7800x3d cache project

[MindCrime]

I built a new computer for my mom, she just uses it to browse web scan business docs.. nothing serious at all. So i mocked up a cheap budget build and i figured integrated graphics would be fine. but she said she wanted to go a little extra.
So it's a 7800x3d 32gb ddr5, and the big upgrade for her is going from an HDD to an NVME SSD. (still on CPU graphics, i have a 750ti i can throw in, but haven't)

So these 7800x3ds seem to be doing pretty damn good in certain benchmarks and games. I want to see how it does on higher cache boinc projects, so which ones would utilize the larger cache the best so i can compare it to other cpus?

Also i guess i should try how it is GPU wise, any projects with good AMD integrated graphics support?

[MindCrime]

I'm having problems getting into (finding) the BIOS XMP settings. first of all I can't find anything that says XMP in all the settings "advanced" or not. When I get into more advanced settings menus i do not see XMP or DOCP, there are some weirdly named Mem settings under AI Tweaker, but'll have to take phone images and post them 3rd pary and bad quality.. SO just assume they're garbo, why is XMP so god damn hard?

[Ron Shurtz]

I think it is in the "AI Tweaker" Under the "AI Overclock Tuner" options.

I have no experience or knowledge in this area, so take it with a grain of salt.

:D

[zombie67]

So these 7800x3ds seem to be doing pretty damn good in certain benchmarks and games. I want to see how it does on higher cache boinc projects, so which ones would utilize the larger cache the best so i can compare it to other cpus?


PG is the project to test this out on. In the project preferences, it lists the cache requirements for each of the CPU sub-projects. In general, you want to run as many MT tasks as you can, while keeping under the cache total amount. And you want the MT setting to be an even multiple of your Cores. For example, PSP uses 28mb of cache per task. 96/28=3.4 tasks could be run. So you would want to run 3 tasks. But that CPU has 8 cores. So that doesn't work. Your best bet is to run two tasks at 4 cores each. That is still way better than CPUs that cannot fit even one task in the cache, since most have only 16mb or less.

Another example is SR5, which uses just over 10mb. So you could fit up to 9.6 tasks based on the cache. In this case, you can fit 8 single-thread tasks at the same time easily. A non-3D chip has only 32mb, which means the best it could do is only two 4-thread tasks at a time. Because 2 3-thread tasks leaves 2 cores idle, and 4 two-thread tasks exceeds the cache size and everything slows WAY down if the tasks overflow and have to go to RAM.