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

[DisEnchantment]

Speculate at will

 

[FlanK3r]

@Det0x

https://twitter.com/x/status/1816336252846891466

hello Domdtxdissar This guy is on Overclock.net forum and also here with Det0x nick ,-)
and he is using watercooling setup. So no LN2 or DICE. This means 5950 MHz is max boost for some of cores.
His watercooling for 7950X


+ another score propably of him https://browser.geekbench.com/v6/cpu/7045730

 

[gaav87]

If anyone is interested here are some answers from the 9900x italian guy. Full answer on YT. The tests were done with 2:1 but he managed to boot into windows with 3600uclk 2400IF.
Is 3600 uclk real with proper tunning ? (i dont own zen4 does it even turn on with 3600uclk and 2400IF ?)

 

[gaav87]

@Hail The Brain Slug Quick question does zen4 even boots to windows with 2400IF 3600uclk ? I talked with the italian guy and he said he run 3600/3600/2400 but it was not stable in stress tests.

 

[Philste]

Some Bits from Computerbase:

-nobody had review samples, so nobody can tell if the SKUs are really taken back and switched out
-AMD knew about a problem for weeks, so they put 31th as release and tried fixing it until then.
-AMD specifically told the press on AMD Tech day over 2 weeks ago that supply of Ryzen 9 SKUs will be (extremely) tight.
-with these info, Computerbase Editor definitely thinks what AMD said to the Verge isn't the truth.
-However, he doesn't know the real reason

Overall it makes me think my Yield Theory gets increasingly more likely. There must be something different for the big SKUs and the only real difference is higher boost clocks.

 

[exquisitechar]

Some Bits from Computerbase:

-nobody had review samples, so nobody can tell if the SKUs are really taken back and switched out
-AMD knew about a problem for weeks, so they put 31th as release and tried fixing it until then.
-AMD specifically told the press on AMD Tech day over 2 weeks ago that supply of Ryzen 9 SKUs will be (extremely) tight.
-with these info, Computerbase Editor definitely thinks what AMD said to the Verge isn't the truth.
-However, he doesn't know the real reason

Overall it makes me think my Yield Theory gets increasingly more likely. There must be something different for the big SKUs and the only real difference is higher boost clocks.

Source is forum posts or what?

 

[CakeMonster]

HUB talked about this a day or two before the news broke, Steve was waiting for a package with the CPU's but there was radio silence about when it would ship or arrive, and they were quite overdue compared to earlier launches that had a fixed date.

 

[MS_AT]

I have read rumours about solder quality between dies and IHS being the problem

 

[MS_AT]

Some Bits from Computerbase:

-nobody had review samples, so nobody can tell if the SKUs are really taken back and switched out
-AMD knew about a problem for weeks, so they put 31th as release and tried fixing it until then.
-AMD specifically told the press on AMD Tech day over 2 weeks ago that supply of Ryzen 9 SKUs will be (extremely) tight.
-with these info, Computerbase Editor definitely thinks what AMD said to the Verge isn't the truth.
-However, he doesn't know the real reason

Overall it makes me think my Yield Theory gets increasingly more likely. There must be something different for the big SKUs and the only real difference is higher boost clocks.

Tight for 9950X only or both 9900X and 9950X?

 

[CakeMonster]

I've read several places that 9950X supply was expected to be less than the rest.

 

[Det0x]

The only thing i will say is that i dont use LN2 nor dry ice, runs were done with my large loop watercooling.
GB runs were done with PBO CO + baseclock (not allcore static OC)

 

[cebri1]

So paper lunch? Why not delay it another 15-30 days? After all the PR damage is already done after the recall , and if then there are not nearly enough units available for a launch… what’s the point?

 

[FlameTail]

Strix Point is on schedule right?

 

[stayfrosty]

Strix Point is on schedule right?

yup, you can actually buy them in many countries now. for example here in switzerland the following shop has many asus models in stock since yesterday:

https://www.brack.ch/asus-vivobook-s-15-oled-m5506wa-ma012w-1682611

(this specific model has 26 in stock right now and 104 more arriving in 10 days)

 

[Josh128]

The only thing i will say is that i dont use LN2 nor dry ice, runs were done with my large loop watercooling.
GB runs were done with PBO CO + baseclock (not allcore static OC)

Lol, the real burning question is, how'd you get your grubby mitts on one?

 

[StefanR5R]

Some Bits from Computerbase:

-nobody had review samples, so nobody can tell if the SKUs are really taken back and switched out
-AMD knew about a problem for weeks, so they put 31th as release and tried fixing it until then.
-AMD specifically told the press on AMD Tech day over 2 weeks ago that supply of Ryzen 9 SKUs will be (extremely) tight.
-with these info, Computerbase Editor definitely thinks what AMD said to the Verge isn't the truth.
-However, he doesn't know the real reason

Overall it makes me think my Yield Theory gets increasingly more likely. There must be something different for the big SKUs and the only real difference is higher boost clocks.

Source is forum posts or what?

Yes, the source is a forum post.
https://www.computerbase.de/forum/t...ugust-verschoben.2204061/page-8#post-29618706
(post #149 by Volker, in German)

Edit, plus
https://www.computerbase.de/forum/t...ugust-verschoben.2204061/page-9#post-29618818
(post #169 by Volker)

Edit 2,
https://www.computerbase.de/forum/t...gust-verschoben.2204061/page-11#post-29619173
(post #218 by Volker)
gustlegga: "When did you receive notice to return the review kit?"
Volker: "We did not have a review kit! Just like everybody else did not have any."

 

[leoneazzurro]

The fact the delay is only two weeks and affects desktop parts only makes me believe there is no hardware problem. Especially with the heat spreader that is one of the easiest things to keep under check (and it is a proven process, especially on desktop parts).
My guess is that either.

- AMD wants to get the BIOSes for older mainboards as polished as they can before the launch. Especially when Intel is releasing a microcode update that might reduce a little their performance, ths making the 9xxx series look better.
- There was a problem with the production testing equipment or testing conditions for the desktop parts that was detected late, that could have caused some unstable CPUs to be passed as good ones. Testing is not an easy task, especially on complex and power hungry devices as CPUs are nowadays.

 

[gaav87]

The only thing i will say is that i dont use LN2 nor dry ice, runs were done with my large loop watercooling.
GB runs were done with PBO CO + baseclock (not allcore static OC)

Is 3600uclk and 2400fclk possible or is the italian guy delusional ?

 

[Det0x]

Review embargo dates are out, 1 day before release

9600X + 9700X @ august 7
9900X + 9950X @ august 14


Taken from this HUB video

 

[CakeMonster]

You'd think they'd want the higher end parts to get media coverage first. But they may not have a choice in this situation.

 

[igor_kavinski]

+ another score propably of him https://browser.geekbench.com/v6/cpu/7045730

Doubt it. Why would Det0x use older Geekbench? And why balanced power plan?

 

[igor_kavinski]

Lol, the real burning question is, how'd you get your grubby mitts on one?

Bribed someone with BIG watermelons

 

[Hans de Vries]

5.95 GHz 9950X in Geekbench 3, 4, 5 and 6:

domdtxdissar's Profile - Geekbench

browser.geekbench.com

 

[poke01]

5.95 GHz 9950X in Geekbench 3, 4, 5 and 6:

domdtxdissar's Profile - Geekbench

browser.geekbench.com

5.95GHz stock would have been possible with N3E thanks to FinFlex. Sucks that desktop won’t get 3nm anytime soon.

 

[Josh128]

5.95GHz stock would have been possible with N3E thanks to FinFlex. Sucks that desktop won’t get 3nm anytime soon.

Maybe, maybe not. Surface area, power density, and core architecture have a lot to do with it as well. Considering Intels current plight, I think all companies are now very wary of chasing frequencies >2.5x that of microwave ovens.

 

[Josh128]

5.95 GHz 9950X in Geekbench 3, 4, 5 and 6:

domdtxdissar's Profile - Geekbench

browser.geekbench.com

That GB4 score is impressive to me. GB4 is the one I cut my Ryzen benchmarking teeth on. I remember being impressed when 2700X was just able to crack 5000 ST (still ~15% behind 8700K at the time).