Any source on Granite Ridge being able to barely maintain boost clock in R23? And could you give a context to barely?Almost certainly the "core parking" thing they are implementing during gaming is required because of the increased cross CCD latency, not the cause of it. With it over doubling from Zen 4, any game threads that crept onto CCD 1 could bring a massive drop in performance.
Also, regarding X3D, the chance that they launch with a greater than 5.3G boost is almost zero. In fact, I'd be more willing to bet that it might be 5.2G out of the box and they "will allow the user" to try to obtain more at their own peril. Think about it, Zen 5 barely holds its 1T boost in something like R23 without a delicate X3D chip stacked on top of it, why in the world would we expect significantly higher boosts than Zen 4 X3D?
Cinebench R23 1T scores. 5.7GHz should be almost 2400 according to AMDs own official R23 IPC uplift data. Its nowhere close to that.Any source on Granite Ridge being able to barely maintain boost clock in R23? And could you give a context to barely?
That doesn't meant the boost doesn't hold, it might mean the performance target for the boost clock was overstated. Plus top boost clock always was given as something attainable under basically idle load conditions For reference http://www.numberworld.org/blogs/2024_8_7_zen5_avx512_teardown/#throttling Zen5 maintains usually higher clocks and lower power draw than Zen4Cinebench R23 1T scores. 5.7GHz should be almost 2400 according to AMDs own official R23 IPC uplift data. Its nowhere close to that.
Yes I agree. I simply meant that some people thought driver is the reason for high latency and since you are not using it, we know it's not.
That would make sense if this was burst behaviour but I guess you can keep all cores active [disable power saving options in the bios should prevent lower power states? maybe something in the windows power planes] and the picture wouldn't be changed fundamentally. Might be that test itself is enough to keep the cores awake.
Almost certainly the "core parking" thing they are implementing during gaming is required because of the increased cross CCD latency, not the cause of it. With it over doubling from Zen 4, any game threads that crept onto CCD 1 could bring a massive drop in performance, so they were basically forced to do it or risk Bulldozer'esque gaming regressions.
Also, regarding X3D, the chance that they launch with a greater than 5.3G boost is almost zero. In fact, I'd be more willing to bet that it might top out at 5.2G out of the box and they "will allow the user" to try to OC to obtain more at their own peril. Think about it, Zen 5 barely holds its 1T boost in something like R23 without a delicate X3D chip stacked on top of it, why in the world would we expect significantly higher boosts than Zen 4 X3D?
Hmm, not sure but their 7950X seems about 200MHz slow.That doesn't meant the boost doesn't hold, it might mean the performance target for the boost clock was overstated. Plus top boost clock always was given as something attainable under basically idle load conditions For reference http://www.numberworld.org/blogs/2024_8_7_zen5_avx512_teardown/#throttling Zen5 maintains usually higher clocks and lower power draw than Zen4
Ah, the reason is author is using y-cruncher sub-part that is insensitive to memory bandwidth for this comparison. This is tightly optimized math routine that will generate much more load than R23 loop can, that is doing mostly scalar computations. Plus if you read the page further there was a silent stability fix that lowers clocks on Zen4 when certain mix of instructions is usedHmm, not sure but their 7950X seems about 200MHz slow.
The author might be a bit biased towards what "real-world" workloads are as he deals mostly with heavy avx stuff, so cinebench that is using scalar SSE most of the time might not count as such.September 2022: AMD launches Zen4. The 7950X clocked at 5.7 GHz which was easily achievable in many lightly threaded and low-intensity workloads.
However, this 5.7 GHz would crash the chip under certain AVX workloads. The work-around was to lower the clock speed to 5.5 GHz by applying a -200 MHz offset to the PBO (Precision Boost Overdrive).
January 2023: AMD releases the AGESA 1.0.0.4 microcode update. It effectively drops the clock speed from 5.7 GHz to 5.5 GHz to fix the AVX instability - thus formalizing the workaround. While 5.7 GHz was still possible, it required operating conditions that were unrealistic and unlikely to be sustainable in any real-world workloads.
August 2024: AMD launches Zen5. The 9950X is also clocked at 5.7 GHz. But unlike the 7950X, the 9950X's 5.7 GHz is both easily achievable in normal operating conditions and is fully stable with all workloads.
I can't speak for any video game tests as I don't look at those at all. But as far as workstation benchmarks go: It is completely safe to go there if you recall in time that according to Igor W., workstations are basically CAD workstations; there ain't such a thing as a computing workstation, developer workstation, etc..in many cases it's Zen 4 minus minus.
Everybody, do yourself a favor and DON'T check out the igorslab.de 9950X review.
You don't deserve depression. No one does.
Don't GO THERE!
We got 2277 https://www.hwcooling.net/en/amd-ryzen-9-9950x-at-the-very-top-with-both-1-and-32_threads-review/21/Cinebench R23 1T scores. 5.7GHz should be almost 2400 according to AMDs own official R23 IPC uplift data. Its nowhere close to that.
9950X clearly doesnt hold as well as 7950X in this example, plus its about 50MHz lower. 9700X looks fine, but is this R23 or R24? 24 seems to require less power per/Mhz than 23. If it is 23, then AMD just flat out lied about Zen 5s +17% IPC gain in R23. 7700X scores ~2000 ST @ basically the same 5.5GHz boost clock (even though its rated for 5.4GHz). This should put the 9700X at a cool 2340 pts. I have yet to see a >2300 R23 score for any 9700X @ stock, but I may have missed it. 9950X should be touching 2400 but also comes nowhere close to that.The C2C measurements are based on burst behavior. I'm not sure what you can do to keep all cores awake on Zen 5, it should be possible, but I don't think anyone's actually tried yet to confirm.
I don't think we have enough data or insight into this to say for sure. It doesn't make much since that the cross CCD latency would go up so much given that nothing has changed in the IF or IOD except if there is some kind of power saving feature that has been implemented on the CCDs. So far, the best candidate seems to be that the cores on an entire CCD are parked when low threading activity is happening and it takes a bit of time to wake them up when checking the latency from CCD to CCD. Even without the extra CCD latency, CCD thread migration was always an issue before, especially with the X3D chips, so I doubt this is something they just barely thought of with Zen 5. I'm more inclined to believe that their experience with the X3D chips and working on the driver side to keep games on 1 CCD led them to the decision that they can park a whole CCD on Zen 5 to save power by implementing the same kind of driver. Hopefully AMD covers this at Hotchips or in some papers at some point.
Zen 5 holds its 1T boost just fine? The axis marking suck, but it is holding just a hair over 5.7 GHz the whole run.
View attachment 105324
Temperatures and clock rates
Once again, I have deliberately put this section on the last page because no one else will find it. However, I have once again shortened it a little, because it is relatively easy to explain. The Ryzen 9 9950X is definitely easier to cool than the Ryzen 9 7950X. At 76 °C (ok, 35 watts less), it was even cooler in the Blender run than the Ryzen 9 7950X at 88 °C. AMD can take credit for this. The CPU is therefore thermally rather inconspicuous and the technical solution can be found in the deep dive on page one. Even under load, my sample boosted often enough up to 5.7 GHz and thus also meets the specifications. What was noticeable, however, was the significantly more hectic clock behavior under rapidly changing loads.
Honestly @Markfw for what you do it'll probably be a great chip, provided you aren't using Windows anymore. Especially since you already have a bunch of AM5 boards it's really not price-practical to consider switching to a different socket.So 26% more across the board for 20 watts less (average across all test) than the 14900k, and all I see are negative results and reactions. And no specific avx-512 results that I saw yet.
9950X clearly doesnt hold as well as 7950X in this example, plus its about 50MHz lower. 9700X looks fine, but is this R23 or R24? 24 seems to require less power per/Mhz than 23. If it is 23, then AMD just flat out lied about Zen 5s +17% IPC gain in R23. 7700X scores ~2000 ST @ basically the same 5.5GHz boost clock (even though its rated for 5.4GHz). This should put the 9700X at a cool 2340 pts. I have yet to see a >2300 R23 score for any 9700X @ stock, but I may have missed it. 9950X should be touching 2400 but also comes nowhere close to that.
Well is this an improvement or not?Also seems 9950X used more power to get there, ~17%, likely hitting its 230W PPT which 7950X doesn't. Watt values include VRM losses etc, it's a measurement on 12V cable.)
I do not agree with him at all, linux test are more like sever grade test, linux is not used as windows as home user do use them, with xx and xxy service and appz running at same time...From phoronix review:
Simply put, I am extremely impressed with the Ryzen 9 9900 series. If you are a creator, developer, or just doing any heavy lifting on your desktop across a range of workloads, the Ryzen 9 9950X and Ryzen 9 9900X proved to be terrific options. These new AMD Ryzen 9900 series processors deliver great generational uplift and better power efficiency than the prior Ryzen 9 7900 series parts and the Intel Core 14th Gen competition. At $499 USD for the Ryzen 9 9900X and $649 USD for the Ryzen 9 9950X these processors are also priced fair.
So 26% more across the board for 20 watts less (average across all test) than the 14900k, and all I see are negative results and reactions. And no specific avx-512 results that I saw yet.
Edit : it appears these bad reviews are all based on windows feedback, where there could be drivers or kernel issues.
We got 2277 https://www.hwcooling.net/en/amd-ryzen-9-9950x-at-the-very-top-with-both-1-and-32_threads-review/21/
13% uplift over 7950X (2024) with the same DDR5-6000 CL30 memory for both CPUs..
However, the performance increase is better in multithread testing. We also get +13 %, but 9950X does it at lower clocks. Including logged clocks with MT scores into the calcultion, I get MT IPC increase of 16.4%. In ballpark/margin of error of AMD's claim (which was +17% in june).
(Edit: Also seems 9950X used more power to get there, ~17%, likely hitting its 230W PPT which 7950X doesn't. Watt values include VRM losses etc, it's a measurement on 12V cable.)
So what is the default value for PPT? Some people report 200W others 230W PPT inline with 7950x.(Edit: Also seems 9950X used more power to get there, ~17%, likely hitting its 230W PPT which 7950X doesn't. Watt values include VRM losses etc, it's a measurement on 12V cable.)
Those parts of the Phoronix tests which are computing throughput tests would most likely fare well on Windows too.Honestly @Markfw for what you do it'll probably be a great chip, provided you aren't using Windows anymore.
It is reasonable to test a thirtytwo-threaded big-core CPU with computing throughput oriented loads.I have one 13700k desktop use
and 12 ryzen linux cli more server kind of load, he do the same its not fair to compare it with desktop windows use at all
No I did not. I said it can barely hold its 1T boost. I said this, because Zen 5 5.5GHz = Zen 4 5.4GHz, and Zen 5 5.7GHz is 30-50 MHz less than Zen 4 5.7GHz and more erratic (per GN and Chips N Cheese). Also, despite having a more power efficient variant of process node, Zen 5s 9950X top 1T boost pulls about 5 to 10W more power than 7950X top 1T boost per Toms Hardware. The point is, it appears it takes more power to get Zen 5 to match Zen 4 clocks, whether 1T or nT, so theres no logical reason to expect miracles in clock speed from Zen 5 X3D.You said that Zen 5 can’t hold its boost, this clearly shows it can. Anything above 5.7 GHz is above stated boost and will depend on the specific sample being tested, just like with Zen 4.
Grabbed this before the updated treat these like vcache cpus guidance.With all of these mixed results, I think I'm going to hold out for the userbenchmark review before making any decision.
61 when X3D comes out.Disagree on that. At first thread director didn't work on my 12600K because the antique BSD didn't support it. But the next version did and performance improved.
It can sometimes happen. Probably not for Windows though, Microsoft doesn't care about all 60 Zen 5 DIY users.
Find me on OC.net and see my signature from a user that said Bulldozer would have been fine if it ran at 6Ghz plus another quote.This launch shares many parallels with Bulldozer, although the absolute performance and power characteristics of the product are nothing like Bulldozer.
Anyways, both got:
* Horrible hype and projections
* Slow and painful hype train derailing process
* Incompetent marketing fluff
* Unorthodox x86 architecture solutions
* Last minute software "adjustments"
* Linux performance gains more impactful than the Windows ones
* etc.
But probably the worst part having striking resemblance to Bulldozer are the expectations for future products - "fixing the shortcomings" and "unlocking the true potential". This sounds exactly like all those people talking about the great foundation for the future. It remains to be seen.
No I did not. I said it can barely hold its 1T boost. I said this, because Zen 5 5.5GHz = Zen 4 5.4GHz, and Zen 5 5.7GHz is 30-50 MHz less than Zen 4 5.7GHz and more erratic (per GN and Chips N Cheese). Also, despite having a more power efficient variant of process node, Zen 5s 9950X top 1T boost pulls about 5 to 10W more power than 7950X top 1T boost per Toms Hardware. The point is, it appears it takes more power to get Zen 5 to match Zen 4 clocks, whether 1T or nT, so theres no logical reason to expect miracles in clock speed from Zen 5 X3D.
No I did not. I said it can barely hold its 1T boost. I said this, because Zen 5 5.5GHz = Zen 4 5.4GHz, and Zen 5 5.7GHz is 30-50 MHz less than Zen 4 5.7GHz and more erratic (per GN and Chips N Cheese). Also, despite having a more power efficient variant of process node, Zen 5s 9950X top 1T boost pulls about 5 to 10W more power than 7950X top 1T boost per Toms Hardware. The point is, it appears it takes more power to get Zen 5 to match Zen 4 clocks, whether 1T or nT, so theres no logical reason to expect miracles in clock speed from Zen 5 X3D.
Precisely-- now do you think that bodes well for X3D boost clocks?More throughput at a given frequency imply more power in proportion, what matter is the efficency.
AMD Ryzen 9 9950X – ett litet prestandalyft för mer pengar - Test - Effektmätning
Trots att flaggskeppet Ryzen 9 9950X är både energieffektiv och kompetent faller den mot konkurrensen från AMD:s egna X3D-processorer.www.sweclockers.com
I'm curious.Precisely-- now do you think that bodes well for X3D boost clocks?
isnt X3D issue voltage and not power ?Precisely-- now do you think that bodes well for X3D boost clocks?