How many CPU cores do I need?

[RasmusP]

Thanks for the detailed post and questions!

As part of my GPU upgrade, I went from 180 W to 330 W adapter. The extra 150 W may all go to the GPU, I'll need to see how that goes...

They're about $100 on eBay. That's really why I asked the original question. It seems like going from two cores to four would be worth it to me if we can make it work.

Yes, I'm going to test before and after the new GPU install, with stress tests and while racing, to see how much power I have left for a new CPU.

The chipset is either H87 or Z87. I'm on vacation so I can't check. I'm not planning on overclocking because of power and heat limitations. The thermal management in the case is pretty good, but I don't want to push it, especially with the extra heat from the GPU.

I hear what you're saying about building a new machine with the good parts I have. However, part of the fun for me is tinkering with parts and getting the most out of what I have. I'm very pleased by the participation in this thread. Let's see how much we can soup up this old beater...

All this sounds like you should just grab the 4th Gen i7 for a good price (if the k ist much more, go with that as you'd get the higher automatic boost) and get happy for a bit more time.
It will definitely improve your performance. Maybe not by that much and depending on the perspective, not worth the money but it seems that you wanna keep your case etc so yeah, you'll have a nice boost in performance but not be blown away by it.
A cheap 9600k(f) build would feel like going from a bicycle to a sportscar but the 4th Gen i7 would at least be a decent little car to move from A to B without getting wet in the rain

 

[John-Eric Saxén]

For what it's worth...I just upgraded my old system as I was holding out as long as I could but it just struggled with all the new titles.
I5 3570k @ 4.5ghz with a gtx 970...the CPU ran at 100% in AC, ACC and pc2.
I found some deals and was able to upgrade to a i5 9600kf with msi budget board for $350 cad..$260 USD and was able to reuse everything from my old build.
Personally I'd save a $100-200 and wait for a deal and upgrade to something newer, cause you'll just be in the same situation in a yrs time upgrading to used i7 4790

i5-3570k is running DDR3 RAM as is OP's 4th gen i3, so you need to factor in RAM price in upgrade, or if you already did I can't see how you can get i5-9600kf + board + RAM for $260 when MSRP for that CPU alone is $200.

 

[Kurupt CDN]

i5-3570k is running DDR3 RAM as is OP's 4th gen i3, so you need to factor in RAM price in upgrade, or if you already did I can't see how you can get i5-9600kf + board + RAM for $260 when MSRP for that CPU alone is $200.

Price is in Canadian, I just converted it to USD with the current currency exchange nor did I factor in the 5%sales tax.
The parts were bought at Memory Express in Calgary a month ago
CPU was $255
Msi A-pro board was $100
I didn't factor in Ram to the above cost but it cost me $79 for a set of 16gb 2666mhz Corsair ram...usually $99 on Amazon.
I also grabbed a NVMe M.2 Western digital 1tb card for $125
All together I paid just under $550, re used my 240gb SSD for windows, Corsair gamer 850 psu, 2tb hhd drive, Corsair h80 water cooler and case

This week they have AMD stuff on sale.
You can grab a r5 3600x for $270.

Processors at Memory Express - Memory Express Inc.

www.memoryexpress.com

I've notice they will run an AMD sale for a week or two than do a Intel sale afterwards.
They also do online sales and delivery, just not sure if they do international?

 

[dud]

If somebody has plenty of cores and wants to help the community it would be nice to disable more and more cores in the BIOS and see what that does.

 

[RobertR1]

If somebody has plenty of cores and wants to help the community it would be nice to disable more and more cores in the BIOS and see what that does.

Already been done:

Test: CPU Core count and RAM scaling in ACC, AC1 and R3E

Hello guys Couldn't find decent data so I did these tests myself during the holidays. Max settings @ 720P (Max distances, Max shadows, Max mirrors etc) to push for the max cpu drawcalls. Heavy duty multiplayer race replays were used, with 20~30 cars. 1 entire lap per test. With the camera set...

www.racedepartment.com

 

[RasmusP]

Already been done:

Test: CPU Core count and RAM scaling in ACC, AC1 and R3E

Hello guys Couldn't find decent data so I did these tests myself during the holidays. Max settings @ 720P (Max distances, Max shadows, Max mirrors etc) to push for the max cpu drawcalls. Heavy duty multiplayer race replays were used, with 20~30 cars. 1 entire lap per test. With the camera set...

www.racedepartment.com

Thanks for posting this so I didn't have to

One crucial thing for this thread is missing though:
2 cores vs 2 cores+ HT vs 4 cores

It's crucial because the question is how much HT improves the performance of 2 cores in comparison to 4 real cores.
We can see in that thread, that 3 cores are already a lot closer to 4 cores and more than it is close to 2 cores.

I'll do this test throughout the next days

 

[RobertR1]

Thanks for posting this so I didn't have to

One crucial thing for this thread is missing though:
2 cores vs 2 cores+ HT vs 4 cores

It's crucial because the question is how much HT improves the performance of 2 cores in comparison to 4 real cores.
We can see in that thread, that 3 cores are already a lot closer to 4 cores and more than it is close to 2 cores.

I'll do this test throughout the next days

Once you're down to 2 cores, it's virtually a point of irrelevance as you'll have basic windows processes competing for resources and that's before you run any background apps (TS, discord, A/V, etc).

 

[RasmusP]

Once you're down to 2 cores, it's virtually a point of irrelevance as you'll have basic windows processes competing for resources and that's before you run any background apps (TS, discord, A/V, etc).

I agree. Still, for the creator of this thread, the needed test would be 2 cores + HT vs 4 real cores (and 4 cores + HT).
He has a hyperthreading i3 and wants to spend money on an i5 or i7 of the same generation.
Theres no benchmark for 1-3 thread games anywhere (I had a look).

From my view the question is whether or not having 4 virtual cores is already enough to maintain enough headroom for TS, Discord etc. or not.

 

[Terry Rock]

I'll just leave this here....
This is just with me running rFactor2 laps...nothing else.

 

[Trebormoore84]

I'll just leave this here....
This is just with me running rFactor2 laps...nothing else.

Thanks, for reference what Resolution and fps are you running here? Also were you hot lapping solo? Or in a race Ai or otherwise?


.

 

[Terry Rock]

1920 x 1080@ 60 Hz....hot-lapping solo ...every setting max in game.

 

[2stains]

Do the same with ACC at sunset at Bathurst! But full grid.

 

[Terry Rock]

Do the same with ACC at sunset at Bathurst! But full grid.

 

[Trebormoore84]

Nice, I think RF 2 does more multithreading when you race with a full grid compared to solo. I dont know why exactly but I think its the shadows which might explain why its higher when you race around other cars compared to not.

This was bothering me after my last posts so I asked this question was told on the s397 forum RF2 uses 2 cores for single threaded loads which are apparently gpu draw calls (all graphics except shadows) and ai on one, and then physics on the other. Could be wrong though so don't quote me.

The graphics one goes up when you increase the graphics settings, but they both scale up together dramatically when you increase the fps above 60, on mine one is loaded at virtually 100% and the other between 60 and 90% as I essentially uncap the fps (200fps cap with these monitors) and increase graphics settings until the gpu is the bottleneck again and settle at a range between 100-160. This is fine for me as long as that loaded core doesnt clash with anything else and why i use process lasso. Makes sense for people running high fps setups.

Only things I was told are multithreaded are shadow draw calls (which was a s397 optimisation to help people with VR and triples), and the FFB which varies in load depending on the wheel you use. I use a high frequency OSW and coupled with the high fps I actually see big spikes over the multithreads when there is a lot of input and force feedback.

You can see that ACC has one core higher than the rest which must be the graphics again, but the other threads are loaded a lot more so is far more efficient at multithreading.

I'd be interested in seeing how PC2 and AMS2 scales over multithreads, that engine seems quite efficient.

 

[David Mann]

I'd like to thank everyone for their contributions and discussion. This has turned into a good thread! In case you're interested, I posted some benchmark results from Phase 1 of my upgrade (GPU upgrade). The CPU upgrade is yet to come. My plan is to upgrade to the i7-4790K.

Potato upgrade, Part 1 (GPU)

In a previous thread (see below here), I asked about upgrading the CPU on my old Alienware X51 R2. I'm going to get to that eventually, but before that, I'd like to tell you about the GPU upgrade that I just did. I'd also like to thank everyone who contributed to the other thread for the...

www.racedepartment.com