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

[DisEnchantment]

Speculate at will

 

[CouncilorIrissa]

In b4 AMD is sandbagging hard by sending out old B0 silicon to reviewers that is buggy/clocks worse than the retail samples on a newer stepping.

What would be the point?

 

[Hail The Brain Slug]

What would be the point?

I'm just trying to explain the existence of this ancient sample. The crazier the speculation, the better!

I have as of yet not participated in wild speculation, so I figured I should contribute.

Actual retail silicon will be Zen50% confirmed!

 

[StefanR5R]

I'm more inclined to believe a misprint or mislabel of some sort rather than AMD holding on to inventory for.... 5 quarters?

Or it was done for leak detection.
Or it's not a photograph but made up.

I'm just trying to explain the existence of this ancient sample. The crazier the speculation, the better!

I agree with the last part here. :-)

 

[Hail The Brain Slug]

Or it's not a photograph but made up.

Actually I am inclined to believe that. If you look at the spacing between the '99' of the '9950X' they look closer together than AMD's font width. Too close together. Almost touching.

 

[igor_kavinski]

I'm still confident that Zen 5 delivers 62% IPC under certain conditions such as:

Near absolute zero temperature.
Close to the Earth's core.
In deep space.
Just before getting sucked into a blackhole.
When touched by Thor.
On the dark side of the moon.
On a sunny day in Siberia.
When no one's looking.
When threatened by Darth Vader.
When all the planets of the solar system align.
When a gravitational wave runs through it.

 

[Hail The Brain Slug]

I'm still confident that Zen 5 delivers 62% IPC under certain conditions such as:

Near absolute zero temperature.
Close to the Earth's core.
In deep space.
Just before getting sucked into a blackhole.
When touched by Thor.
On the dark side of the moon.
On a sunny day in Siberia.
When no one's looking.
When threatened by Darth Vader.
When all the planets of the solar system align.
When a gravitational wave runs through it.

Prompt: Rewrite the following text as a poem.

Result:

In the chill of near absolute zero, Or the heat where Earth’s core does glow, In the vastness of deep space, Or near a black hole’s embrace.

When touched by Thor’s mighty hand, Or on the moon’s dark, silent land, On a sunny day in Siberia’s cold, Or when secrets are left untold.

When Darth Vader’s threat is near, Or when planets align, crystal clear, When a gravitational wave flows through, Zen 5’s power shines anew.

 

[igor_kavinski]

Wouldnt it be something if at some point they surprise with a 16 core unified L3 X3D 3nm chip? Depending on clocks, seems like it could be a potent SKU.

Even with 4.5 GHz clocks, it would still be a pretty decent chip. No more inter-CCD latency!

And if they make it the most powerful APU (even just a little faster than 8700G in graphics horsepower but with 32 CPU threads), it would sell really well.

 

[Nothingness]

Very strange there would even exist a retail etched sample from so long ago. It doesn't seem like it's necessarily one of the review samples, that article is kind of all over the place saying bitwit kyle "found it on a plane". The photos of the actual review samples have the details blurred on the IHS.

Yes, I wonder if the picture used on the site is an old one randomly picked that just shows an early ES.

 

[gaav87]

If the performance uplift falls off at 160W+ im still coping for BIOS uplift

 

[StefanR5R]

I wonder if the picture used on the site is an old one randomly picked that just shows an early ES.

This article shows a CPU with the very same etchings, and same PCB layout:
TechPowerUp on June 4, Hands On with the AMD Ryzen 9 9950X "Zen 5" Desktop Processor
Edit: even with the same scratch mark on the left side of the heat spreader.

At its Computex 2024 booth, AMD showed us their latest flagship desktop processor, the AMD Ryzen 9 9950X. [...]

 

[Hail The Brain Slug]

This article shows a CPU with the very same etchings, and same PCB layout:
TechPowerUp on June 4, Hands On with the AMD Ryzen 9 9950X "Zen 5" Desktop Processor
Edit: even with the same scratch mark on the left side of the heat spreader.

So it seems like it was likely MFG specifically as a showpiece. That leads me to believe no conclusions can be drawn from the dates. They probably made it a long time ago out of a really early ES sample to show privately and it's just been around for a while.

 

[Josh128]

If the performance uplift falls off at 160W+ im still coping for BIOS uplift

Interesting. You have a link to that article??

 

[lightmanek]

The earlier they get their work done, the quicker they can slap Intel around

Remember, rear view mirrors are hard to deal with, work smart, not hard.

 

[Abwx]

View attachment 102682

Looks like clock really tanks with low limit on Zen5, i've had like 4 ghz on my system with 90W/90W/90W limits

It tanks because the guy increased exageratly the IOD TDP, this way all numbers are truncated and shifted to an apparent lower efficency, Computerbase measured the IOD TDP at 15W with stock RAM and about 20W with 6000 RAM, far from the 35W we can see here.

At 60W PPT he s actually at 45W PPT, his 90W is actually 70W PPT and his 120W is 100W PPT, that s the real numbers, and still, that would be with DDR 6000, at stock RAM speed the actual figures are 40W/65W/95W.

 

[gaav87]

Interesting. You have a link to that article??

AMD Ryzen 3000 Post-Review BIOS Update Recap: Larger ST Gains, Some Gains, Some Losses - article name​

 

[Hail The Brain Slug]

It tanks because the guy increased exageratly the IOD TDP, this way all numbers are truncated and shifted to an apparent lower efficency, Computerbase measured the IOD TDP at 15W with stock RAM and about 20W with 6000 RAM, far from the 35W we can see here.

At 60W PPT he s actually at 45W PPT, his 90W is actually 70W PPT and his 120W is 100W PPT, that s the real numbers, and still, that would be with DDR 6000, at stock RAM speed the actual figures are 40W/65W/95W.

at 60W PPT my 7950X only gets about 28-29W core power. Only a few watts more than he's getting. I'm completely stock with only 6000C30 EXPO enabled.

 

[Makaveli]

AMD Ryzen 3000 Post-Review BIOS Update Recap: Larger ST Gains, Some Gains, Some Losses - article name​

That's a zen 2 article.

 

[gdansk]

That's a zen 2 article.

That's in line with the conversation, they posted a Zen 2 graph.

 

[Abwx]

at 60W PPT my 7950X only gets about 28-29W core power. Only a few watts more than he's getting.

Because you overclocked the RAM with user settings, beside if you have a 7950X3D the V-cache will drain some power, at Computerbase for their 7950X review and depending of the MB they get 14-16W at stock and 19-21W when overclocking the RAM to 6000, numbers like 35W are obviously either bad power management or badly implemented settings when overclocking the RAM.

AMD Ryzen 7000 im Test: So schnell sind 7950X und 7700X: Leistungsaufnahme und Effizienz

Ryzen 9 7950X & Ryzen 7 7700X im Test: Leistungsaufnahme und Effizienz / TDP und PPT steigen deutlich

www.computerbase.de

Edit : From 90W PPT to 230W PPT the 7950X get 23% higher perf, so it should be the same for the 9950X, when starting from the 120W PPT figure of the leaker we should expect this score to increase by about 23% once the 9950X is set at stock 230W PPT.

 

[tsamolotoff]

Perhaps they want to debut Zen 5 with a really stable AGESA at launch, for once.

Very funny joke, haha

Here's some 160W results (basically runs at full clocks that are permissible with x3d limits)





Won't go any higher, sadly. Thanks Lisa et al for blocking proper OC, garrrr.

 

[yuri69]

Zen 5 IHS shows 2023 date. Disqus commenter Senpai noted the visible code seems to indicate 15th week of 2023. Thats over a year ago.

Review samples of Zen 4 were labelled with 26th week 2022. Reviews themselves happened on the 16th week of 2023 or so. This shows there usually is a notable gap.

 

[Hail The Brain Slug]

Review samples of Zen 4 were labelled with 26th week 2022. Reviews themselves happened on the 16th week of 2023 or so. This shows there usually is a notable gap.

16th week 2023?
Weren't the reviews 39th week 2022? Sept 26/27?

 

[gaav87]

That's a zen 2 article.

Guy asked about bios pre-launch and post-launch uplift from earlier screenshot i linked for 3900x launch so i provided.

 

[Rheingold]

Review samples of Zen 4 were labelled with 26th week 2022. Reviews themselves happened on the 16th week of 2023 or so.

In your linked article, there are just 3 months between the date on the chip and the review. This time, it's more than a year.

 

[Thibsie]

And for the fun : maybe zen5 DT is so-so 'cos it has an older rev with those bugs fixed in Epyc silicon. There. Did my speculation for the year lol