between Zen5 vs Zen4 is not as big as the >20% you claimed at 170W TDP. So for those people intending to run at 120W TDP, Zen5 will not be such a big performance improvement compared to Zen4.

 

[adroc_thurston]

It matters, because a lot of people don't want to run the CPU at higher TDP than necessary. On Zen4, you would not get much lower performance if running at 120W compared to 170W.

People run stuff stock.
Stock is 170W aka what GNR ships at.

If that is not the case for Zen5, then it means that if running at 120W TDP the perf/watt difference between Zen5 vs Zen4 is not as big as the >20% you claimed at 170W TDP. So for those people intending to run at 120W TDP, Zen5 will not be such a big performance improvement compared to Zen4.

You're shadowboxing with stuff that's not real.
Don't.

 

[Fjodor2001]

People run stuff stock.
Stock is 170W aka what GNR ships at.

Not everyone does. Zen5 can run at 120W if desired by changing the max TDP BIOS setting, just like Zen4.

You're shadowboxing with stuff that's not real.

It's reality for those that will run at 120W.

Since you're avoiding the question, my assumption is that the perf/watt difference between Zen5 vs Zen4 will not be so big at 120W, compared to at 170W. So for those people, a discounted Zen4 might be a better option than Zen5.

 

[gdansk]

a discounted Zen4 might be a better option than Zen5

If Zen 5's PPW isn't good at the astronomically high limit of 10W per core instead of 14.5W per core, then they are really going to struggle in even lower watt per core like servers.

Somehow it's good at 14.5W and 5W but sucks at 10W? Unlikely.

 

[adroc_thurston]

Not everyone does. Zen5 can run at 120W if desired by changing the max TDP BIOS setting, just like Zen4.

Again, people run chips ootb.

Since you're avoiding the question, my assumption is that the perf/watt difference between Zen5 vs Zen4 will not be so big at 120W, compared to at 170W. So for those people, a discounted Zen4 might be a better option than Zen5.

You're shadowboxing stuff that doesn't exist once again.

 

[Fjodor2001]

Again, people run chips ootb.

Again, not everyone does.

You're shadowboxing stuff that doesn't exist once again.

Correct doesn't exist since not released yet. But will not be shadowboxing once released.

But I think you avoiding the question answers it indirectly anyway.

 

[adroc_thurston]

Again, not everyone does.

Outliers don't matter.

Correct doesn't exist since not released yet. But will not be shadowboxing once released.

Because you fight whatever imaginary stuff that exists in your brain and is not real.

 

[CakeMonster]

Is 6000 still the memory sweet spot or will that be possible to run faster reliably? I plan on getting the vanilla 16c since I don't want to bother with thread prioritization schemes anyway, and faster memory is preferable when not on X3D.

 

[StefanR5R]

Since you're avoiding the question,

No he did not. He answered it fully.
– There are no such projections.
– There are no such samples.

 

[adroc_thurston]

Is 6000 still the memory sweet spot or will that be possible to run faster reliably?

Yea.

and faster memory is preferable when not on X3D.

No, this core copes helluva lot better with it.

No he did not. He answered it fully.
– There are no such projections.
– There are no such samples.

He's a bit obsessed so no need to engage him.

 

[Fjodor2001]

No he did not. He answered it fully.
– There are no such projections.
– There are no such samples.

He did not answer the question about perf/watt at 120W TDP for Zen5 vs Zen4.

If there are Zen5 samples, they can be set to max 120W TDP in the BIOS settings, and then tests can be performed.

If he does simply not know, and does not have access to samples for testing, then he might as well say so. I'm fine with that.

 

[adroc_thurston]

they can be set to max 120W TDP in the BIOS settings,

You can't.
It's a ref platform.

 

[moinmoin]

they can be set to max 120W TDP in the BIOS settings

I wish that were as commonplace as you seem to assume it is. Even plenty consumer systems don't offer such BIOS settings. Don't know what makes you think all reference boards would.

 

[Fjodor2001]

You can't.
It's a ref platform.

You mean the Eng CPU sample you got only runs on a ref MB that you got along with it, and that ref MB does not allow changing max TDP setting from stock? Yeah, in that case there’s no way of knowing.

Then we can only make educated guesses of the perf/watt improvement at 120W TDP for Zen5 vs Zen4, based on e.g. CPU arch and chip process tech changes from Zen4 to Zen5. Anyone that has any guesstimates?

Finally, if you only got Eng sample, there’s no way of knowing other results from final CPU version anyway either. So that should apply for your other Zen5 perf info predictions too, if you’re basing it on Eng sample testing alone.

 

[adroc_thurston]

You mean the Eng CPU sample you got only runs on a ref MB that you got along with it, and that ref MB does not allow changing max TDP setting from stock?

It's not really tied to that RVP physically but no UEFI should currently support GNR besides the AMD ref one, so yeah.

 

[Thunder 57]

Wish I was.

It doesn't matter you do.
All that matters is what ships fused out of the fab.
And that is 170W GNR parts.

120W GNR doesn't exist.

Why would you wish you wrong arong about another "Conroe moment"? You're just being a curt (removed).



We do not allow profanity in tech section.

-Moderator Aigo

 

[CouncilorIrissa]

This thread is in dire need of mod intervention.

 

[adroc_thurston]

Why would you wish you wrong arong about another "Conroe moment"?

Because it makes the market will be toast for a while.

 

[clemsyn]

Because it makes the market will be toast for a while.

Intel is toast

 

[adroc_thurston]

Intel is toast

no but it's not good

 

[FlameTail]

Looking at this chart, Strix Point is clearly not a direct successor to Phoenix/Hawk Point. Rather Strix Point is the beginning of. whole new tier of APU. Real successor of Hawk Point will be Kraken Point.

Speaking of which, we just got a leak:

Kraken Point has 4x Zen 5 + 4x Zen 5c core config

https://x.com/Olrak29_/status/1750579185180119157?s=20

 

[adroc_thurston]

Strix Point is clearly not a direct successor to Phoenix/Hawk Point.

It is and it isn't.
More expensive for more expensive laptops but that's it.
Same niche otherwise.

 

[Abwx]

View attachment 92487
Looking at this chart, Strix Point is clearly not a direct successor to Phoenix/Hawk Point. Rather Strix Point is the beginning of. whole new tier of APU. Real successor of Hawk Point will be Kraken Point.

Speaking of which, we just got a leak:

https://x.com/Olrak29_/status/1750579185180119157?s=20

Kracken point is the successor of Phoenix2, that is , a low cost solution, Strix point is to replace Phoenix1, aka Hawk point, i bet that the former will have either 6 or 8 CUs GPU.

 

[adroc_thurston]

Kracken point is the successor of Phoenix2, that is , a low cost solution

It's cheaper yes but it doesn't target the same laptop SRP.
KRK1 is lot more expensive.

 

[Abwx]

It's cheaper yes but it doesn't target the same laptop SRP.
KRK1 is lot more expensive.

I edited my post to add my expectations for the GPU part, beside there will be cut down versions with 6C to serve the lower end of the market, at this point 6C should be the minimal offering unless yields are horrendous, but i dont think that there will be a native 6C part.