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

[DisEnchantment]

Speculate at will

 

[inf64]

Oh, really?
1800x -> 3700x +32%
3900x -> 5900x +19%
5900x -> 7900x +33%
7900x -> 9900x +14%

Well we don't have the official numbers yet, you are just using "leaks" as numbers. Not that long to go, just 3 weeks.

 

[DrMrLordX]

They are focusing on enterprise stuff, and nothing related to Zen 5 (yet)

AMD has been doing that since at least as far back as Zen3.

 

[Josh128]

Oh, really?
1800x -> 3700x +32%
3900x -> 5900x +19%
5900x -> 7900x +33%
7900x -> 9900x +14%

Bad math for your Zen 4 vs Zen 5 comparison. Its >17%. Also, its the only one out of them all that achieved all its gains with the combination of no clock uplift and no node advancement.

7900X PBO = 29.3K, 9900X PBO = 34.5K. Thats a difference of 17.7%, not 14%. Source: https://www.techpowerup.com/review/amd-ryzen-9-7900x/7.html

 

[poke01]

no node advancement.

Depends if its N4P then it should offer 11% performance increase over N5.

To say there is no node advancement is not true, its just N4P is the same N5 family of nodes but you do get performance increase from N4P.

TSMC Expands Advanced Technology Leadership with N4P Process|Taiwan Semiconductor Manufacturing Company Limited

Hsinchu, Taiwan, R.O.C., Oct. 26, 2021 - TSMC (TWSE: 2330, NYSE: TSM) today introduced its N4P process, a performance-focused enhancement of the 5-nanometer technology platform.

pr.tsmc.com

 

[poke01]

Did AMD use N4P for Zen 5? If AMD used N4 then yeah most of the increase is from urarch.

 

[Fjodor2001]

New AMD firmware for Zen5 released with performance optimizations.

I wonder how much that’ll improve performance. Hopefully some at least, but I don’t think it’ll make Zen5 reach the 40% IPC increase vs Zen4.

New AMD firmware brings performance optimizations for Ryzen 9000 CPUs

Gigabyte's AGESA 1.2.0.0a BIOS updates add performance optimizations for Ryzen 9000 processors.

www.tomshardware.com

 

[SarahKerrigan]

New AMD firmware for Zen5 released with performance optimizations.

I wonder how much that’ll improve performance. Hopefully some at least, but I don’t think it’ll make Zen5 reach the 40% IPC increase vs Zen4.

New AMD firmware brings performance optimizations for Ryzen 9000 CPUs

Gigabyte's AGESA 1.2.0.0a BIOS updates add performance optimizations for Ryzen 9000 processors.

www.tomshardware.com

But I thought 32%, at least, was inevitable! Tier 1 OEMs don't lie and neither do Anandtech posters!

 

[Philste]

I hope the good reviews point out the real efficiency gains of ZEN5 (which seem to be not that big) and not do this "9900X is 15% faster at 120W compared to 7900X at 170W" thing. The detailed reviews show that 7900X needs like 170-180W for it's numbers, but loses below 10% even with 88W. 9900X has 120W TDP, which equals 162W MTP, so it's definitely able to draw 162W. Suddenly the gains don't look that impressive, right?

I start getting the feeling ZEN4s horrendous stock profiles were made to make ZEN5 on N4P look better 2 years later.

 

[Abwx]

The detailed reviews show that 7900X needs like 170-180W for it's numbers, but loses below 10% even with 88W.

Detailed reviews show that at 88W it lose 12%, and in Cinebench R23 it lose even 15%, wich is certainly not below 10%.

Intel Core i9-14900K, i7-14700K & i5-14600K im Test: Anwendungs-Benchmarks

i9-14900K, i7-14700K & i5-14600K im Test: Anwendungs-Benchmarks / Testmethodik / Multi-Core-Leistung / Single-Core-Leistung

www.computerbase.de

 

[Philste]

Detailed reviews show that at 88W it lose 12%, and in Cinebench R23 it lose even 15%, wich is certainly not below 10%.

Your link has only one 7900X result in the test, so what are you talking about? In PCGH release test 88W 7900X is 8% slower than unlocked 7900X (which also only pulled like 180W).

 

[coercitiv]

New AMD firmware for Zen5 released with performance optimizations.

I wonder how much that’ll improve performance. Hopefully some at least, but I don’t think it’ll make Zen5 reach the 40% IPC increase vs Zen4.

They are porting the performance improvements from the Rocket Lake microcode and the power improvements from the Meteor Lake firmware.

Folks on this forum don't need to learn much from history, they just need to remember what happened yesterday.

 

[PJVol]

Bad math for your Zen 4 vs Zen 5 comparison. Its >17%. Also, its the only one out of them all that achieved all its gains with the combination of no clock uplift and no node advancement.

7900X PBO = 29.3K, 9900X PBO = 34.5K. Thats a difference of 17.7%, not 14%

PBO? Why not LN OC? )
Anyway, if you haven't read the article (highlighted the key words for you, just in case)

In terms of scores, the chip was first evaluated at its default 120W mode and scored around 33,000 points in the Cinebench R23 multi-threaded test. For comparison, the AMD Ryzen 9 7900X scores around 29,000 points on average at the default operation at its 170W TDP. So you are seeing a 14% uplift

Oh, wait... they stole my "bad math" ?

 

[Abwx]

Your link has only one 7900X result in the test, so what are you talking about? In PCGH release test 88W 7900X is 8% slower than unlocked 7900X (which also only pulled like 180W).

The link has the 7900X at stock, at 142W and at 88W, look better.

 

[leoneazzurro]

PBO? Why not LN OC? )
Anyway, if you haven't read the article (highlighted the key words for you, just in case)

The 7900X score Wcfftech used has PBO, too.
"Also, the 170W TDP (score) disclosed above for the Ryzen 9 7900X has PBO enabled by default."
Just because I read the article.

We also don't know what PBO means for the 9900X in term of power consumption.

 

[PJVol]

The 7900X score Wcfftech used has PBO, too.
"Also, the 170W TDP (score) disclosed above for the Ryzen 9 7900X has PBO enabled by default."
Just because I read the article.

Great! Here is your reward:

 

[PJVol]

We also don't know what PBO means for the 9900X in term of power consumption

To be honest, it seems like some people still have a vague idea of what PBO is (judging by the confusion in that artcle)

 

[Abwx]

Great! Here is your reward:

View attachment 102513

Those charts mean nothing, what matter is the actual power, in Cinebench R23 the 7900X is at 175-180W.

The 9900X will be at 162W at most since it s 120W TDP/162W PPT but i doubt it since the 7900X3D wich has also a 162W PPT doesnt exceed 120W in CB.

From the numbers given by WCCFTech we can iterate that without PBO the 9900X is around 125W and at most at 162W with PBO.

Basically it s 20% faster than the 7900X at 0.9x the power and 14% faster at 0.72x the power, in the first case this amount to 33% better perf/watt and 58% bettter perf/watt for the second case.

 

[leoneazzurro]

To be honest, it seems like some people still have a vague idea of what PBO is (judging by the confusion in that artcle)

One of them seems to be you, as you should be aware that PBO is either thermally or power limited, so in the case of using a insufficient cooling you will have no effect in moving from a stock to PBO setting.

Great! Here is your reward:

View attachment 102513

Great, and you point is (other than showing that Tom is using a bad cooler)?
Article says that 7900X (with PBO enabled, which in case of not enough cooling does not improve vs stock) at 170W TDP is worse than a 9900X stock, and said 9900Xstock with PBO (which with all probability is still lower than 170W) s even higher.
This tells two things: Zen 5 runs cooler than Zen4, and the +14% is achieved at a lower power consumption, with some more power (likely less than standard TDP of 7900X) the difference increases.

 

[igor_kavinski]

Presenting with permission from our wonderful ES user:

 

[PJVol]

Those charts mean nothing, what matter is the actual power, in Cinebench R23 the 7900X is at 175-180W.

This chart was intended to illustrate the fact that the 7900X gains very little from enabling PBO in CB-type workloads, so the 14% MT uplift assumption still stands.
I believe the only thing that matters to most potential buyers is stock vs. stock performance increase, and they don't give a damn **** about all your fancy perf/watt numbers.

 

[igor_kavinski]

But I thought 32%, at least, was inevitable! Tier 1 OEMs don't lie and neither do Anandtech posters!

Rapydmark score (probably unoptimized since it depends on a VC2017 runtime and possibly not using any SIMD) seems to suggest that software will need to be optimized for Zen 5 to unlock its potential. That's my thinking at the moment. The score? Not earth shattering so no point in revealing it or we'll get a WTFtech article saying Zen 5 disappoints!

 

[PJVol]

The 9900X will be at 162W at most since it s 120W TDP/162W PPT but i doubt it since the 7900X3D wich has also a 162W PPT doesnt exceed 120W in CB.

7900X3D is restricted by the additional VID limit for V-cache CCD, which is often reached before the thermal one.

One of them seems to be you, as you should be aware that PBO is either thermally or power limited, so in the case of using a insufficient cooling you will have no effect in moving from a stock to PBO setting.

As long as we compare CPUs in cinebench nT this will always be the case, unless water cooling is used, which is kinda rare amongst the reviewers.

Great, and you point is (other than showing that Tom is using a bad cooler)?

See above.

 

[PJVol]

Presenting with permission from our wonderful ES user:

So, Fmax is 5850... nice )
Although only 25W for cores from the 60W budget is far too little. This is where Zen6 will shine (I hope)

 

[Rheingold]

Your link has only one 7900X result in the test

It's an interactive diagram. There are more entries to be seen when clicking the "+ 183 entries" at the top right. Using the Developer Tools, you can even reduce the diagram to just the 7900* entries:

The wattages are PPT values. Overall, in multi-core-tests, the 7900X loses 12% performance when lowering the limit from 230 to 88 watts. When clicking "Edit", the individual benchmarks can be turned on and off, showing that the biggest performance loss of -15% happens with Cinebench R20 and R23.

You can also see that a limit of 142W costs almost no performance. This makes it clear that they could have released the 7900X as a 120W TDP model with basically no difference in performance. AMD adjusting the TDP of the 9900X down to 120W is just acknowledging that.

 

[igor_kavinski]

I hope we see a mobo design that places these SODIMMs at the back of a mobo. They could also put it in the regular place. Should be great for office/home PCs whose users don't typically tinker and these should be great for Zen 5 too. Seem like the perfect way to get rid of the pesky heatsink clearance issue.

https://www.kingston.com/datasheets/KF564S38IBK2-32.pdf