Discussion Zen 7 speculation thread

[mikegg]

Fun fact, M4 Max MBP running macOS running Parallels running Windows on Arm is still the fastest Windows laptop in the world: https://browser.geekbench.com/v6/cpu/compare/13494385?baseli...

Note that Parallels is only using 14/16 total cores.

 

[mikegg]

If the last 1% of performance was at expense of 10% of power you would not be reporting anything useful. Since you don't know where you are on the curve on either of the processors.

I can't wait until you demonstrate that if you set Strix Halo's clocks low enough to match M4 Pro's efficiency, you'll get the ST speed of an iPhone 10.

 

[Joe NYC]

Why just the core? Will real users experience special physics that their computer will only consume power from the core?

I think you are now arriving (on your own) at the uselessness and futility of measuring ST power efficiency, with many unknown variables that can skew the results.

 

[mikegg]

I think you are now arriving (on your own) at the uselessness and futility of measuring ST power efficiency, with many unknown variables that can skew the results.

It isn't that hard. Plug in an external monitor, close the laptop lid, measure idle power from the wall, measure ST benchmark power from the wall. Subtract power during benchmark from idle power.

This is how Anandtech reviews used to do it. This is how Notebookcheck does it.

It gives a real world experience of ST efficiency.

Just because the test doesn't favor AMD doesn't mean it is useless.

 

[adroc_thurston]

Fun fact, M4 Max MBP running macOS running Parallels running Windows on Arm is still the fastest Windows laptop in the world: https://browser.geekbench.com/v6/cpu/compare/13494385?baseli...

Yo brudda so are Atom Xeons amazing?
You know, lots of cores at slow clocks. Must be how you win in server right?

 

[Bigos]

Why do we talk about apple cores for several pages in a zen 7 thread?

 

[Joe NYC]

Fun fact, M4 Max MBP running macOS running Parallels running Windows on Arm is still the fastest Windows laptop in the world: https://browser.geekbench.com/v6/cpu/compare/13494385?baseli...

Note that Parallels is only using 14/16 total cores.

Still comparing with a tablet.

BTW, there are differences in performance you can derive from tablet, full notebook and Mini PC. Still using GeekBench

For example you can look at these results, doing the same test, Timed LLVM

Strix Halo Framebook Mini PC: 348 sec
Strix Halo HP Laptop: 401 sec
Apple M4 Mac Mini: 498 sec

(I don't understand why Phoronix is using completely different sets of results in different tests, with very little overlap)

Comparison including M4 Mac Mini:

Apple M4 Mac Mini With macOS vs. Intel / AMD With Ubuntu Linux Performance Review - Phoronix

www.phoronix.com

Comparison including 2 Strix Halo form factors

Framework Desktop With AMD Ryzen AI Max Offers Excellent, Linux-Friendly Performance Review - Phoronix

www.phoronix.com

 

[Joe NYC]

Why do we talk about apple cores for several pages in a zen 7 thread?

Because Zen 7 will be the Apple Slayer.

 

[adroc_thurston]

Why do we talk about apple cores for several pages in a zen 7 thread?

People are thirsty for low-effort bait here.

 

[poke01]

I’m sorry but this is a zen7 thread.

I made an arm vs x86 thread for this exact reason, go take the arguments there

 

[Joe NYC]

It isn't that hard. Plug in an external monitor, close the laptop lid, measure idle power from the wall, measure ST benchmark power from the wall. Subtract power during benchmark from idle power.

This is how Anandtech reviews used to do it. This is how Notebookcheck does it.

It gives a real world experience of ST efficiency.

Just because the test doesn't favor AMD doesn't mean it is useless.

M1 has impressive ST performance
M1 has impressive power efficiency

You just are just getting into LaLa land when you try to multiply these 2 to get ridiculous figures.

Sanity check is that these ridiculous numbers don't coincide with MT efficiency results. Which should tell you that you have departed sanity with this line of argument.

If MT is telling you the efficiency of M4 could be 35-40% higher, and ST tells you it is 400% better, one of the 2 is within reason and the other is not.

 

[mikegg]

M1 has impressive ST performance
M1 has impressive power efficiency

You just are just getting into LaLa land when you try to multiply these 2 to get ridiculous figures.

Sanity check is that these ridiculous numbers don't coincide with MT efficiency results. Which should tell you that you have departed sanity with this line of argument.

If MT is telling you the efficiency of M4 could be 35-40% higher, and ST tells you it is 400% better, one of the 2 is within reason and the other is not.

Once again, MT efficiency is a function of how many cores you have. If Apple cares enough, they can stuff 64 cores in their consumer chips as well and get out of this world MT efficiency.

 

[adroc_thurston]

Once again, MT efficiency is a function of how many cores you have

Does that make Atom-based Xeons awesome then.
They gotta right, gajillion cores@low clk.

 

[itsmydamnation]

Once again, MT efficiency is a function of how many cores you have. If Apple cares enough, they can stuff 64 cores in their consumer chips as well and get out of this world MT efficiency.

if AMD cared to have st perf above all else they could size there oooe engine the same size as apple and have out of this world T1.........

 

[poke01]

if AMD cared to have st perf above all else they could size there oooe engine the same size as apple and have out of this world T1.........

That’s interesting. Honestly, AMD could probably focus on ST too. Their cores are very small compared to Intels but as long as Intel is stuck in a rut, AMD can just save costs and focus Zen on server.

 

[Joe NYC]

Once again, MT efficiency is a function of how many cores you have. If Apple cares enough, they can stuff 64 cores in their consumer chips as well and get out of this world MT efficiency.

So you understand that if you take 14 cores, add 1 core by reducing the power of the 14 cores, you bring the 14 cores to higher efficiency zone.

Then why is it so difficult to understand that 1 core running ST, if you limit the power it can use, you bring it to higher efficiency zone?

BTW, the difference between 14 core Mac Pro and 16 core Strix Halo (in number of cores) is not that great. By going from 14 cores to 16 cores, power efficiency of M4 is not going to jump from 37% advantage to 400% advantage (you have been dreaming about in your ST line of argument)

 

[511]

That’s interesting. Honestly, AMD could probably focus on ST too. Their cores are very small compared to Intels but as long as Intel is stuck in a rut, AMD can just save costs and focus Zen on server.

Intel can grow Atom in size it's merely 1.1mm2 Core+L2 N3B and even smaller on 18A like <1mm2 P core is just very difficult to scale down.

 

[Joe NYC]

That’s interesting. Honestly, AMD could probably focus on ST too. Their cores are very small compared to Intels but as long as Intel is stuck in a rut, AMD can just save costs and focus Zen on server.

If die size if going up by 40% (from 70 mm2 to 98 mm2), the core count is going up by 33% and there is some density increase from new node, there should be some extra transistors and die size per core with Zen 7

 

[mikegg]

Then why is it so difficult to understand that 1 core running ST, if you limit the power it can use, you bring it to higher efficiency zone?

Why do you think I don't understand this? When did I ever disagree with this statement?

All I'm asking you is when you lower the clocks to achieve the same perf/watt as an M4 Pro, what is the speed of Strix Halo?

 

[adroc_thurston]

Why do you think I don't understand this? When did I ever disagree with this statement?

All I'm asking you is when you lower the clocks to achieve the same perf/watt as an M4 Pro, what is the speed of Strix Halo?

So yeah about SRF and CWF.
Atom Xeons must be winning the nT race by a country mile right?
After all, you only need tons of cores clocked lower.

 

[mikegg]

So yeah about SRF and CWF.
Atom Xeons must be winning the nT race by a country mile right?
After all, you only need tons of cores clocked lower.

You keep repeating this but that's exactly what would happen if Intel puts enough atoms into a single chip.

What you're missing is area efficiency, which is a factor when it comes to MT scaling. Ultimately what matters for a server chip is area x efficiency x performance of one core x # of cores. Area directly correlates to cost of the chip.

So you put enough atoms into a single chip and it can win nT race by a mile. It's likely terribly inefficient area wise for the performance.

 

[adroc_thurston]

You keep repeating this but that's exactly what would happen if Intel puts enough atoms into a single chip.

Which is why it got stomped by AMD dense parts. Wonder why.

So you put enough atoms into a single chip and it can win nT race by a mile. It's likely terribly inefficient area wise for the performance.

I wonder if cores had some kinda power floor and other cores had it better n stuff.

 

[adroc_thurston]

(newsflash, core power isn't defined by v/f placement alone)

 

[LightningZ71]

Er, there's a point in there where the communication mesh between that "ton" of cores starts to both consume more and more power AND reach saturation enough to negatively affect performance. In addition, once you start to push past 16-24 cores, unless you have a very embarrassingly parallel workload that doesn't exist but in tiny niches or benchmark labs, or in specific hyperscalar DCs, the efficiency of adding more cores from a software throughput point of view begins to taper off.

 

[Meteor Late]

Translation: AMD is so far behind in ST efficiency and performance that burying one's head in the sand is a valid option.

This is literally what he is doing, yes.

It would be better to say AMD prioritizes server much more than client, thus focusing more on extracting more MT with SMT, etc than Apple, which is client focused. But no, gotta say ST efficiency doesn't matter, of course.