My poor 3900x experience upgrading from i7-4790k
This is a copy-paste from my reddit post. I hope to fix my 3900x with its poor perfromance by going back to AGESA 220.127.116.11. I'm using Asus C7H with 18.104.22.168AB bios (2602) and ram oc'ed: https://i.ibb.co/9vLfrvb/3733best.jpg
This is not purely a CPU comparison. My purpose was to see how much of a difference it will make on my normal usage with all the apps running in background like xbox/steam/epic/battle.net/origin/bitdefender etc. So this was not a clean install or in benchmark mode. By clean install i meant the OS is new installed that too on the 970 evo plus, but all the apps and programs installed as i used it before. Not just benchmarking apps to get the highest scores.
I oc'ed my 3900x to 4.2ghz all-core as well tightening the ram at 3733mhz. The i7-4790k had an all-core oc of 4.6ghz. My MSI 2080 X Trio with +70core/+1000mem and Predator XB271HU 1440p 165hz G-Sync remained the same. Nvidia driver was 461.30 for both comparisons. For 3900x i'm using 1.07.29.0115 chipset driver with 22.214.171.124AB AGESA bios (2602 for C7H). I'd didn't want to use older ones even if some say they performed better as i firmly like to keep everything latest and updated.
It hasn't been easy to make 3900x work properly out of the box. The chipset drivers and bios are not mature enough i think for the not so impressive performance in gaming even with same GPU and driver. Some background apps can have a huge effect on its single-threaded performance. For example when i had just setup the PC, single-core CPUZ bench score was around 545 but now its just 516. This is even after removing apps like Aura Sync, I-cue etc. Of course benchmarks that can fully use the cores showed massive gains.
Even though Avg and Max FPS dropped for most of the games, Min FPS did increase for more than half of the games and the biggest benefit was that it was all micro-stutter free. Two best examples are the battle scene in FF:XV benchmark where Shiva is summoned and Deus Ex: Mankind Divided bench. Even if the scores didn't show there was micro-stutter easily observable on my 4790k.
All the benchmarks below:
So this has been a mixed bag. The main improvement for me has been the freedom from very annoying micro-stutter probably from consistent min fps and cpu not being a bottleneck. I'm pretty sure if i went with i9-9900K, there would have been much more gains in gaming. But i bought 3900x to future-proof for the next 5 years at least like the 4790k before. As its always said get the best gpu you can in your budget and a cpu that can run it. If i was going all-new build then i probably would have went 3700x+2080ti instead.
Will definitely run the benchmarks again when improved chipset drivers and bios with better AGESA updates are available.
Edit: people saying the 4.2 all core is an underclock, it's not. I ran at stock and the benchmarks were even lower. You don't know how many times i re-ran benchmarks and trying to debug the low performance so i could show the 3900x trumping 4790k, i just couldn't. I made a thread few days back asking for the low performance and found out that background apps have so much effect on the processor. Maybe it is the broken Agesa 126.96.36.199AB so will be rerunning the benchmarks on 188.8.131.52.
I did get 3900x seeing all the release benchmarks too but didn't know how just normal usage with the apps you normally run can lower the performance so much. I didn't buy it just to bench on a clean os with only benchmarking apps installed.
Edit 2: i will install 184.108.40.206 and re-run the benchmarks and compare again.
Asus Crosshair VII Hero x470
MSI RTX 2080 Gaming X Trio
T-Force Nighthawk Legend RBG
Samsung 960 Evo Plus
Samsung 960 Evo Plus
Antec High Current Game Extreme
Corsair H115i Platinum RGB
Windows 10 x64
Acer Predator XB271HU IPS G-Sync
Logitech G512 Carbon Romer-G Tactile
▲ hide details ▲