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

[DisEnchantment]

Speculate at will

 

[Goop_reformed]

So is zen 5 good yet?

 

[igor_kavinski]

So is zen 5 good yet?

It will be. In Turin

 

[Goop_reformed]

It will be. In Turin

I don't care about Turin, just want the x3d part to be extremely good.

 

[Markfw]

I don't care about Turin, just want the x3d part to be extremely good.

The Turin is the basis for all DT parts, including the x3d parts. You want Turin to be good, just less power, and it will be.

 

[DaaQ]

• X870E is a renamed X670E with mandatory USB4 capabilities, which costs 4 lanes.
• X870 is a renamed B650E (just one Promontory chip), so just 8 additional lanes from the start, and it also has mandatory USB4

So if you were going to build on AM5 with Z5, would you go for the 670E or see what 870E brings from the Mobo makers?

Also regarding RAM?

 

[Josh128]

These guys are just f'n with us now. New 9700X run. 3395 , 16909. This is a decent showcase, but who knows what trickery was used to achieve it. I hate Geekbench. +16.5% ST, +10.7% MT vs official 7700X scores in the browser.

Micro-Star International Co., Ltd. MS-7D75 - Geekbench

Benchmark results for a Micro-Star International Co., Ltd. MS-7D75 with an AMD Ryzen 7 9700X processor.

browser.geekbench.com

 

[Thunder 57]

These guys are just f'n with us now. New 9700X run. 3395 , 16909. This is a decent showcase, but who knows what trickery was used to achieve it. I hate Geekbench. +16.5% ST, +10.7% MT vs official 7700X scores in the browser.

Micro-Star International Co., Ltd. MS-7D75 - Geekbench

Benchmark results for a Micro-Star International Co., Ltd. MS-7D75 with an AMD Ryzen 7 9700X processor.

browser.geekbench.com

To be fair, if I had the hardware and was not under NDA, I wouldn't mind screwing with people just to see the reaction. Maybe that makes me a psychopath?

 

[Goop_reformed]

These guys are just f'n with us now. New 9700X run. 3395 , 16909. This is a decent showcase, but who knows what trickery was used to achieve it. I hate Geekbench. +16.5% ST, +10.7% MT vs official 7700X scores in the browser.

Micro-Star International Co., Ltd. MS-7D75 - Geekbench

Benchmark results for a Micro-Star International Co., Ltd. MS-7D75 with an AMD Ryzen 7 9700X processor.

browser.geekbench.com

That is just probably high OC

 

[Hitman928]

That is just probably high OC

It's at 5.55 GHz when it's spec'd at 5.5 GHz. Zen 4 could run at up to 150 MHz above it's max spec'd boost clock so it's probably not overclocked, but is a really good sample that's able to operate (slightly) above listed max boost.

 

[StinkyPinky]

Asus have updated their website with another laptop

This time a 14" model

ASUS Vivobook S 14 OLED (M5406)｜Laptops For Home｜ASUS USA

Colors that match your vibe. Bold and youthful ASUS Vivobook laptops are designed to represent who you are. With top-notch performance, the innovative ErgoLift hinge, and NanoEdge display, ASUS Vivobooks help you get the job done.

www.asus.com

 

[Hail The Brain Slug]

Blender Benchmark 3.3.0 runs: Fully stock 7950X, 6000C30 EXPO (no high bandwidth, no low latency modes).

When comparing these scores to the 7950X3D runs here, I think we need to bear in mind the 7950X was very sloppy with the V/F curve, taking a hammer to it to get max boost clocks at whatever the power cost. The 7950X3D is, comparatively, extremely stringently tuned with regards to V/F curve and will manage to boost higher clocks under all circumstances until the 7950X exceeds its power draw considerably.

This is a launch day 7950X sample. Its possible they improved the V/F curve since launch in newer samples. But I think this is a fair comparison to the alleged 9950X. Much more so than an incredibly-finely-tuned-out-of-the-box 7950X3D.

60W PPT


90W PPT


120W PPT


230W PPT (Stock)

 

[Philste]

Blender Benchmark 3.3.0 runs: Fully stock 7950X, 6000C30 EXPO (no high bandwidth, no low latency modes).

So that would put ZEN5 exactly where i expected it in terms of efficiency. Something between "basically same as ZEN4" and "same clock at same power -> ~16% faster at same power".

 

[Hail The Brain Slug]

So that would put ZEN5 exactly where i expected it in terms of efficiency. Something between "basically same as ZEN4" and "same clock at same power -> ~16% faster at same power".

In the screenshots of the 9950X, it's clearly using a number more watts for uncore, giving the cores fewer watts than in my case. Might be because it's an ES, might be because of whatever weird tuning has been done with it.

If you compare my performance at the core power shown vs the 9950X performance at the core power, it makes the 9950X look a bit better

 

[Det0x]

Blender Benchmark 3.3.0 runs: Fully stock 7950X, 6000C30 EXPO (no high bandwidth, no low latency modes).

When comparing these scores to the 7950X3D runs here, I think we need to bear in mind the 7950X was very sloppy with the V/F curve, taking a hammer to it to get max boost clocks at whatever the power cost. The 7950X3D is, comparatively, extremely stringently tuned with regards to V/F curve and will manage to boost higher clocks under all circumstances until the 7950X exceeds its power draw considerably.

This is a launch day 7950X sample. Its possible they improved the V/F curve since launch in newer samples. But I think this is a fair comparison to the alleged 9950X. Much more so than an incredibly-finely-tuned-out-of-the-box 7950X3D.

60W PPT
View attachment 102787

90W PPT
View attachment 102788

120W PPT
View attachment 102789

230W PPT (Stock)
View attachment 102790

Thanks for comparison 👍

Your 230w numbers also pretty much matches the 230w 7950X here: (taken for WCCFTech)

But sadly its still alittle unclear to me if the 7950X3D are benefiting from the v-cache on CCD0 in blender, or if its all down to voltage tuning for that SKU, since it seems to outperform the 7950X in the powerlimited scenario we're looking at in this thread

*edit*

One more picture from the commet section over there to make it clearer: (@Hail The Brain Slug results used)


It seems like the junkshop sub-bench is standing out with the largest delta, what makes it different compared to the other two ? 🤔

 

[Det0x]

In other news: canada-pc only ordered 50 six cores from AMD

 

[Gideon]

While this is more appropriate to put in the Zen 4 thread, i'll post it here, as I made the initial post here, about Strix Halo comparison.

The mentioned Minisforum mini-PC "AtomMan G7" launched that has the 16 core Zen4 CPU and Radeon 7600M XT, which should give a pretty good idea of Strix Halos gaming preformance:

G7 itself looks great. Is pretty slim and quiet. This in turn seems to indicate that Halo will be decent mobile workstation CPU. something in between M3 Pro and M3 Max (at least on the HW side)

 

[CouncilorIrissa]

These guys are just f'n with us now. New 9700X run. 3395 , 16909. This is a decent showcase, but who knows what trickery was used to achieve it. I hate Geekbench. +16.5% ST, +10.7% MT vs official 7700X scores in the browser.

Micro-Star International Co., Ltd. MS-7D75 - Geekbench

Benchmark results for a Micro-Star International Co., Ltd. MS-7D75 with an AMD Ryzen 7 9700X processor.

browser.geekbench.com

It's the first run with High performance power plan. Might explain the higher score.

 

[tsamolotoff]

High performance power plan. Might explain the higher score.

Probably just variance, you can go a few pages back where I've posted several runs of GB6 with 200 pts of difference between the best and worst for ST score. And 6% of difference is enough to claim crushing victory over your foe on wccf and vcn comments section, mind you!

 

[CouncilorIrissa]

Probably just variance, you can go a few pages back where I've posted several runs of GB6 with 200 pts of difference between the best and worst for ST score. And 6% of difference is enough to claim crushing victory over your foe on wccf and vcn comments section, mind you!

Variance is an annoying part of GB6, yes.

 

[Abwx]

It's the first run with High performance power plan. Might explain the higher score.

Nothing to do with high performance power plan as such, that s just a MSI set up, previous GB scores where with an Asus board that could well be using the same settings and TDP.

 

[igor_kavinski]

In other news: canada-pc only ordered 50 six cores from AMD

Since it isn't expected to sell well, please please tell me who do I contact at AMD to give me a free 9600X?

 

[igor_kavinski]

60W: http://www.portvapes.co.uk/?id=Latest-exam-1Z0-876-Dumps&exid=thread...ranite-ridge-ryzen-9000.2607350/post-41246800
90W: http://www.portvapes.co.uk/?id=Latest-exam-1Z0-876-Dumps&exid=thread...ranite-ridge-ryzen-9000.2607350/post-41248191
120W: http://www.portvapes.co.uk/?id=Latest-exam-1Z0-876-Dumps&exid=thread...ranite-ridge-ryzen-9000.2607350/post-41248591

 

[Hail The Brain Slug]

View attachment 102817

60W: http://www.portvapes.co.uk/?id=Latest-exam-1Z0-876-Dumps&exid=thread...ranite-ridge-ryzen-9000.2607350/post-41246800
90W: http://www.portvapes.co.uk/?id=Latest-exam-1Z0-876-Dumps&exid=thread...ranite-ridge-ryzen-9000.2607350/post-41248191
120W: http://www.portvapes.co.uk/?id=Latest-exam-1Z0-876-Dumps&exid=thread...ranite-ridge-ryzen-9000.2607350/post-41248591

View attachment 102819

FFS how many times do I have to run this benchmark!

 

[igor_kavinski]

FFS how many times do I have to run this benchmark!

Speculation: AMD has implemented n PPT levels with max n value depending on how crazy the mobo power circuitry is

Contact your nearest Russian/Chinese/North Korean dealer for portable nuclear reactor

 

[SmokSmog]

60W: http://www.portvapes.co.uk/?id=Latest-exam-1Z0-876-Dumps&exid=thread...ranite-ridge-ryzen-9000.2607350/post-41246800
90W: http://www.portvapes.co.uk/?id=Latest-exam-1Z0-876-Dumps&exid=thread...ranite-ridge-ryzen-9000.2607350/post-41248191
120W: http://www.portvapes.co.uk/?id=Latest-exam-1Z0-876-Dumps&exid=thread...ranite-ridge-ryzen-9000.2607350/post-41248591

230W PPT when?