- Mar 3, 2017
- 1,747
- 6,598
- 136
If the audio ecosystem (daw, vst, etc) would be more robust on Linux, I'd switch in a secondAlso.
This fiasco also gives more fuel to NOT use Windows. Like ever.
I would say they tested games only because that is where they saw very little improvement from Zen4 to Zen5. That has been the focus since the release of Zen5, the lack of a performance uplift in gaming. There was improvement in synthetic benchmarks. And while they may be "easy and quick to run", the 40 gaming benchmarks he did run were not. Neither were any of the other 4-5 videos they have posted on the topic. If it's so easy why don't you run them?Several people so far (including a user here) seem to be seeing no change with the 2H24 Win 11 build. Very wonky. Im starting to agree with MarkPost here that HWUBs results are due to something else. Odd thing is, KitGuru got ~ similar uplifts in the few games he tested, but his testing is difficult to follow with VBS ON/OFF, 2H24/2H23, etc.
An average of 10% gains in a 40 game test suite should DEFINITELY show up in some synthetic benchmarks, whether that be Geekbench, 3DMark, CPU-z, or whatever. I cant believe that someone like HWUB that goes through all the trouble of testing those games doesnt throw up token Geekbench, Cinebench, CPU-z benchmarks with every configuration change to track whats changing, considering they are so easy and quick to run. SMH.
Im starting to think this is all a big nothingburger.
Ryzen and Windows is difficult, HUB could have had a bugged Windows 23H2 Installation. Computerbase also had difficulties with their 9900X despite a fresh Windows. It performed worse than 7600X and suddenly after another fresh Windows it performed like it should.Several people so far (including a user here) seem to be seeing no change with the 2H24 Win 11 build. Very wonky. Im starting to agree with MarkPost here that HWUBs results are due to something else.
So we have two reputable benchmarking youtubers getting similar results.Several people so far (including a user here) seem to be seeing no change with the 2H24 Win 11 build. Very wonky. Im starting to agree with MarkPost here that HWUBs results are due to something else. Odd thing is, KitGuru got ~ similar uplifts in the few games he tested, but his testing is difficult to follow with VBS ON/OFF, 2H24/2H23, etc.
An average of 10% gains in a 40 game test suite should DEFINITELY show up in some synthetic benchmarks, whether that be Geekbench, 3DMark, CPU-z, or whatever. I cant believe that someone like HWUB that goes through all the trouble of testing those games doesnt throw up token Geekbench, Cinebench, CPU-z benchmarks with every configuration change to track whats changing, considering they are so easy and quick to run. SMH.
Im starting to think this is all a big nothingburger.
That seems to be a problem with AM5 platform, with dual CCD, CCD with extra L3$ vs CCD L3$, etc.Ryzen and Windows is difficult, HUB could have had a bugged Windows 23H2 Installation. Computerbase also had difficulties with their 9900X despite a fresh Windows. It performed worse than 7600X and suddenly after another fresh Windows it performed like it should.
All tests were performed after a clean installation of Windows 11 23H2 (Build 22621.3958) together with the latest AMD chipset drivers and Nvidia Game Ready Driver 560.81.
NBC got numbers quite better than average, here their conditions, notice the bolded :
All tests were performed after a clean installation of Windows 11 23H2 (Build 22621.3958) together with the latest AMD chipset drivers and Nvidia Game Ready Driver 560.81.
how can't they all just make a standard WINDOWS BENCHMARK ISO so they all use the same thing
I thought only the Snapdragon Elite Copilot+ PCs shipped with 24H2, in order to enable the Copilot+ features.Strix is slower in Windows compared to Linux anyway, and all the laptops are shipped with the 24H2 update.
I recall a comment from David Huang but that seemed weird they would ship it to laptops if desktop version is still in preview 1.5 month after Strix releaseI thought only the Snapdragon Elite Copilot+ PCs shipped with 24H2, in order to enable the Copilot+ features.
And the HX370 laptops shipped with 23H2, which is why they couldn't ship as Copilot+ laptops with the features enabled.
Source on HX370 laptops coming with 24H2?
Even if STX laptops shipped with preview 24H2, there's no guarantee that version of the preview included these fixes. Unless someone explicitly tests the known fixed version of 24H2 I will assume these fixes were not present for benchmarking.I recall a comment from David Huang but that seemed weird they would ship it to laptops if desktop version is still in preview 1.5 month after Strix release
Tried this? https://ubuntustudio.org/tour/audio/If the audio ecosystem (daw, vst, etc) would be more robust on Linux, I'd switch in a second
No, I need my windows programs ported to Linux, to work there just like they work on Macs. Not re-learn or buy other things, because I need all my old projects working not only the future ones. So until the vendors start working for Linux too... I'm stuck on Windows.
Healthy increase in 1% low in some titles thohttps://www.hardwareluxx.de/index.p...rsage-zen-5-mit-windows-11-24h2-getestet.html
Seems like they found a significant performance improvement with W11 24H2, though not quite like Hardware Unboxed.
Not really. The results are all over the place, 3D even has a few major performance losses and there isn't a clear sign at any game, sometimes all ZEN5 behaves the same, sometimes Single CCD gets bigger Boost, sometimes dual CCD has bigger Boost, random regressions in lost games. That's a completely different result to Hardware Unboxed.Seems like they found a significant performance improvement with W11 24H2, though not quite like Hardware Unboxed.
Synthetic benchmarks may not see a big gain. You have to remember that the source of the performance gains* is the performance drops that are introduced by speculative attack vulnerabilities mitigations. Those mitigations are applied for OS code that handles context switches, kernel calls and that sort of stuff. So you can expect gains in places that saw heavy hits back when Spectre and "friends" were getting patched.An average of 10% gains in a 40 game test suite should DEFINITELY show up in some synthetic benchmarks, whether that be Geekbench, 3DMark, CPU-z, or whatever. I cant believe that someone like HWUB that goes through all the trouble of testing those games doesnt throw up token Geekbench, Cinebench, CPU-z benchmarks with every configuration change to track whats changing, considering they are so easy and quick to run. SMH.
Reviews.I thought only the Snapdragon Elite Copilot+ PCs shipped with 24H2, in order to enable the Copilot+ features.
And the HX370 laptops shipped with 23H2, which is why they couldn't ship as Copilot+ laptops with the features enabled.
Source on HX370 laptops coming with 24H2?
ASUS Zenbook S 16 (2024) can come with Windows 11 Home or Pro preinstalled. The unit I’m reviewing had the Pro edition, version 24H2, which Microsoft hasn’t finalized yet. This is because ASUS, Microsoft, and AMD wanted this laptop and similar devices to have access to some of the AI goodies built into this version of Windows.
The drivers on this PC are only compatible with Windows 11 version 24H2 or later. To ensure optimal functionality and avoid any potential issues, please do not manually install an older version of the operating system.
I can definitely say that disabling spectre/meltdown fixes via registry alleviates the difference between 'true' admin mode and local admin mode for me personally, and the uplift in CP2077 and FC6 is similar to what HW Unboxed and the other tech media gained (and I have fully tuned system and generally know what I'm doing), and I run Win10 22h2 instead of Windoze 11 of whatever variety. The only thing I couldn't replicate is the Gears of War 5 35% discrepancy, but I've already theorized that this might be related to how the benchmark disrespects the in game video settings and uses something else instead of what you set.An average of 10% gains in a 40 game test suite should DEFINITELY show up in some synthetic benchmarks, whether that be Geekbench, 3DMark, CPU-z, or whatever.
I would say they tested games only because that is where they saw very little improvement from Zen4 to Zen5. That has been the focus since the release of Zen5, the lack of a performance uplift in gaming. There was improvement in synthetic benchmarks. And while they may be "easy and quick to run", the 40 gaming benchmarks he did run were not. Neither were any of the other 4-5 videos they have posted on the topic. If it's so easy why don't you run them?
What's stopping you from getting them and running the tests. It's so "easy and quick" to go on Amazon and click buy now?Where is a link to these improvements in synthetic benchmarks you speak of? Geekbench is very easy to run, it takes less than 5 minutes, and if I had the processors, I would run them?