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

[DisEnchantment]

Speculate at will

 

[eek2121]

No he wasn't. He said SPEC rate. And neither FP nor int show that.

It is quite possible that is affected as well. Zen 4 is starved for bandwidth as well.

We can actually determine this pretty easily by simply waiting for Turin benchmarks.

If non-SPEC INT “INT” tests are showing a 30-35% improvement, that suggests the problem is SPEC being limited in some way.

Note that I am it is the current chiplet design (especially 1 CCD parts!) that is limiting bandwidth, NOT DDR5. It also means Zen 6 should rectify the situation. 😈

🚂 ( hype train)

 

[yuri69]

Geekerwan managed to get spec results, +8.6% int & 26% fp.

The INT figure makes Zen 5 the lowest Zen gen-to-gen improvement given the Zen+'s 3% gain is excluded. This certainly wasn't expected.

 

[adroc_thurston]

So if the below is true, @adroc_thurston was correct.

View attachment 104680

Not quite.
It's very fast in branchy web stuff, but somehow not in video games?
Somebody arm the profiler.

 

[csbin]

It is quite possible that is affected as well. Zen 4 is starved for bandwidth as well.

We can actually determine this pretty easily by simply waiting for Turin benchmarks.

If non-SPEC INT “INT” tests are showing a 30-35% improvement, that suggests the problem is SPEC being limited in some way.

Note that I am it is the current chiplet design (especially 1 CCD parts!) that is limiting bandwidth, NOT DDR5. It also means Zen 6 should rectify the situation. 😈

🚂 ( hype train)

GMI3-Wide

 

[fastandfurious6]

Where Zen 5 actually shines is temperature:
View attachment 104662
That's -35°C from 7700X to 9700X and still -15°C when both are running at 142W PPT. Thank god, this will hopefully end the thousands of "my CPU is running too hot" posts in German forums that I had to endure for Ryzen 7000 during the last two years.

Bingo, lots of OC headroom and proper mobile (Halo / Fire range) will really shine

zen5 is a perfect specimen to have perma-OC

 

[CouncilorIrissa]

So if the below is true, @adroc_thurston was correct.

View attachment 104680

Nah, the claim was +32% iso-clock in SIR 2017.
What he did get right is a double-digit increase in x87 lol.

 

[biostud]

Per Phoronix:



Those are some impressive numbers and suggest Zen 5 server parts are going to be beastly. Since the architectural changes seem to favor work more so than gaming.
9700x is 15% better than the 7700X despite 65W TDP (88W PPT).
9600x is 25% better than the 7600X. And a full 82% better than the 5600X.

It does seem like the improvements primarily is workstation/HPC/power related. On the other hand if they can keep the high clocks then the 3D cache models might be interesting.

 

[Hitman928]

As a small note, guru3d also ran browser benchmarks on Windows and confirm the same (if not better improvement) that Phoronix saw in Linux. I find this result very interesting in light of all the other results. This is a near 40% improvement in performance per clock.

 

[static shock]

It looks like the both processors, 6 or 8 cores have focus at professional loads but gaming was meh. I talk this because games heavily use the FP.

 

[gaav87]

What I don’t get is how / why Tom’s is showing a significant uplift in gaming with PBO on? Basically conflicting with every other source.

Just leaves me puzzled

• Ryzen 7 9700X: Default power limits, DDR5-5600
• Ryzen 7 9700X PBO: Precision Boost Overdrive (PBO) — advanced motherboard, 10X scalar, +200 MHz CPU clock, -20 Curve Optimizer, DDR5-6000 EXPO profile (fabric 2000 MHz, mclk/uclk 1:1)
  
  stock 5600mt/s vs expo, pbo +200mhz -20curve manualy tuned xD you answer

 

[adamge]

its efficient and they have priced it well. We obsess too much with benchmarks anyway.

This is all I would have ever asked for in a new CPU generation.

 

[fastandfurious6]

Consolation prize is that overclocking becomes a thing again with these reduced TDP parts.

9700x OC /w $100 cooler + $100 32gb ram

GB6 ST - 3550
GB6 MT - 20000

we have a winner

 

[CakeMonster]

Some speculation: If Zen6 arrives in ~18 months or less, like very late 2025, I could imagine this looking better with hindsight. The IOD/memory issues could see a lot of improvement, keeping 8c CCX with N3* node improving transistor budget, as well as 'sweet spot' memory/sync being raised significantly above 6000.

 

[DavidC1]

So if the below is true, @adroc_thurston was correct.

View attachment 104680

He claims "40% IPC improvement in SpecInt" which turned out to be totally false. The picture shows the same thing. Geekerwan tests of both laptop and desktop parts show 9% gain in guess what? SpecInt.

 

[Kepler_L2]

Some speculation: If Zen6 arrives in ~18 months or less, like very late 2025, I could imagine this looking better with hindsight. The IOD/memory issues could see a lot of improvement, keeping 8c CCX with N3* node improving transistor budget, as well as 'sweet spot' memory/sync being raised significantly above 6000.

Zen6 is 2027

 

[Markfw]

He claims "40% IPC improvement in SpecInt" which turned out to be totally false. The picture shows the same thing. Geekerwan tests of both laptop and desktop parts show 9% gain in guess what? SpecInt.

maybe not He was comparing either 9950x vs 7950x or Turin vs Genoa. I am guessing Turin. Until those come out, I will not contest that number.

 

[CakeMonster]

Zen6 is 2027

Well, then we have a problem.

 

[DavidC1]

maybe not He was comparing either 9950x vs 7950x or Turin vs Genoa. I am guessing Turin. Until those come out, I will not contest that number.

Nope. He clearly says "Scalar Integer" which is what SpecINT is. Somebody faked the numbers and passed it around as fact.

 

[StinkyPinky]

It's almost like AMD has decided the non-3D is for productivity and the 3D is for gaming.....

 

[Hitman928]

Zen6 is 2027

If Zen 6 is a 2027 product, AMD is going to have a major problem remaining competitive in 2026.

 

[StinkyPinky]

If Zen 6 is a 2027 product, AMD is going to have a major problem remaining competitive in 2026.

Intel can't even stop their CPU's melting at the moment, I'd wait and see before I get excited about their promises.

 

[DisEnchantment]

Zen6 is 2027

This time you will be correct.

 

[CakeMonster]

Might be DDR6 then? I had absolutely called Zen6 as being still 8core CCX + DDR5, but I had also imagined it would arrive at maximum 24 months after Zen5.

 

[Markfw]

Nope. He clearly says "Scalar Integer" which is what SpecINT is. Somebody faked the numbers and passed it around as fact.

My comparison could still be for these CPUs. I did not say what tests, so he still could be right.

 

[ondma]

If Zen 6 is a 2027 product, AMD is going to have a major problem remaining competitive in 2026.

Why? Most likely nearly 2025 for ARL, which probably will at best (for Intel) trade blows with Zen 5. Most of 2026 will be ARL refresh, which probably wont bring much to the table. Maybe 2027 will bring Nova Lake, which is rumored to be a significant step up, but the way Intel is executing now, I would not count on it either being on time or delivering the expected performance uplift. So if Zen 6 is late 2026 or early 2027, they should be fine.

I am talking on desktop. Mobile, I have no idea. I am not interested in that area, and the line-ups from both manufacturers are seriously confusing. So I dont know how things will stack up there, although AMD has a serious power advantage now, so even if Intel makes big gains in efficiency, I think AMD should also be fine in mobile.