AMD Threadripper 2950X versus i9 line?

[MiRai]

I forgot how to get ingame readings like what you posted MiRai. The main is in Boralus and my alts are in Dalaran--- beating on a turnip. I found that while dpsing it puts a realistic stress on the CPU/GPU to give a more accurate reading.

I lowered my CPU OC to 4.4ghz to match your cpu better. I removed the GPU's OC entirely.
Renderscale 100%. Slaves set to 1. DX11. Physics Interaction Player only. Main settings 10, 3440x1440p. Here
*edit* Slaves are 20fps background and the foreground is 144fps *edit* 30 'slaves' and my main. So 31.

I then thought maybe it's some weird 21:9 vs 16:9 bug. So I messed with the renderscale and had it output @ 5120x2880 (renderscale 200% - 2560x1440, XB270HU- old monitor) . Had my girls fighting a turnip in dalaran, took my main into LFR. 45-55fps consistent. I then had my friends pull all the trash and unload AOE style~ that caused my poor little 980ti to be overwhelmed; stuttered and dropped frames (20fps). Here

Notice how under normal WoW gameplay (first picture) my GPU doesn't really even come out of powersaving. Goes up 30w. When I changed the resolution to 5120x2880 it went to 250w.

FWIW a friend is playing 20 on a 3570k. Although he has optimised his setups to allow it to work out ok. (renderscale 0.1 on alts etc etc).

Some things to note...

• Your first screenshot shows all cores being worked, when you said you were only using four out of eight. This isn't a big deal since your overall load isn't much different than what you originally said, but I just wanted to point that out.
• Are you changing focus before taking screenshots? The reason I ask is because the moment you pull focus away from the game clients they usually fall back to their background framerate, which can greatly reduce load.
• The first screenshot shows you only using 1.8GB of VRAM. When I log in a solo character and use a render scale of 2986x1680 that almost matches the pixel count of 3440x1440, I eat up almost 1.3GB at an in-game slider setting of 7 (no AA). You're running 29 more game clients and you're only eating up 500MB more than my one game client.
• You're pumping out 724 frames across 30 game clients (29 x 20 + 144) and your GPU load is only ~30%. Using the same in-game settings from my previous bullet, I barely achieve 320 FPS and I'm zoomed into first-person while staring at the ground with my UI disabled, and my 1080 Ti is pegged at 100% load at this point. If I try to play normally (still solo), my FPS bounces around between 120 and 190.

Do you see how things aren't adding up... at all? Ultimately, what I would like to know is whether or not you're using ISBoxer. This is a huge missing detail that may reveal why you're able to do what you can do. If you are using ISBoxer, then I'm going to, once again, ask to see your profile, and if you aren't, then I'm going to ask that you post the bit of your script that has your layout in it, since you're obviously doing something much different with your setup, that the vast majority of multiboxers aren't.

*second edit*
Due to RAM being so overpriced I just bought a extra SSD and use it as a dedicated storage for Pagefile.sys.
Frees up physical RAM and the performance hit is really negligable. Maybe that helps too? I don't know.

It wouldn't make any difference in loads being produced on the CPU and GPU while standing still. Not only that, but a PageFile is unlikely to be used when you still have RAM available, which, according to your screenshots, you do.

*third edit*
I had a look at MadMilitia's setup as you mentioned.. the 1080 gtx is better than my 980ti's for sure, but the CPU is very different. 8mb cache vs 20mb. 16 pcie lanes vs 40. The Memory bandwidth is half- 34gb/s vs 64gb/s. I'd expect it to struggle a little bit-- but not as much as it was.

Both chipsets have PCIe 3.0, and any piece of hardware can only have access to 16 lanes at any given time. Just because you have access to more lanes, overall, doesn't mean his GPU will be bottlenecked—your GPU is plugged into an x16 slot, and his GPU is plugged into an x16 slot.

If there was any difference, it would be between 8x and 16x, and even then, the difference between those, for the past several generations of GPUs, and has been shown to be negligible (plenty of reviews out there on this topic). Also, the cache is not going to be the deciding factor between 30 games clients and 8 game clients.

----------

I do mainly EVE Online these days, but recently dipped back into wow to try it out. I'm certainly not running everything at a buttery smooth 60fps constantly but it is certainly playable. I have 5 wow accounts all running with default graphics (7) @ 1080p, with background limited to 30, and the main client averages around 40fps. CPU utilization hovers around 50%. This is on an i7 930 oc'd to 4.22ghz and a gtx 1070.

In EVE I do 20 clients but you more or less need 0 graphics on your alts, so setting them to absolute minimum settings is common.

There we go, more details!

What I take issue with is that, in your previous post you said you had zero issues, but, for me, not being able to maintain 60 FPS (when using a 60 Hz panel) would most definitely be an issue. You may not find that to be an issue, but I guarantee that someone else will, and by throwing a blanket statement of "I have zero issues" when referring to only a 9-year old CPU can easily give people this false sense of what they need to buy—not to mention you also have a 1070, which is not a 9-year old GPU.

Now, to be fair, the OP doesn't mention how many clients they'll be playing, nor what resolution they'll be running at, so we don't have that to go off of, but I do have two final questions:

1) What zone(s) are you in where you're seeing the performance you report?
2) Why not just drop the slaves down to 20 FPS to achieve that buttery smooth 60 FPS on the main?