Question Zen 6 Speculation Thread

[adroc_thurston]

You do know that it is the botched L3 dragging the entire platform down alongside the bad fabric it's not the tile tech or 3nm at fault

Cores suck ppballs too.
Fabric-L3 drags the gaming result down harder, though.

 

[Fjodor2001]

nvm

 

[Fjodor2001]

If the cores had the same freq ceiling and v/f characteristics, I'd agree, but people here claim fmax will be above 6GHz. Some saying well above 6GHz. If thats true, sweet spot is going to be more than just +50% IMO. 200W sounds about right, but if thats the sweet spot, I highly doubt AMD is leaving an easy 10% nT perf on the table, if only for the optics. 230W or bust. I guarantee it.

So you think that percentage-wise you will lose more perf when limiting TDP from 230W (assumed default TDP) to ~150-160W on 24C Zen6, compared to when limiting TDP from 170W (default TDP) to ~105W on 16C Zen5?

And to percentage-wise not lose more perf than 16C Zen5 9950X did when going from 170W to ~105W, you’ll have to stay at 200W on Zen6 24C (assuming default TDP is 230W)?

Interesting that the predictions on this vary so widely. Recently 511 said he thought it would be sufficient with 120-135W instead of the 200W you mentioned. That’s a huge difference!

 

[OneEng2]

So you think that percentage-wise you will lose more perf when limiting TDP from 230W (assumed default TDP) to ~150-160W on 24C Zen6, compared to when limiting TDP from 170W (default TDP) to ~105W on 16C Zen5?

And to percentage-wise not lose more perf than 16C Zen5 9950X did when going from 170W to ~105W, you’ll have to stay at 200W on Zen6 24C (assuming default TDP is 230W)?

Interesting that the predictions on this vary so widely. Recently 511 said he thought it would be sufficient with 120-135W instead of the 200W you mentioned. That’s a huge difference!

Zen 5 16c was thermally limited on CB23. It is kinda hard NOT to imagine that Zen 6 24c will not also be thermally limited.

I just assumed that AMD would stick with 200W as they did with Zen 5 for Zen 6 .... this may not be the case, but lets stick with it for a minute.

Zen 6 is getting:

1. A few tweaks to the core
2. A new IOD
3. New and faster memory
4. A new process that should be ~25% PPA improvement over N4P

If you consider that 50% more cores will be running in Zen 6, I think it really comes down to how efficient people believe Zen 6 and N2 will be compared to Zen 5.

It seems difficult to see how fewer cores running at faster clocks could provide better performance within the same power window. My assumption is that all 24 cores will run to max out the score.

 

[Fjodor2001]

Zen 5 16c was thermally limited on CB23. It is kinda hard NOT to imagine that Zen 6 24c will not also be thermally limited.

I just assumed that AMD would stick with 200W as they did with Zen 5 for Zen 6 .... this may not be the case, but lets stick with it for a minute.

Zen 6 is getting:

1. A few tweaks to the core
2. A new IOD
3. New and faster memory
4. A new process that should be ~25% PPA improvement over N4P

If you consider that 50% more cores will be running in Zen 6, I think it really comes down to how efficient people believe Zen 6 and N2 will be compared to Zen 5.

It seems difficult to see how fewer cores running at faster clocks could provide better performance within the same power window. My assumption is that all 24 cores will run to max out the score.

Could it perhaps be that we are talking about different things? Sounds like you’re talking about what the default TDP is likely to be, but that was not what I meant by ”sweet spot” for TDP.

By sweet spot I mean how far you as a user will be able to limit/lower the TDP in BIOS settings (compared to default TDP) without losing much MT perf. This is what I wrote originally:

I’m not so fond of these high 200/230W numbers. I guess it’s to squeeze out the last bit of perf to win some benchmark crown.

I’m hoping that it’ll perform good in perf/watt if not the pushing power consumption that high. E.g. for 9950X you could limit TDP from 170W to ~105W without losing much MT perf. I wonder what the TDP sweet spot will be for 24C Zen6.

So if default TDP will be 200/230W for 24C Zen6, how far can you as a user limit/lower that TDP without losing much MT perf? Similar to the case for 9950X in the quote above.

 

[Det0x]

*double post*

 

[Det0x]

View attachment 132984

My current 16 core Zen5 scores ~45k at 200w PPT @ 5.1ghz static


My prediction for Zen6 in Cinebench R23 MT: (pretty much zero IPC increase in this benchmark)
24 cores, 230w PPT (same as zen4), ~5.5ghz allcore = 73k points

All maxed out OC on watercooling:
80+k points (~6ghz allcore)

93+k points on LN2 (~7ghz allcore)

Will see bigger IPC gains in other benchmarks since Cinebench R23 dont care about latency/bandwidth at all

 

[Hulk]

My current 16 core Zen5 scores ~45k at 200w PPT @ 5.1ghz static
View attachment 133038

My prediction for Zen6 in Cinebench R23 MT: (pretty much zero IPC increase in this benchmark)
24 cores, 230w PPT (same as zen4), ~5.5ghz allcore = 73k points

All maxed out OC on watercooling:
80+k points (~6ghz allcore)

93+k points on LN2 (~7ghz allcore)

Will see bigger IPC gains in other benchmarks since Cinebench R23 dont care about latency/bandwidth at all

How are your temps so low?!!

 

[eek2121]

If you are already in the chart please provide the rest of your prediction data. Moving forward I'm only going to enter data if all data in red font is submitted so the chart is complete.

View attachment 132961

24C + 2C, 230W PPT 5.6ghz avg (6.2 peek), 70,500

 

[Hulk]

Might as well add the ST data.

 

[Det0x]

How are your temps so low?!!

Direct die + LM + mora in window
And this sample is colder than my other delidded samples, even at same powerdraw

You dont want to included my prediction to the chart ?

My prediction for Zen6 in Cinebench R23 MT: (pretty much zero IPC increase in this benchmark)
24 cores, 230w PPT (same as zen4), ~5.5ghz allcore = 73k points

 

[Hulk]

Direct die + LM + mora in window
And this sample is colder than my other delidded samples, even at same powerdraw

You dont want to included my prediction to the chart ?

Sorry, oversight.

I'll add you in. I got preoccupied with something and forgot to put in your data!!!

 

[OneEng2]

Could it perhaps be that we are talking about different things? Sounds like you’re talking about what the default TDP is likely to be, but that was not what I meant by ”sweet spot” for TDP.

By sweet spot I mean how far you as a user will be able to limit/lower the TDP in BIOS settings (compared to default TDP) without losing much MT perf. This is what I wrote originally:

Yep. I was considering the max score Zen 6 24c could achieve at stock settings. You are asking a different question.

Does AMD dynamically lower core voltage in situations where it is thermally limited? If it doesn't, seems like it might be a good feature to add .

 

[Josh128]

Sorry, oversight.

I'll add you in. I got preoccupied with something and forgot to put in your data!!!

1T lets go! Im sticking with a 4.7% IPC gain over Zen 5 here in R23, going "conservative" on the average boost clock at 6.1GHz, for a combined 12% uplift over your 9950X score, 2515.

Looks like you are expecting the absolute worst case scenario for 1T, lol. 2375 is what Ultra 9 285K is doing right now. I CERTAINLY hope its going to be better than that, or AMD is in trouble.

 

[Hulk]

Okay here we go! I added a ST category for all of you thinking Zen 6 is going to get close to 7GHz. Here's your chance to show us you mean what you write

Post all of the data with headers in red font and I'll add your predictions. There is a very good chance I'll be buying this part so I can pop it in the same rig and see how it stacks up to my 9950X.

 

[Hulk]

1T lets go! Im sticking with a 4.7% IPC gain over Zen 5 here in R23, going "conservative" on the average boost clock at 6.1GHz, for a combined 12% uplift over your 9950X score, 2515.

Looks like you are expecting the absolute worst case scenario for 1T, lol. 2375 is what Ultra 9 285K is doing right now. I CERTAINLY hope its going to be better than that, or AMD is in trouble.

I think the 285K architecture is inherently better suited for CB. Zen 6 will improve a bit in this one, but not much without HT in on the action. Just my thoughts. We'll see how it all pans out.

Of course this is just one silly benchmark. Again, all for fun and to put a little skin in the game.

 

[Josh128]

I think the 285K architecture is inherently better suited for CB. Zen 6 will improve a bit in this one, but not much without HT in on the action. Just my thoughts. We'll see how it all pans out.

Of course this is just one silly benchmark. Again, all for fun and to put a little skin in the game.

You have my ST guess in jpeneiro's row, lol.

 

[Det0x]

Okay here we go! I added a ST category for all of you thinking Zen 6 is going to get close to 7GHz. Here's your chance to show us you mean what you write

Post all of the data with headers in red font and I'll add your predictions. There is a very good chance I'll be buying this part so I can pop it in the same rig and see how it stacks up to my 9950X.

View attachment 133075

i get 2301 points at PBO 5725mhz
(static 5725 would be higher)


And i still think we will see pretty much zero IPC gain in Cinebench R23, which means Zen6 should score ~2612points @ PBO 6.5ghz

 

[Hulk]

 

[poke01]

View attachment 133082

Hulks chart is evolving like a Pokémon.

Probably at Level 30 now

 

[Abwx]

2650 pts in ST if they manage to hit 6GHz.

 

[IronLynx]

Hulk, 2800 ST with 6.35 of clock.

 

[Hulk]

2650 pts in ST if they manage to hit 6GHz.

Is that your prediction for the chart or your hope?

 

[Abwx]

Is that your prediction for the chart or your hope?

That s my prediction for both score and peak ST frequency.

 

[Josh128]

2650 pts in ST if they manage to hit 6GHz.

Going all in on on those IPC gains, eh?