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

[DisEnchantment]

Speculate at will

 

[igor_kavinski]

ASUS System Product Name vs ASUS System Product Name - Geekbench

WTF, is this the year of unbalanced uarchitectures or what???

Does ARL look like an INT heavy uarch from these scores?

 

[gdansk]

ASUS System Product Name vs ASUS System Product Name - Geekbench

WTF, is this the year of unbalanced uarchitectures or what???

Does ARL look like an INT heavy uarch from these scores?

No one in human history has complained about having more integer performance. It's what you want.

 

[Thibsie]

Now stupid question, just for the fun.

Since GNR is of a lower rev than Turin,
A later could (huh, can dream hu?) bring some performance,
Maybe X3D would be later rev ,???

 

[gdansk]

Now stupid question, just for the fun.

Since GNR is of a lower rev than Turin,
A later could (huh, can dream hu?) bring some performance,
Maybe X3D would be later rev ,???

No

 

[DavidC1]

ASUS System Product Name vs ASUS System Product Name - Geekbench

WTF, is this the year of unbalanced uarchitectures or what???

Does ARL look like an INT heavy uarch from these scores?

"Unbalanced" chip is actually Zen 5. It got ~10% in Int and 25% in FP performance. Hence Arrowlake looks comparatively better in Int.

 

[Josh128]

Get back to us with your analysis in October when Intel releases their Arrow Lake CPU's. I think we may be looking at Conroe 2.0

Based on what? Definitely not the Geekbench scores released thus far. Looks like Zen 5 with better MT...and thats probably exactly what it will end up being.

 

[fastandfurious6]

how do both companies know in advance the average scores of each other so they are so similar?

on cutting edge CPU archs? the most complex devices on planet earth (and manufacturing lithography ASML etc etc)

some insane espionage going on

 

[Thibsie]

No

Not funny. Find something longer with 40% in it. 😁

 

[Josh128]

how do both companies know in advance the average scores of each other so they are so similar?

on cutting edge CPU archs? the most complex devices on planet earth (and manufacturing lithography ASML etc etc)

some insane espionage going on

I thought the same thing when Zen 4 hit 5.7GHz then Raptor Lake came two months later and hit 5.8GHz, lol. Its...quite strange.

 

[Abwx]

Based on what? Definitely not the Geekbench scores released thus far. Looks like Zen 5 with better MT...and thats probably exactly what it will end up being.

13900K was better than the 7950X in GB MT as well, here how it ended once you threw apps in the pipelines, and this average comprise both CB R20 and 23, wich is to say to count R23 twice, as well as the AMD crippling Povray, yet it wasnt enough even with the 253W brake removed :

Core i9-13900K, i7-13700K & i5-13600K: Gaming-Könige im Test: Benchmarks in Apps

Intel Raptor Lake im Test: Benchmarks in Apps / So wurde getestet / Leistung in Single-Core-Lasten (klassisch)

www.computerbase.de

 

[Markfw]

13900K was better than the 7950X in GB MT as well, here how it ended once you threw apps in the pipelines, and this average comprise both CB R20 and 23, wich is to say to count R23 twice, as well as the AMD crippling Povray, yet it wasnt enough even with the 253W brake removed :

Core i9-13900K, i7-13700K & i5-13600K: Gaming-Könige im Test: Benchmarks in Apps

Intel Raptor Lake im Test: Benchmarks in Apps / So wurde getestet / Leistung in Single-Core-Lasten (klassisch)

www.computerbase.de

First, this is a Zen 5 thread and you don't compare it to Zen 5. Second, if Zen 5 is faster than its totally irrelevant.

 

[Abwx]

First, this is a Zen 5 thread and you don't compare it to Zen 5. Second, if Zen 5 is faster than its totally irrelevant.

This was an answer to the assumption that ARL being better in GB MT than the 9950X would imply that it would be better in MT apps as well, so this comparison make 100% sense, and even more with the core count scaling limited GB 6 than with GB5.

 

[itsmydamnation]

Nope, the official Zen 5 SOG confirmed the decoder works as intended - it is there to boost SMT.

no it doesn't , it details the way it works.
Which when you look at there own description you can tell its foobar and they are running it the best way they can. but i wouldnt expect it to get fixed in any Zen5 core revision etc.

 

[Markfw]

This was an answer to the assumption that ARL being better in GB MT than the 9950X would imply that it would be better in MT apps as well, so this comparison make 100% sense, and even more with the core count scaling limited GB 6 than with GB5.

which is totally wrong since Zen 5's single core is much better than Zen 4's. Post a real benchmark or stay out of the Zen 5 thread with irrelevant posts.

 

[poke01]

Post a real benchmark

Gee that's gonna be hard for a lot of people

 

[Abwx]

which is totally wrong since Zen 5's single core is much better than Zen 4's. Post a real benchmark or stay out of the Zen 5 thread with irrelevant posts.

I m talking of MT since the post i was referencing to mentioned ARL s GB MT scores as being a prove of likely better MT scores than Zen 5 in apps, dunno why you are talking of ST, here is the post i was answering to :

Page 791 - Discussion - Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)

Page 791 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

 

[Markfw]

I m talking of MT since the post i was referencing to mentioned ARL s GB MT scores as being a prove of likely better MT scores than Zen 5 in apps, dunno why you are talking of ST, here is the post i was answering to :

Page 791 - Discussion - Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)

Page 791 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

But Zen5 has great avx-512 support and that could make it a lot faster. Again, without a real benchmark and the windows patch, by the time we see ARL it could blow the doors off of it.

 

[Abwx]

But Zen5 has great avx-512 support and that could make it a lot faster. Again, without a real benchmark and the windows patch, by the time we see ARL it could blow the doors off of it.

Set apart y-Cruncher i dont know anything that use AVX512 when it comes to the usual apps and benches, so it s unlikely to make any difference, FTR Cinebench 2024 require AVX256 support at most, so that s the highest AVX that is of widespread use currently.

The advantage of Zen 5 could lie in a better SMT yield, but so far no one made tests about this possibility and yet that s very easy to check, wich tell me that most reviewers are not really interested by the underlying tech but by clicks on their pages and Youtube channels, last time i saw such a test was at Hardware.fr back in 2017.

 

[RnR_au]

Gee that's gonna be hard for a lot of people

eekBench isn't out yet...

 

[deasd]

ASUS System Product Name vs ASUS System Product Name - Geekbench

WTF, is this the year of unbalanced uarchitectures or what???

Does ARL look like an INT heavy uarch from these scores?

More important question, how do you read it INT or FP heavy since GB6 doesn't distinguish INT and FP?

 

[igor_kavinski]

More important question, how do you read it INT or FP heavy since GB6 doesn't distinguish INT and FP?

GeekBench | Virtual Client Platform

GeekBench5 and GeekBench6 are third party tools that runs its own, pre-defined set of workloads to measure CPU performance. GeekBench is often used to compare

microsoft.github.io

GeekBench6 reports four different types of scores, a value that's calculated by comparing the device's performance against a baseline.

Subsection score
The workloads are grouped by how the workload exercise the system, and are calculated using the individual workload scores. The subsections are Integer,and Floating-Point workloads.

Maybe the subsection score is available in paid version or while logged in?

 

[CouncilorIrissa]

GeekBench | Virtual Client Platform

GeekBench5 and GeekBench6 are third party tools that runs its own, pre-defined set of workloads to measure CPU performance. GeekBench is often used to compare

microsoft.github.io

Maybe the subsection score is available in paid version or while logged in?

Page 11 - Question - Incredible Apple M4 benchmarks...

Page 11 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

Here's the split between INT and FP workloads present in GB6 suite. You'll have to calculate the geomean scores yourself though.

 

[jdubs03]

Already seeing price drops. Look at that 9900X. 10% discount lol.

AMD Ryzen 9 9950X drops to $623, 9900X now $449 in the US, with prices slashed up to 12% in Germany - VideoCardz.com

Ryzen 9000 gets first price cuts Not official AMD price cuts, but retailer price cuts. The AMD Ryzen 9000 series has arrived, with at least four SKUs already announced and launched. However, a few things need to be finalized before the full transition to the new platform can be declared...

videocardz.com

 

[Josh128]

Already seeing price drops. Look at that 9900X. 10% discount lol.

AMD Ryzen 9 9950X drops to $623, 9900X now $449 in the US, with prices slashed up to 12% in Germany - VideoCardz.com

Ryzen 9000 gets first price cuts Not official AMD price cuts, but retailer price cuts. The AMD Ryzen 9000 series has arrived, with at least four SKUs already announced and launched. However, a few things need to be finalized before the full transition to the new platform can be declared...

videocardz.com

Needs more.

 

[JustViewing]

Already seeing price drops. Look at that 9900X. 10% discount lol.

AMD Ryzen 9 9950X drops to $623, 9900X now $449 in the US, with prices slashed up to 12% in Germany - VideoCardz.com

Ryzen 9000 gets first price cuts Not official AMD price cuts, but retailer price cuts. The AMD Ryzen 9000 series has arrived, with at least four SKUs already announced and launched. However, a few things need to be finalized before the full transition to the new platform can be declared...

videocardz.com

All the negative talk, a win for the consumer.