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

[DisEnchantment]

Speculate at will

 

[igor_kavinski]

geekbench. Need I say more.

True. I suspect 9600X will be much better with PBO and Curve Shaper due to having lower temps than 7600X. Stock performance difference will not be enough to warrant an upgrade, unless the user is running something FP-intensive.

 

[fastandfurious6]

How is it possible all internals got beefed up even doubled but its only 10%-20% at most

everyone's in disbelief!

 

[CouncilorIrissa]

9600X vs. 7600X on the same mobo: https://browser.geekbench.com/v6/cpu/compare/7182575?baseline=7181563

Looks like nothing to write home about

4800MT/s for the 9600X vs 7600MT/s for the 7600X and GB6 is very membw sensitive.

I expect a 13-14% score difference using the same memory.

 

[igor_kavinski]

This shows the 9700X in a much better light over the 7800X3D: https://browser.geekbench.com/v6/cpu/compare/7158172?baseline=7196206

 

[FlameTail]

How is it possible all internals got beefed up even doubled but its only 10%-20% at most

everyone's in disbelief!

We are in the era of diminishing gains.

In 2018 when ARM unveiled the Cortex A76 with a colossal >50% IPC gain, they prophesied that going forward there will be only smaller gains. It has proved true. We haven't seen such a large IPC gain from them since.

 

[Timmah!]

How is it possible all internals got beefed up even doubled but its only 10%-20% at most

everyone's in disbelief!

Clearly not all of them were doubled. They forgot to double exactly the ones that would fetch 2x speed-up :-D

 

[CouncilorIrissa]

Huang got his STX Zenbook to boost properly lmao.

Linux providing 2/3 worth of a gen-on-gen improvement will never be not funny.

Ryzen AI 9 HX 370 P-cores (28W, 4C8T) - Geekbench

Benchmark results for an ASUSTeK COMPUTER INC. ASUS Zenbook S 16 UM5606WA_UM5606WA with an AMD Ryzen AI 9 HX 370 processor.

browser.geekbench.com

 

[Josh128]

Was about to post. Theres the fabled 3K on Strix. Linux FTW!! Did Huang ever post part 2 of his review? He references it in part 1, but Ive yet to see it.

 

[ErichZannMusician]

4800MT/s for the 9600X vs 7600MT/s for the 7600X and GB6 is very membw sensitive.

I expect a 13-14% score difference using the same memory.

Makes absolutely no difference to the Single Core score:

ASRock X670E Taichi Carrara - Geekbench

Benchmark results for an ASRock X670E Taichi Carrara with an AMD Ryzen 5 9600X processor.

browser.geekbench.com

 

[gdansk]

Huang got his STX Zenbook to boost properly lmao.

Linux providing 2/3 worth of a gen-on-gen improvement will never be not funny.

Ryzen AI 9 HX 370 P-cores (28W, 4C8T) - Geekbench

Benchmark results for an ASUSTeK COMPUTER INC. ASUS Zenbook S 16 UM5606WA_UM5606WA with an AMD Ryzen AI 9 HX 370 processor.

browser.geekbench.com

That's better. About 1.15x my 7840U in Linux.

 

[Josh128]

Makes absolutely no difference to the Single Core score:

ASRock X670E Taichi Carrara - Geekbench

Benchmark results for an ASRock X670E Taichi Carrara with an AMD Ryzen 5 9600X processor.

browser.geekbench.com

That is absolutely not true.

 

[ErichZannMusician]

That is absolutely not true.

Literally proved in the link I posted, which is with 6400 MT/s memory.

 

[Philste]

This shows the 9700X in a much better light over the 7800X3D:

5.0 vs 5.5GHz for those SKUs, so clock normalised we also end at like 9%

 

[Josh128]

Literally proved in the link I posted, which is with 6400 MT/s memory.

Latency isnt shown and is as important as clock. How else can you explain the massive descrepancies in ST and MT scores reporting same CPU clocks for same SKUs?

 

[ErichZannMusician]

Latency isnt shown and is as important as clock. How else can you explain the massive descrepancies in ST and MT scores reporting same CPU clocks for same SKUs?

MT score of GB6 is irrelevant.

 

[Hail The Brain Slug]

MT score of GB6 is irrelevant.

Maybe overall, especially when comparing two different products. But comparing the exact same product against itself can provide some value. The core configuration and uarch is unchanged so differences in score can be attributed to more specific differences like boost clocks and memory.

 

[ErichZannMusician]

Maybe overall, especially when comparing two different products. But comparing the exact same product against itself can provide some value. The core configuration and uarch is unchanged so differences in score can be attributed to more specific differences like boost clocks and memory.

The 'consensus' - if you will - in this forum is that GB6 multi-core score is useless as representative of your typical multi-core workload.

 

[igor_kavinski]

AMD Ryzen 9 7900X Zen 4 AM5 12Cores/24Threads Processor | 100-100000589WOF in UAE | Oman Online

AMD Ryzen 9 7900X Zen 4 AM5 12Cores/24Threads Processor | 100-100000589WOF in uae | oman for 1685 Play now pay later next day delivery for Dubai, Al Ain, Muscat

gccgamers.com

AMD Ryzen 9 9900X 5.6 GHz 12-Core AM5 - CPU | 100-100000662WOF in UAE | Oman Online

AMD Ryzen 9 9900X 5.6 GHz 12-Core AM5 - CPU | 100-100000662WOF in uae | oman for 2050 Play now pay later next day delivery for Dubai, Al Ain, Muscat

gccgamers.com

7900X priced at $458 while 9900X at $557.

Should get delivered in max 4 days, if I pull the trigger, which I won't

 

[Hail The Brain Slug]

The 'consensus' - if you will - in this forum is that GB6 multi-core score is useless as representative of your typical multi-core workload.

I can see nuance is lost on you.

 

[Josh128]

4800MT/s for the 9600X vs 7600MT/s for the 7600X and GB6 is very membw sensitive.

I expect a 13-14% score difference using the same memory.

For some unknown reason, a quite good result for 9600X posted after the above run, different system. This one appears to be hitting and maintaining full ST boost at least, but dont know if this is PBO or what.

Gigabyte Technology Co., Ltd. X670E AORUS XTREME - Geekbench

Benchmark results for a Gigabyte Technology Co., Ltd. X670E AORUS XTREME with an AMD Ryzen 5 9600X processor.

browser.geekbench.com

 

[CouncilorIrissa]

Makes absolutely no difference to the Single Core score:

ASRock X670E Taichi Carrara - Geekbench

Benchmark results for an ASRock X670E Taichi Carrara with an AMD Ryzen 5 9600X processor.

browser.geekbench.com

At 6400 MT/s you might be running FCLK:MCLK at a ratio of 1:2.
In general best GB6 runs are either running DDR5-6000 (likely 1:1 ratio), or DDR5-8000. The reported clocks are very similar to your run.

ST score isn't as memory bandwidth sensitive, sure, but it still matters, so I wouldn't compare two CPUs with different memory setups on something as variable as gb6.

Anyway, only a few days left to wait to see how DOA GNR is.

 

[ErichZannMusician]

I can see nuance is lost on you.

Nah, I find it funny that a forum whose members overwhelmingly decided that GB6 MT is useless suddenly splitting hairs about perceived differences between submissions because reality is turning out to be quite different from the hype they have been feeding on for months.

 

[Hail The Brain Slug]

Nah, I find it funny that a forum whose members overwhelmingly decided that GB6 MT is useless suddenly splitting hairs about perceived differences between submissions because reality is turning out to be quite different from the hype they have been feeding on for months.

Again, nuance is lost on you.

 

[Nothingness]

The 'consensus' - if you will - in this forum is that GB6 multi-core score is useless as representative of your typical multi-core workload.

The reality is always more complex, no matter what the consensus is. This has been discussed ad nauseam. And this proves that no matter what technical arguments are given, someone will come and make the same approximative claim again and again.

If you want to see perfect scaling, pick GB6 ray tracer score, it scales about as well as Cinebench.

 

[CouncilorIrissa]

The reality is always more complex, no matter what the consensus is. This has been discussed ad nauseam. And this proves that no matter what technical arguments are given, someone will come and make the same approximative claim again and again.

If you want to see perfect scaling, pick GB6 ray tracer score, it scales about as well as Cinebench.

And then people wonder why more and more knowledgeable people are leaving the forum.
How is "do not treat different memory configs as an apples-to-apples comparison" a controversial statement.

edit: and by "knowledgeable" I don't mean myself -- I'm far from it.