Best CPU available: AMD Ryzen 9 7950X3D?

[azesmbog]

Very interesting that the game uses 125GB when loading if it actually has that much memory available. No wonder it took me 20 minutes to load and plenty of swap.

Please check this map, the results for processors are very interesting, and if you make a rating at current prices, it will turn out even more interesting
https://factoriobox.1au.us/results/cpus ... =1.0.0&vh=

[Bloodred217]

Here's that map, I got 48 UPS: https://factoriobox.1au.us/result/d43fd ... f485ece27e
I also ran it in a Windows VM on another computer I have with a 5950X (the VM was the only one set up to run Factorio), I got 29UPS there. Makes sense with a 5800X3D getting 41. I don't have a link for this one since the script didn't like running in a VM I guess, it says it couldn't parse the results.

The results look very strange to me, slowest by far is a Zen 3 CPU, albeit a mobile version it's still pretty new so it shouldn't be THAT slow. Old i5 6400 faster than a 12700K and 11700F, that 7800X3D is slightly slower than a 5800X3D and a fair bit behind my 7950X3D even though they should be almost the same. The 7800X3D seems to be using very similar memory to mine as well. What's up with this map? None of those results make sense to me.

[azesmbog]

I added my result with 11600K to make it even more interesting

[Bloodred217]

I see a 30 UPS and 2 40 UPS results for the 11600k, big differences once again, same CPU this time too. I wonder if it's just different RAM settings, or what? One CPU being 33% faster than another when they're the very same model is not something you usually see in benchmarks.

[azesmbog]

My results are just twice 40.
30 is not my result, it's someone who can't tune their computer
The results of the i5-6400 (35 UPS) are also mine, correct.
Some modern processors should be ashamed to play Factorio with a bunch of processor cores and low UPS
Life is pain.

[azesmbog]

In order not to make new threads, let it be here, maybe someone is interested.
A little comparison of two rather old 11th generation Intel processors.
Both processors were installed in turn on the same motherboard
with unchanged parameters in the bios:

Core Speed 4944 MHz
128 Gb memory DDR4 3498 MHz 16-20-20-38

Therefore, the difference between them was only in the number of cores and the amount of cache memory:

11th Gen Intel Core i5-11600K
6 cores / 12 threads
Cache memory 12 MB Intel® Smart Cache
Release Date Q1'21
vs
11th Gen Intel Core i7-11700K
8 cores / 16 threads
Cache memory 16 MB Intel® Smart Cache
Release Date Q1'21
vs
6th Gen Intel® Core™ i5-6400
4 cores / 4 threads
Basic CPU clock frequency 2.70 GHz @ 4.5 GHz
Cache memory 6 MB Intel® Smart Cache
Release Date Q3'15

Here are the results of testing on the most difficult maps in Factorbox:

--------------------------------------------------------------
227 UPS
Map: azesmbog - 10000x10000 DW marathon
11th Gen Intel Core i5-11600K

241 UPS
Map: azesmbog - 10000x10000 DW marathon
11th Gen Intel Core i7-11700K

217 UPS
Map: azesmbog - 10000x10000 DW marathon
Intel Core i5-6400 CPU
--------------------------------------------------------------

201 UPS
Map: flame_Sla 10k - 10x1000spm Belt Module
11th Gen Intel Core i5-11600K

208 UPS
Map: flame_Sla 10k - 10x1000spm Belt Module
11th Gen Intel Core i7-11700K

186 UPS
Map: flame_Sla 10k - 10x1000spm Belt Module
Intel Core i5-6400 CPU
--------------------------------------------------------------
69 UPS
Map: Flame_Sla's 30k SPM belt map (30x1000SPM) v3
11th Gen Intel Core i5-11600K

70 UPS
Map: Flame_Sla's 30k SPM belt map (30x1000SPM) v3
11th Gen Intel Core i7-11700K

62 UPS
Map: Flame_Sla's 30k SPM belt map (30x1000SPM) v3
Intel Core i5-6400 CPU
--------------------------------------------------------------
63 UPS
Map: High UPS 40k cell base
11th Gen Intel Core i5-11600K

65 UPS
Map: High UPS 40k cell base
11th Gen Intel Core i7-11700K

57 UPS
Map: High UPS 40k cell base
Intel Core i5-6400 CPU
--------------------------------------------------------------
58 UPS
Map: Smurphy1 40.5k monolithic belt base
11th Gen Intel Core i5-11600K

60 UPS
Map: Smurphy1 40.5k monolithic belt base
11th Gen Intel Core i7-11700K

53 UPS
Map: Smurphy1 40.5k monolithic belt base
Intel Core i5-6400 CPU
--------------------------------------------------------------
40 UPS
Map: 2023-04-13 1498400 rockets 10000spm new blueprints for lds blues reds
11th Gen Intel Core i5-11600K

41 UPS
Map: 2023-04-13 1498400 rockets 10000spm new blueprints for lds blues reds
11th Gen Intel Core i7-11700K

36 UPS
Map: 2023-04-13 1498400 rockets 10000spm new blueprints for lds blues reds
Intel Core i5-6400 CPU
--------------------------------------------------------------
40 UPS
Map: 712000 rawket lawnchairs 10000spm 260 trains
11th Gen Intel Core i5-11600K

41 UPS
Map: 712000 rawket lawnchairs 10000spm 260 trains
11th Gen Intel Core i7-11700K

36 UPS
Map: 712000 rawket lawnchairs 10000spm 260 trains
Intel Core i5-6400 CPU
--------------------------------------------------------------
37 UPS
Map: napo 10000spm 10000city blocks switching out reactors 1,1 million rockets
11th Gen Intel Core i5-11600K

39 UPS
Map: napo 10000spm 10000city blocks switching out reactors 1,1 million rockets
11th Gen Intel Core i7-11700K

34 UPS
Map: napo 10000spm 10000city blocks switching out reactors 1,1 million rockets
Intel Core i5-6400 CPU
--------------------------------------------------------------
34 UPS
Map: 475475 rockets - 10k stable at 6800 city blocks 100x60
11th Gen Intel Core i5-11600K

34 UPS
Map: 475475 rockets - 10k stable at 6800 city blocks 100x60
11th Gen Intel Core i7-11700K

31 UPS
Map: 475475 rockets - 10k stable at 6800 city blocks 100x60
Intel Core i5-6400 CPU
--------------------------------------------------------------

What can I say from the test results.
At more "light" and simple map difference in the amount of cache is still important, but the "heavier" map - there is more of this difference is leveled.

upd:
Also out of competition added i5-6400 "non K" - officially not overclocked ancient processor from eight years ago, for comparison We do not pay attention to the results with it (especially not every modern processor can achieve such results

Thank you for your attention

[datarza]

azesmbog, your tests once again proves that calcus in Factorio does not depends on CPU speed or number of cores. Primary depends on memory speed, secondary depends on CPU cache. I just curious, it you can also post `task manager` screens with `Performance` page on `CPU` for i5-6400 on big Factorio map. Something tells me that CPU's loading will be around 75%.

At more "light" and simple map difference in the amount of cache is still important, but the "heavier" map - there is more of this difference is leveled.

It does not matter how many cache the CPU has, during a long playing game it is possible to build a base that is bigger than the CPU cache

[datarza]

It doesn't really matter what CPU or memory you have. In Factorio with the default map settings without surprizing blueperints adopted for Factorio's calculation, the maximum that can be achieved will be 3500 trains and 10k sciense packs at 5UPS +/- 2UPS (depends on CPU+Memory).

[TheKillerChicken]

I have a ryzen 9 7950x with 64GB DDR5-5200. But there is a caveat, I do ultra-extreme things to factorio like 90 trillion native bugs, a radar that can reveal 128 chunks, and modified modules that can create 100k science/minute, and miners that can harvest ~1 million ore/minute with the fact that I also process 500 million fluids/second. So I bring this game to its' knees upon this extreme processing.. Usually hammering all the bugs will speed me up again.