I am reading it as AMD really pushed for great improvement with efficiency. It was interesting to see such a divergence of opinion between Tech Jesus (Steve) and Wendell. I personally think Wendell is pretty much the most old school uber tech nerd journalist out there, so I am inclined to align my opinions with his at the moment. The X3D SKUs will be the real tell if Zen 5 is a worthy upgrade to 4. I am curious to see the efficiency of those dies with the extra cache.
We don't need X3D to see Zen 5 is a significant step forward. When given the same amount of power as Zen 4, Zen 5 easily meets and beats the AMD uplift claim. Don't be tricked by benchmarks done in eco mode and compared to Zen 4 benchmarks done fully powered. You may notice, there are plenty of Zen 5 benchmark articles showing a big step forward. Have a closer look at the shit talkers. AMD has said X3D will be identical to Zen 4 X3D but there have been leaked comments the new X3D will have significantly lower latency. Perhaps the sheer amount of cache will be the same but I am expecting a performance boost and possibly X3D that will assist both CXX in a dual chiplet complex.
Understood. Then my next question to this is why did AMD not set the TDP to match Zen 4 SKUs and let the IPC increases shine? AMD was already being praised for its efficiency versus Intel. Why not twist the knife and make Intel look worse? It just seems like a wasted opportunity to gain momentum while Intel is down. Their GPU division has not been able to compete with NVDA in a long time, so they need to stick it to Intel every chance they get.
Maybe they didn't want to burn out their ring bus and ruin their reputation? Sorry... that was a cheap shot. I don't know. What I do know is the 9700X is a 65W TDP part (recommended settings) and that causes me to lust a 9700X build. 65W allows the 9700X to be spec'd for SFF devices. It's a huge win, IMO. The 9700X is unlocked so users are free to power up and dominate benchmarks.
I read that one angle AMD might be taking is perhaps using the lower TDP will lower idle wattage and make these SKUs more attractive for power conscious business machines with large contracts. Perhaps taking market share away from Intel that way? I don't know, but that would be an angle worth pursuing I would think. AMD is depressingly underrepresented in business machines.
File this under, DUH! https://www.techpowerup.com/325568/...-boost-agesa-update-increasing-tdp-up-to-105w I am also seeing that Zen 5 performance in Linux is a very nice step up from Zen 4, but Windows not so much with the current TDP limitations. Windows really needs to get their scheduler nonsense fixed. CPUs are not getting less complex and there are so many little tweaks that need to happen now for CPUs to shine. Microsoft really needs to get its ass in gear and hire all the QC staff back.
I really hope this blows up and brings low cost Zen 5 to my door. I'd buy the hell out of a 9900X but I'm a linux guy. Did you see the Level1Techs video from earlier today? Wendell found the 9950X lost by quite a bit to the 14900k at 1440p but beat it at 4K resolution. Very odd. Also, he found that he could make the 9950X faster than the 14900k at 1440p with some of the standard tricks like turning off security or running as admininstrator. You would expect an uplift from these things but he found a tremendous uplift, far beyond what was expected. Both of these things show strong uplift over Zen 4 without adding power. Something is going on at ring 3 that is not going on at ring 0. I believe this is fixable by AMD but time will tell.
HA! I am watching that right now. It seems like the actual CPU is good stuff, but either AMDs software division or Microsoft need to get it optimized to show what it can do. It's very frustrating, as this kind of stuff needs to happen before the launch or it really takes the wind out of the sails. The average layperson does not understand much of this, and if they get their purchasing opinions from someone who only follows the launch headlines, that is a lost sale. Most people are not going to sit around waiting for months while issues are ironed out. Only enthusiasts do that. But then again, Intel is a dumpster fire right now, so it's fun times. I will say though that Linux users who use AVX-512 instructions are over the moon with AMD right now!
Testing shows literally every Windows app runs faster in Linux than in native windows, under Zen 5. Brutal.
It appears the 2H24 W11 patch brings the promised Zen 5 performance. Now we can look back at the drama and panic caused by an extremely minor performance issue that lasted 3 weeks, as we mock and laugh.
Sage words, Rayak. I've been looking through benchmark data and find it interesting that linux scales beyond AMD/Intel claims while Windows scales less. If you look at compile time benchmarks done by Phoronix, Zen 5 is almost twice as fast as Zen 3. Some of that is the 16% clock speed increase, however, Zen 5 does it while using 62% less power. Further, improvements are somewhat linear from generation to generation. Meanwhile, Windows benchmarks look to me like we are up against the limits of the software. It does not scale linearly and seems to hit a bit of a wall. I'm sure Microsoft will improve their pile of junk, given time.
I'm taken back by the smear campaign on AMD. It's remarkably similar to that done to Tesla, in the early years. Back then, fake news kept pouring out for years until it was complete comedy and they turned off the tap. I'm seeing reports of the Zen5 launch failure, AMD choking on unsold inventory, etc. Of course, none of it is true. Personally, I'm waiting for a Zen 5 APU. I'm no gamer.
AMD just released some BIOS changes that make a nice little improvement. Phoronix testing of the new BIOS looks good.
YouTube consensus seems to be that Zen 5 did not meet it's performance targets. It's difficult to find anyone discussing Zen 5 who doesn't crap on it, saying it's only 2~3% faster than Zen 4. This is a bold faced lie that is propagating like oral herpes at the RNC. If we look at linux benchmarks (Phoronix) or compute benchmarks, the 15% uplift is there. Also, we need to compare the 9900X to the 7900X (not the 7900X3D). Even here, the 9900X shows major gains over the 7900X3D in many benchmarks. There are only a few benchmarks with marginal gains due to being GPU limited. I'm a linux user, so I'm heavily biased, but I disregard the current talking points. Nobody seems to be talking about gen 4 m.2 SSD operating at gen 1 speeds on AMD X670E motherboards. These are AMD's top motherboards (sort of... X870E is top but it's identical silicone to the X670E. The only change is the MB manufacturing license forcing USB 4 where the X670E supports USB 4 for MB sellers to use optionally.) This could be an SK Hynix controller issue. Apparently, some SSD manufacturers are replacing the drives and full performance is restored. There is a clear pattern here. Both Intel and AMD CPUs are going to have disappointing performance results on Windows until Microsoft improves their platform. To be fair, I expect Intel's next gen to also be a monster on linux but with disappointing Windows results. For some reason, Windows benchmarks are heavily biasted toward system functions and graphics. Pure compute benchmarks on Windows show great uplift on new CPUs.
