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

[DisEnchantment]

Speculate at will

 

[Mopetar]

Apparently I woke up on the darkest timeline side of the bed today since I've learned that MLID was apparently correct.

Tech bros, I'm not feeling so well right now.

 

[Hitman928]

Did they have the same memory BW and/or memory? The latency curves were different.

I’d have to double check. I know the A15 actually had significantly more SLC but the rest I’d have to go back and review.

 

[igor_kavinski]

since I've learned that MLID was apparently correct.

I don't know. Maybe it's all an elaborate prank and Lisa is in on it.

Are we sure that adroc and Kepler don't know Lisa Su personally?

 

[gdansk]

I don't know. Maybe it's all an elaborate prank and Lisa is in on it.

Lisa had the real slide leaked in an attempt to keep expectations realistic.
But it didn't matter.

 

[Hitman928]

Given the variability in the results and how sensitive they are to external factors I don't trust point estimates anymore. Just a 100Mhz clock difference can have a massive effect on IPC estimation (e.g. given a score of 11 points 3 Ghz vs 2.9 Ghz give you 3% difference in IPC!). That's why if I want to compare iso-performance across different models or architectures, I need results from dozens of machines with multiple samples and real-time frequency estimation. Geekbench gives me that. Anandtech database sadly doesn't.

GB database has almost no controls and way too much noise. When Anandtech was giving us reviews every gen with Spec numbers, that was way more reliable. Unfortunately that is going away but I’d still take a known reviewer’s numbers over trying to get something trustworthy out of the GB data base.

 

[trivik12]

1st Strix Point laptop listed for sale at BestBuy.

https://www.bestbuy.com/site/asus-proart-px13-13-3k-touch-screen-laptop-amd-ryzen-ai-9-hx-370-32gb-memory-nvidia-geforce-rtx-4050-1tb-ssd-nano-black/6584436.p?skuId=6584436

Its saying 7/15 release date. Very impressive indeed.

 

[Hulk]

As much of an Intel guy as I am, based on AMD's recent Zen 5 info, Zen 5 is going to be great. Zen 4 is already darn good even 2 years after its release. Zen 5 is a nice bump.

I'll wait for reviews but Zen 5 could be my first AMD rig. Those Handbrake scores in particular are wicked fast.

 

[gdansk]

1st Strix Point laptop listed for sale at BestBuy.

https://www.bestbuy.com/site/asus-proart-px13-13-3k-touch-screen-laptop-amd-ryzen-ai-9-hx-370-32gb-memory-nvidia-geforce-rtx-4050-1tb-ssd-nano-black/6584436.p?skuId=6584436

Its saying 7/15 release date. Very impressive indeed.

I assume that's sarcasm?
But by previous AMD mobile track record they're way ahead this time.

 

[igor_kavinski]

But by previous AMD mobile track record they're way ahead this time.

Yeah mobile and desktop launching together. I don't know AMD history that well. Has this ever happened before?

 

[inquiss]

Yeah there's ondevice cope-pilot exclusivity for Qualcomm.

Why though? Not that I care for the feature myself, but this feels like a backstab from Microsoft. Thinking back to Lisa on stage at the AI event and saying that "they're gonna give TOPS so Microsoft is definitely going to use them right?", then this

 

[Glo.]

Apparently I woke up on the darkest timeline side of the bed today since I've learned that MLID was apparently correct.

Tech bros, I'm not feeling so well right now.

You are experiencing this.

Imagine what Mike Clark has to feel now, after all those years of sleep...

 

[trivik12]

I assume that's sarcasm?
But by previous AMD mobile track record they're way ahead this time.

I said its impressive that we are seeing a laptop listed for sale already. That is unprecedented. No sarcasm.

 

[Glo.]

In all seriousness.

Either AMD marketing team is a bunch of incompetent fools, or AMD as a whole is a bunch of incompetent fools.

 

[exquisitechar]

As much of an Intel guy as I am, based on AMD's recent Zen 5 info, Zen 5 is going to be great. Zen 4 is already darn good even 2 years after its release. Zen 5 is a nice bump.

I'll wait for reviews but Zen 5 could be my first AMD rig. Those Handbrake scores in particular are wicked fast.

I'd definitely wait to see Arrow Lake-S first. There've been some positive news and rumors about it recently and with how underwhelming Zen 5 is, it could very well be the better buy despite the initially disappointing leaked performance projections.

 

[adroc_thurston]

Why though? Not that I care for the feature myself, but this feels like a backstab from Microsoft. Thinking back to Lisa on stage at the AI event and saying that "they're gonna give TOPS so Microsoft is definitely going to use them right?", then this

They want to make WoA real, no matter the cost.

 

[jpiniero]

Yeah mobile and desktop launching together. I don't know AMD history that well. Has this ever happened before?

I suspect Granite Ridge was held back cuz no AI.

 

[TwistedAndy]

I'd definitely wait to see Arrow Lake-S first. There've been some positive news and rumors about it recently and with how underwhelming Zen 5 is, it could very well be the better buy despite the initially disappointing leaked performance projections.

We will get some performance numbers during the upcoming Lunar Lake launch. Desktop Arrow Lake will be announced this fall.

As for the Zen 5, it looks pretty good but boring. Arrow Lake can be a much more interesting product.

 

[CakeMonster]

In all seriousness.

Either AMD marketing team is a bunch of incompetent fools, or AMD as a whole is a bunch of incompetent fools.

IIRC at least two of their previous hypemen have moved on to competitors.

 

[inquiss]

They want to make WoA real, no matter the cost.

Microsoft is such a redactedpartner....



No profanity in tech.


esquared
Anandtech Forum Director

 

[fastandfurious6]

Strix Halo and Fire Range when??

 

[adroc_thurston]

Strix Halo and Fire Range when??

CES'25.

 

[vanplayer]

We will get some performance numbers during the upcoming Lunar Lake launch. Desktop Arrow Lake will be announced this fall.

As for the Zen 5, it looks pretty good but boring. Arrow Lake can be a much more interesting product.

I'd definitely wait to see Arrow Lake-S first.

If you are not interested in AI, Raptorlake would be your long lifecycle product until 2026.

 

[B-Riz]

Apparently I woke up on the darkest timeline side of the bed today since I've learned that MLID was apparently correct.

Tech bros, I'm not feeling so well right now.

That's the fun of it though, isn't it? (Anyone that takes MLID as more than "would be cool if he was right", I got some bridges to sell in NYC)

Back in the Pentium days, we had to rely on magazines, and maybe a tidbit on the early web.

We are spoiled by instant access to information now.

 

[Ghostsonplanets]

I'd definitely wait to see Arrow Lake-S first. There've been some positive news and rumors about it recently and with how underwhelming Zen 5 is, it could very well be the better buy despite the initially disappointing leaked performance projections.

You're in luck then. Less than 9 hours and if Intel doesn't disclose it, others will disclose on their behalf.

 

[SpudLobby]

And I have seen power consumpion of M1 at around 5 watts (3.2Ghz) and M4 at aroudn 7 (4.05 Ghz). So their M4 results I can agree with, the M1 strikes me as extremely low (incidentally i see 3.25 watts when more than one thread is active and the M1 core is runnign at 2.7Ghz, so maybe that's the reading).

I have also seen multplie screenshots from Geekerwan with widly different numbers. So yeah. I don't really see much point in discussing results when they are this imprecise. We need a number of samples with real-time frequency and power readings to talk about these things. The same goes for pretty much any other benchmark or platform of course.

I think these internal measurements be it powermetrics or the internal API geekerwan used are fine for measuring directional shifts *maybe* in CPU power, and just maybe, but overall they’re not that valuable.


To clear this up, I confirmed with someone who I believe can read Chinese (from Chips n Cheese discord) the reason for the different power figures we see — and indeed the 3.62W iPhone A17 Pro result & 7W M4 result — are because those were software modeled power using Apple’s internal APIs.

By contrast, Geekerwan has physically measured the A17 Pro (via removing battery and polling) and the power results are 5.7W minus all idle static/display.

The same (normal and similar to Andrei) methods of power got him to around 4W for the A14 in SpecInt, 4.5W for the A15 in SpecInt, and similar for the A16, and similar ranges for other Android phones, and so this lines up well with Apple pushing clocks on the A17 Pro and a new arch.

I don’t believe M4 is doing 7W total at 4.4GHz ish (minus idle) by any means, it’s just bad software measurement. And given what we know about the iPhone measurements — the software 3.62W vs physically sampled 5.7W - a factor of like 57%, the true M4 system consumption being 11W ballpark makes total sense over the 7W as was measured by a leaker.

https://twitter.com/x/status/1790354528770084945

Like I said, I think these things are okay maybe for measuring internal core power shifts from new generations or even for software metrics on core use, but they should be taken with a grain of salt outside of that.