To anyone who was interested how multiple cpu cores scale in gta V heavily modded, here is a 4k video with statistics in real time

[er557]

This is on a 36 core, 72 thread e5 v3 setup, patched to run @3.5ghz, the OSD shows that multiple cores are in use, and gpu usage is good as well, with sli.
(dont mind the 7980xe lables, it's from another cfg file for MSI AB)

 

[EarthDog]

I cant see a thing on my phone... overlay color matches the scenery, lol.

I'd be more curious to see this on modern Intel instead of older lower clocked xeons. I'd bet money those GPUs are held back by the cpu at 4k due to clocks and IPC.

 

[er557]

Dont watch this on a phone , it requires a 1440p screen, the larger the better, 4k recommended, you can then pause the video and watch those figures on some dark background, I see the OSD perfectly on my 4k 40"

 

[EarthDog]

My phone is 2960x1440. I'll check it out shortly. Thanks!

PS - your thread title leads me to believe there is more than one game... may want to have it read something like this...

Edit: the formatting copied over and I cant fix it on my phone, sorry. Lol

To anyone who is interested how multiple cpu cores scale in a heavily modded GTA V, here is a 4k video with statistics in real time.

 

[er557]

You're right, done!



Also dont mind the cpu wattage below, as it is patched to ignore TDP and think it is 50w even when it's 300W, AKA powercut.

 

[Jetster]

I got 1440 and cant see shit. I can see it but its still really small
4 K is a little better.
So basically there all over the place. 20% on a few 3% on a couple, 15% Not pushing this CPU at all
Most are being utilized but not by much

 

[er557]

First try with MSI AB, sorry about that, definitely the fonts need to be larger.

 

[EarthDog]

Ok, watched it at my PC.

First, if I run the vid at my monitor's native res (2560x1440), I can still barely read that tiny font... colors mix with the background as well. Very difficult to see... hopefully it is just me.

Second...

What do all the numbers mean? What is GPU use.....The first set of % on the left or the ones on the right? Both values seem to vary wildly. At 4K with 2x 1080Ti's I would expect GPU use to be pegged constantly at 99%(ish). Instead it is all over the map with an average of around 45 fps FPS (you are modded, I know - but that tells me the GPUs should be loaded more). Those Xeon's aren't driving those cards properly. I'd bet good money says a 9900K or 3700x would get a lot more than what you are seeing. I don't think GTA V scales past 6 cores and threads?

The AnandTech Coffee Lake Review: Initial Numbers on the Core i7-8700K and Core i5-8400

www.anandtech.com

That OSD is a nightmare to figure out with all those cores, lol. The top half is grouped with multiple columns, while another 18 or w/e are single column... lol. I can't make heads or tails of it all. lol.

 

[Jetster]

 

[er557]

Ok, watched it at my PC.

First, if I run the vid at my monitor's native res (2560x1440), I can still barely read that tiny font... colors mix with the background as well. Very difficult to see... hopefully it is just me.

Second...

What do all the numbers mean? What is GPU use.....The first set of % on the left or the ones on the right? Both values seem to vary wildly. At 4K with 2x 1080Ti's I would expect GPU use to be pegged constantly at 99%(ish). Instead it is all over the map with an average of around 45 fps FPS (you are modded, I know - but that tells me the GPUs should be loaded more). Those Xeon's aren't driving those cards properly. I'd bet good money says a 9900K or 3700x would get a lot more than what you are seeing. I don't think GTA V scales past 6 cores and threads?

The AnandTech Coffee Lake Review: Initial Numbers on the Core i7-8700K and Core i5-8400

www.anandtech.com

That OSD is a nightmare to figure out with all those cores, lol. The top half is grouped with multiple columns, while another 18 or w/e are single column... lol. I can't make heads or tails of it all. lol.

The columns represent groups of 3 cores in the first 18, then another 16 at the bottom, I got a cfg file from a 7980xe layout...
As to the usage, it is mediocre for each core, but most cores are being used and no stress on the system or stuttering from pegging cores to 100%. I think the percentages on the top are fan speed percentages, and the proof of gpu usage is the cards go to 85c/75c for top/bottom card, which is showing little case airflow(they had to be placed adjacent). So temp wise they do stress.

As to other conclusions, I put heavy impact on gpu settings @4k 60hz, so I dont need a 9900k pushing 100 fps, I always stress the gpu @ ultra settings, my style is gpu bound. So I dont mind if the cpu is snapping it's thumbs.


edit:

BTW, 3.5ghz on the cores is enough for todays gaming? I dont think the IPC itself changed much since haswell?




I hijacked my friend's from another forum video, showing his 18 cores in use in an uncapped AAA title, so that's how his scaling looks like

 

[EarthDog]

Not sure if that is it. The fan speeds were not dropping down proportionately with the percentages though and both were different. Not to mention fans don't instantly ramp down and back up in 2 seconds.

Most of the cores weren't tickled and we can see the threads spin through them. It would be even better to see a graphical representation like from task manager in W10 with a core breakdown.

Faster CPUs would increase 1% FPS quite a bit. You'd reach your 60 FPS goal average. It doesn't look like you are hitting it now in most cases.

 

[Deleted member 185158]

edit:

BTW, 3.5ghz on the cores is enough for todays gaming? I dont think the IPC itself changed much since haswell?

Most desktop CPUs are release with a turbo boost into the 4.5ghz range (average) and a lot of Intel do 5ghz quite easily.

I think you'd see significant gains with a 9900K at 5.2ghz actually.

 

[EarthDog]

BTW, 3.5ghz on the cores is enough for todays gaming? I dont think the IPC itself changed much since haswell?

That is a haswell based unit? I thought it was Nehalem/X58... the Intel Ark doesn't have much info on it.

But I agree with Shrimp... 3.5 isn't enough to push those GPUs. SLI... and SLI of high end cards, really needs core clocks. I feel the gain would allow you to reach that 60 FPS average.

 

[er557]

all other less heavily modded titles produce great fps for me...
and btw that is haswell ep, this system is exactly i7 5930k multiplied by six

 

[The Egg]

GTA V was originally released in 2013 for PS3/360. It was slightly updated for PC, but it’s still an old game, running on an old engine with known unpatched bugs/quirks. Whatever can be learned here doesn’t necessarily extrapolate to anything other than GTA V.

 

[EarthDog]

all other less heavily modded titles produce great fps for me...
and btw that is haswell ep, this system is exactly i7 5930k multiplied by six

Gotcha. Still, 3.5 GHz for SLI 1080 Ti's is holding them back.

Anyway, thanks for the info.

 

[er557]

gtx 1080 vanilla, not ti

 

[EarthDog]

gtx 1080 vanilla, not ti

That doesn't change anything about the bottleneck though. You can see GPU use is all over the map.

Thanks again!

 

[er557]

sure, still i wonder if the gpu was not fed properly, the cores would be pegged 100%, not relaxed like now.

so the stress seems to be on the gpus, appropriately, heating them up. i will check in AB graph the gpu usage itself, and maybe related to the game itself or to fast sli vsync setting. what remains to see on YT is, does anyone with dual 1080 get better fps wth these extreme game settings and mod.

 

[EarthDog]

sure, still i wonder if the gpu was not fed properly, the cores would be pegged 100%, not relaxed like now.

That isnt always true.

what remains to see on YT is, does anyone with dual 1080 get better fps wth these extreme game settings and mod.

Good luck with that. One problem is being able to add the same mods as you and run the same settings. Additionally, you completed a manual run through instead of using the integrated benchmark which is repeatable testing.

So like someone said, this doesnt extrapolate out past your own manual run with mods and settings since it cant be replicated easily.

 

[er557]

Ok, here are some screenshots showing my in game settings, and a graph from msi AB showing gpu usage of both gpus in game and power usage of each gpu, and those are pegged just fine while running the game. Also, I found out that disabling msi ab altogether, and closing corsair link monitoring, while avoiding from recording videos in game, completely resolved the stutter/crash after 5 mins, and the fps is much smoother and fast in a substanial way.

 

[EarthDog]

You are killing us with your information and it being so hard to read... lol. The screenshot you put up also proves my point.

Again, you are heavily modded... so while the in-game settings are one thing, the mods are another. Again as well, you did a MANUAL RUN THROUGH instead of running the repeatable integrated benchmark... we cannot compare like to like very easily at all. Move on from the thought of others being able to replicate this.

As far as GPU use... that isn't pegged man. Please check again... I see use varied throughout the run... I used arrows to highlight the ONE part that was 'pegged'. Now compare that to my screenshot. Notice how after the last dip it is much more flat than yours? both GPU use and power? Just to note, prior to that flat part, my settings were turned down as I play the game competitively (PUBG 2560x1440) and there was the minute load/lobby. Once the settings were turned back up, it was pegged and a lot more flat than yours. Your settings are cranked, plus mods, plus 4K UHD, and your GPUs aren't consistently running close to 100% like mine and only getting 30-60 fps... what do you think the issue is if it isn't the lower clocks of your CPU? Also note, my system is running with a 4.4 GHz 7960x running 16c/16t.

 

[MyTechAddiction]

Above all everyone already said YOU ARE THE WORST DRIVER EVER.

 

[er557]

You are killing us with your information and it being so hard to read... lol. The screenshot you put up also proves my point.

Again, you are heavily modded... so while the in-game settings are one thing, the mods are another. Again as well, you did a MANUAL RUN THROUGH instead of running the repeatable integrated benchmark... we cannot compare like to like very easily at all. Move on from the thought of others being able to replicate this.

As far as GPU use... that isn't pegged man. Please check again... I see use varied throughout the run... I used arrows to highlight the ONE part that was 'pegged'. Now compare that to my screenshot. Notice how after the last dip it is much more flat than yours? both GPU use and power? Just to note, prior to that flat part, my settings were turned down as I play the game competitively (PUBG 2560x1440) and there was the minute load/lobby. Once the settings were turned back up, it was pegged and a lot more flat than yours. Your settings are cranked, plus mods, plus 4K UHD, and your GPUs aren't consistently running close to 100% like mine and only getting 30-60 fps... what do you think the issue is if it isn't the lower clocks of your CPU? Also note, my system is running with a 4.4 GHz 7960x running 16c/16t.

View attachment 138435View attachment 138436

Just to check for this issue, I will run time spy 4k , and monitor the same values. Will that be a good starting point?


EDIT: @ mod: I edited my post for pertinent data, and removed the other stuff, so I will be glad if it is back online