Question Zen 6 Speculation Thread

[adroc_thurston]

Everything else I think might be close.

you can't get 'close' when you can't clock.

 

[MadRat]

Caching might drive performance with RAM at the ridiculous prices. Of course every reviewer has buttloads of RAM rather than accurate reflections of Joe Consumer.

 

[Hulk]

you can't get 'close' when you can't clock.

I'm sure they will both have clock domains.

 

[Kryohi]

you can't get 'close' when you can't clock.

A hypothetical 6.5GHz vs 6GHz fmax is an 8% difference, even assuming linear scaling of performance with frequency I would say 8% is close enough.
We're not in the 2GHz era anymore, when a 500MHz difference was huge.

 

[adroc_thurston]

A hypothetical 6.5GHz vs 6GHz fmax is an 8% difference, even assuming linear scaling of performance with frequency I would say 8% is close enough.

NVL-S is 5.7G.

 

[Kryohi]

NVL-S is 5.7G.

So you'd expect a >10% fmax difference, while having similar PPC? That would indeed be bad for Intel

 

[adroc_thurston]

So you'd expect a >10% fmax difference, while having similar PPC?

similar or higher yeah.

That would indeed be bad for Intel

This is *already* a problem for Intel; they're on N3 with worse v/f than AMD on N4p.

 

[techjunkie123]

Can amd clock tho??

similar or higher yeah.

This is *already* a problem for Intel; they're on N3 with worse v/f than AMD on N4p.

Is N3B expected to clock higher at the limit than N4P though? Not sure if has anyone gotten high clocks out of N3B.

N3P is a different story ofc.

 

[adroc_thurston]

Can amd clock tho??

yea

Is N3B expected to clock higher at the limit than N4P though?

Apple gained 17% fmax there so yeah.

Not sure if has anyone gotten high clocks out of N3B.

It's a much much faster node than anything N5/N4.

N3P is a different story ofc.

N3p is maybe 5% (optimistically) faster than N3b.

 

[Hulk]

ST "turbo" clocks aren't important. It's MT clocks that win the day.

 

[adroc_thurston]

ST "turbo" clocks aren't important

most insane statement maybe ever

 

[CouncilorIrissa]

ST "turbo" clocks aren't important. It's MT clocks that win the day.

What.
You do realise that designers of mobile SoCs spend 3-4mm^2 worth of area on a single core with the sole goal of bursting up to 4.7 GHz for a thousandth of a second to deliver responsive experience? Every time you open a web page, or interact with $chatAppName in any way, you're essentially running a 1t burst workload.

 

[511]

N3p is maybe 5% (optimistically) faster than N3b.

I think you meant N3E not N3B

ST "turbo" clocks aren't important. It's MT clocks that win the day.

rare usage compared to ST For client ST matters the most if MT was superior we would be getting 48C Skymont not 8+16 ARL

 

[adroc_thurston]

I think you meant N3E not N3B

nope.

 

[Hulk]

rare usage compared to ST For client ST matters the most if MT was superior we would be getting 48C Skymont not 8+16 ARL

I meant aren't important for the comparison of top parts, they are important for performance of course. I was referring to the comparison of parts that was carrying through the last few posts. For most top end parts today they will boost for a couple of cores within a few hundred MHz of one another. When it comes to frequency I'm more concerned with how high they clock when all of those cores are loaded.

To better illustrate my point see the charts below. Both start at 5.7GHz for ST. No difference. But by the time all cores are loaded there is a 500MHz gap. Now to be fair I don't think they put much if any load on these cores when running this test, but it does show at stock how high the core will clock as it gets loaded.

This, more than ST peak for a core or two is also very important to some people, like myself.

That's what I get for not fully communicating my thought!

 

[msj10]

NVL-S is 5.7G.

ah yes Intel is paying all this money for N2 just to get the same frequency as N3B lmao. sure thing buddy.

 

[adroc_thurston]

ah yes Intel is paying all this money for N2 just to get the same frequency as N3B lmao

Yeah?
They're already paying for N3 to have the same freq as AMD at more voltage.

sure thing buddy.

Skill issue.
Maybe Intel engineering needs to figure out how to Not Suck.

 

[exquisitechar]

ah yes Intel is paying all this money for N2 just to get the same frequency as N3B lmao. sure thing buddy.

Well, it’s a new core and design compared to ARL. There’s no reason that it can’t clock the same (or even lower) on a better node. Presumably, the implication is that Intel messed up the physical design or what have you. No idea whether the 5.7GHz figure is accurate or not, to be clear.

 

[CakeMonster]

Caching might drive performance with RAM at the ridiculous prices. Of course every reviewer has buttloads of RAM rather than accurate reflections of Joe Consumer.

Zen6 and Zen6 3D probably ought to be benchmarked with 4800/5200/5600/6000 in addition to the new default/sweet spot considering the kind of memory market prospective buyers will have to deal with.

 

[Covfefe]

Well, it’s a new core and design compared to ARL. There’s no reason that it can’t clock the same (or even lower) on a better node. Presumably, the implication is that Intel messed up the physical design or what have you. No idea whether the 5.7GHz figure is accurate or not, to be clear.

When's the last time a new intel core lost 10%+ frequency iso-node?

 

[jpiniero]

When's the last time a new intel core lost 10%+ frequency iso-node?

It's not going to be iso-node obviously.

 

[Kaffeekenan]

NVL-S is 5.7G.

Do you mean Singlethread?

 

[Covfefe]

It's not going to be iso-node obviously.

Obviously, but claiming that Nova Lake gets zero clockspeed increase when it's moving from N3B to N2 is tantamount to saying that Coyote Cove clocks 10% worse than Lion Cove, iso-node. It's possible, but it would be a historically bad.

 

[Joe NYC]

you can't get 'close' when you can't clock.

Interesting that the newer Intel CPUs (MTL, LNL, PTL, ARL) all have either same or lower clock speeds compared to years old ADL, RPL. While moving to better process nodes.

 

[Joe NYC]

ST "turbo" clocks aren't important. It's MT clocks that win the day.

Both are quite important. For client, it is ST boost clock that disproportionately affects the performance.