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

[DisEnchantment]

Speculate at will

 

[Mopetar]

Reaching Apple's IPC level does not guarantee reaching Apple's efficiency level.

I think Apple cares more about that then they do chasing after hertz.

I also think that both companies have different approaches to getting their increased IPC and that simply trying to emulate someone else's approach with consideration for your own strengths is a sure way to fail.

 

[adroc_thurston]

I think Apple cares more about that then they do chasing after hertz.

their new P-core is literally designed around chasing hertz.

I also think that both companies have different approaches to getting their increased IPC

Big wide core with chungus reorder buffer and a very mean front-end is a fairly universal recipe.
It's how you execute on it that makes all the difference.

consideration for your own strengths is a sure way to fail.

See CPUs have largely converged on being deep and wide and mean OoO machines.
Funny outliers like Netburst or POWER6 or Bulldozer or Niagara no longer exist.

 

[poke01]

their new P-core is literally designed around chasing hertz

Isn’t Apple latest P core in the M3 just an upgraded Everest?

Edit: Apple uses the same code names for the p and e cores for A17/M3 as they did for A16.

 

[DisEnchantment]

What a dream it was for Greymon/Vegeta and the others.

In the dream, it was an asian dude based in the mainland with connections to the packaging facilities.
In the dreams RDNA3 perf was vivid and absolutely demolished everything on the market.

The RDNA3 dream came to a stop and in the ensuing reality she is an Indian female in Bangalore now discovering out Zen 5 facts .

Wonder she/he/they/them? will wake up and another dream ensues?

This is wild, inception is real. Dream within a dream.
Just head over to the RDNA3 thread, dream in dream. Wake up from RDNA3 in Zen5 thread and wake from Zen5 in Zen6 thread.

Right now in this thread we need to spin the top or let ourselves free fall from the chair just to be sure.

 

[adroc_thurston]

Isn’t Apple latest P core in the M3 just an upgraded Everest?

No, it's a new one, bigger and designed for higher clockspeeds.

In the dreams RDNA3 perf was vivid and absolutely demolished everything on the market.

Real bad bad timing saying that.

 

[APU_Fusion]

I mean …

What a dream it was for Greymon/Vegeta and the others.

In the dream, it was an asian dude based in the mainland with connections to the packaging facilities.
In the dreams RDNA3 perf was vivid and absolutely demolished everything on the market.

The RDNA3 dream came to a stop and in the ensuing reality she is an Indian female in Bangalore now discovering out Zen 5 facts .

Wonder she/he/they/them? will wake up and another dream ensues?

This is wild, inception is real. Dream within a dream.
Just head over to the RDNA3 thread, dream in dream. Wake up from RDNA3 in Zen5 thread and wake from Zen5 in Zen6 thread.

Right now in this thread we need to spin the top or let ourselves free fall from the chair just to be sure.

what if … like Zen 5 … like spelled arrow lake. Zen and arrow lake are like totally symmetrical and in love man. One is the other as the other is the one.

I IPC You

 

[Glo.]

What a dream it was for Greymon/Vegeta and the others.

In the dream, it was an asian dude based in the mainland with connections to the packaging facilities.
In the dreams RDNA3 perf was vivid and absolutely demolished everything on the market.

The RDNA3 dream came to a stop and in the ensuing reality she is an Indian female in Bangalore now discovering out Zen 5 facts .

Wonder she/he/they/them? will wake up and another dream ensues?

This is wild, inception is real. Dream within a dream.
Just head over to the RDNA3 thread, dream in dream. Wake up from RDNA3 in Zen5 thread and wake from Zen5 in Zen6 thread.

Right now in this thread we need to spin the top or let ourselves free fall from the chair just to be sure.

Yep.

Zen 5 will be barely faster than RTX 3070.

 

[RnR_au]

I think its obvious a lot of people stand to gain by grasping at straws to build up clout and thus generate streams of revenue/help bolster their existing income streams.

What I want to know is how much @Kepler_L2 owns of Anandtech....

https://twitter.com/x/status/1773772359586963909

 

[Glo.]

What I want to know is how much @Kepler_L2 owns of Anandtech....

https://twitter.com/x/status/1773772359586963909

he already wrote: 40%.

 

[Ghostsonplanets]

their new P-core is literally designed around chasing hertz.

Big wide core with chungus reorder buffer and a very mean front-end is a fairly universal recipe.
It's how you execute on it that makes all the difference.

See CPUs have largely converged on being deep and wide and mean OoO machines.
Funny outliers like Netburst or POWER6 or Bulldozer or Niagara no longer exist.

So Zen 5 is the final send-off AM5 Platform with a deep, wider and smarter processor. And Zen 6 debuts the AM6 platform with a just as deep, just as wide and even more smart processor, with focus on the back-end and the uncore (New IOD with RDNA 4/5 IGP and others changes), with also the Zen 6 Desktop platform being basically a small MI/Venice in terms of structure?

AMD has a quite interesting roadmap, if so. Basically aiming for leadership from top to bottom. It's an interest contrast to Intel, which appears almost lethargic, if not a very slow ship to steer. But, as they always say, execution is key.

 

[Joe NYC]

What a dream it was for Greymon/Vegeta and the others.

In the dream, it was an asian dude based in the mainland with connections to the packaging facilities.
In the dreams RDNA3 perf was vivid and absolutely demolished everything on the market.

The RDNA3 dream came to a stop and in the ensuing reality she is an Indian female in Bangalore now discovering out Zen 5 facts .

Wonder she/he/they/them? will wake up and another dream ensues?

This is wild, inception is real. Dream within a dream.
Just head over to the RDNA3 thread, dream in dream. Wake up from RDNA3 in Zen5 thread and wake from Zen5 in Zen6 thread.

Right now in this thread we need to spin the top or let ourselves free fall from the chair just to be sure.

There were two possible outcomes:
1. RDNA3 would perform at or near 3 GHz as many (including at AMD) expected
2. It would have to be clocked down to much less - for reasons.

RDNA3 fail was quite unique in recent AMD execution track record.

Making conclusions today, like it was obvious this unique event would take place is a little bit like a lottery winner lecturing those who did not win as being stupid.

 

[Kepler_L2]

So Zen 5 is the final send-off AM5 Platform with a deep, wider and smarter processor. And Zen 6 debuts the AM6 platform with a just as deep, just as wide and even more smart processor, with focus on the back-end and the uncore (New IOD with RDNA 4/5 IGP and others changes), with also the Zen 6 Desktop platform being basically a small MI/Venice in terms of structure?

AMD has a quite interesting roadmap, if so. Basically aiming for leadership from top to bottom. It's an interest contrast to Intel, which appears almost lethargic, if not a very slow ship to steer. But, as they always say, execution is key.

AFAIK AMD has not yet decided if Zen6 will be on AM5 or a new platform.

 

[branch_suggestion]

Only if the update includes Lisa jumping on Nvidia's track, because there's no way RDNA4 will get any closer to Nvidia.

In relevant metrics (PPAC) it should certainly be better than RDNA3 was relative to the comp, as RDNA4 should actually work as intended.
Absolute perf of course not as it is a midrange gen.

 

[Ghostsonplanets]

AFAIK AMD has not yet decided if Zen6 will be on AM5 or a new platform.

That's nice ngl! 3 generations per socket/platform would be really great. AM6 debutting with Zen 6 would also be an awkward timing due to DDR6/LPDDR6 TTM. Would be better to wait a bit and have a more mature solution.

 

[FlameTail]

That's nice ngl! 3 generations per socket/platform would be really great. AM6 debutting with Zen 6 would also be an awkward timing due to DDR6/LPDDR6 TTM. Would be better to wait a bit and have a more mature solution.

LPDDR6 will hit the market in late 2025.

 

[DisEnchantment]

Making conclusions today, like it was obvious this unique event would take place is a little bit like a lottery winner lecturing those who did not win as being stupid.

Excuse me, I have not been making such conclusions or predictions or any statement of fact.

But many are patrolling the forums handing out fact sheets on Zen 5 every third post, according to which my order for the 16C part would be delivered any time now. I am only not sure about the cooler.

In fact, these statements are never incorrect, just AMD had last minute changes indeed.

 

[Joe NYC]

Excuse me, I have not been making such conclusions or predictions or any statement of fact.

But many are patrolling the forums handing out fact sheets on Zen 5 every third post, according to which my order for the 16C part would be delivered any time now. I am only not sure about the cooler.

In fact, these statements are never incorrect, just AMD had last minute changes indeed.

That seems like a little bit tongue in cheek.

There was, in fact, some internal breakdown inside AMD, that lead to RDNA3 underdelivering. Those people / leakers who predicted higher RDNA3 performance were in fact right. Those were the AMD internal design goals and performance targets.

If the upcoming RDNA3+ parts indeed clock at or near 3 GHz, it will be a confirmation of the above.

So, posts claiming some moral high ground, sitting on a high horse, for getting RDNA3 right or not falling for RDNA3 hype while disparaging those who expected AMD to meet its design goals - those posts don't really sit right with me.

 

[Abwx]

That seems like a little bit tongue in cheek.

There was, in fact, some internal breakdown inside AMD, that lead to RDNA3 underdelivering. Those people / leakers who predicted higher RDNA3 performance were in fact right. Those were the AMD internal design goals and performance targets.

If the upcoming RDNA3+ parts indeed clock at or near 3 GHz, it will be a confirmation of the above.

So, posts claiming some moral high ground, sitting on a high horse, for getting RDNA3 right or not falling for RDNA3 hype while disparaging those who expected AMD to meet its design goals - those posts don't really sit right with me.

That s not much difficult to spot what went wrong with RDNA3, the chips still clock quite high but methink that for some reasons there were some critical circuitries within the GPU that didnt clock accordingly with the projected stock voltage.

It can be only 10% of the whole chip but they had to increase voltage by about 10%
for the whole to get the flawed parts working with inconditional stability, and at this rate this did increase the comsumption by 20% wich led to the missed perf/watt target, guess that it should be the main fix that will be done with RDNA3.5.

 

[Doug S]

No, it's a new one, bigger and designed for higher clockspeeds.

No its not. It has been gaining clock speed as expected via process improvements, there has been no design specific to higher clock speeds. Do you really think that Intel and AMD alone are able to design CPUs that clock higher than 4 GHz, and M3 has the highest clocked cores Apple could possibly design? If you do you're hopelessly ignorant.

Do you really not understand why they have targeted lower clock speeds when their cores are explicitly designed to go into phones and the same core is copied unchanged into Apple Silicon CPUs - just with a bit higher clock since it can have a little higher power budget?

 

[Kepler_L2]

No its not. It has been gaining clock speed as expected via process improvements, there has been no design specific to higher clock speeds. Do you really think that Intel and AMD alone are able to design CPUs that clock higher than 4 GHz, and M3 has the highest clocked cores Apple could possibly design? If you do you're hopelessly ignorant.

Do you really not understand why they have targeted lower clock speeds when their cores are explicitly designed to go into phones and the same core is copied unchanged into Apple Silicon CPUs - just with a bit higher clock since it can have a little higher power budget?

M3 increased L1 cache latency to allow for higher clocks.

 

[poke01]

M3 increased L1 cache latency to allow for higher clocks.

The biggest question mark is why didn’t Apple give new codenames for the M3/A17 CPUs cores. They use the codenames as A16.

Anyhow, Apple clocking their cores higher is needed to compete in desktop. This is where AMD and Intel have huge lead.

 

[FlameTail]

Anyhow, Apple clocking their cores higher is needed to compete in desktop. This is where AMD and Intel have huge lead.

Why does it matter? If Apple is able to have higher IPC than AMD/Intel, it makes up for the clock speed disadvantage.

 

[adroc_thurston]

No its not.

Yeah it is, L1 absolute cycle count latency is 33% up (3 -> 4) and average mispredict penalties are up.
The thing has a struggle moment in SIR due to that.

 

[MoogleW]

Only if the update includes Lisa jumping on Nvidia's track, because there's no way RDNA4 will get any closer to Nvidia.

But we have AMD making massive strides of 2 gens in 2 years according to the rumors

 

[Hans Gruber]

The AMD playbook should be to release Zen 5 by June at the latest. That would give AMD the performance crown across all categories for 4-6 months. After Arrowlake releases, AMD should release a Zen 5 refresh on TSMC N3P. They could have a 45w 6C/12T and a 8C/16T cpu as well as a bunch of 65w/95w TDP SKU's. I am thinking Feb/March 2025 for TSMC N3P silicon. A 16C/32T Zen 5 CPU with a 95w TDP would be brutal for Intel no matter how crazy Arrowlake performance could be. People forget that Intel will be on true 5nm silicon with Arrowlake. This means Intel will have a huge efficiency gain over 10nm. The efficiency of Zen CPU's on TSMC has been their strongest point since moving to TSMC for Zen 2.