Benchmark Testing of Spaced out

[Billio]

I just received an AMD 5800x3D chip and decided to benchmark it compared to my previous CPU and thought I'd add a new data point here. First up is the old CPU for comparison. [Edit] Fixed error in RAM timings [16-16-16-36 -> 16-16-16-32].

Time: 122 seconds

CPU: AMD Ryzen5 3600X 

RAM: 32GB RAM (2x16) @ 3800 MT/s

RAM Timings: 16-16-16-32

GPU: Radeon 6900xt

 

Swapping the new system yields the following results:

Time: 102s @ ~25 FPS

CPU: AMD Ryzen 7 5800x 3D

RAM: 32GB RAM (2x16) @ 3800 MT/s

RAM Timings: 16-16-16-32

GPU: Radeon 6900xt

[Saturnus]

6 minutes ago, Billio said:

I just received an AMD 5800x3D chip and decided to benchmark it compared to my previous CPU and thought I'd add a new data point here. First up is the old CPU for comparison.

Time: 122 seconds

CPU: AMD Ryzen5 3600X 

RAM: 32GB RAM (2x16) @ 3800 MT/s

RAM Timings: 16-16-16-36 timings

GPU: Radeon 6900xt

 

Swapping the new system yields the following results:

Time: 102s @ ~25 FPS

CPU: AMD Ryzen 7 5800x 3D

RAM: 32GB RAM (2x16) @ 3800 MT/s

RAM Timings: 16-16-16-36 timings

GPU: Radeon 6900xt

If I'm not mistaken that a record.

[Billio]

I don't think I tracked the time closely enough to qualify it as a record since my reaction time on the stop watch would add some variance, but I ran 2 tests that clocked in at approximately 102s and 103s respectively so I'm fairly confident the result isn't an anomaly.

[Edit] As was mentioned in other benchmark runs, I notice occasional periods where the entire game loop freezes for up to a second or 2, and I think these hitches probably would prevent any CPU from getting a perfect score.

One thing I'm not sure about is whether any game updates may have contributed to overall runtime since even the score for my old system seemed to be on the higher end for that CPU.

Would love to see more data points, but its good to know that the 5800x 3d seems to do well here as in other simulation games (since that's the main reason I decided to grab it).

 

[Edit] One more addition after a quick and dirty increase in memory speed.

Time 103s @ ~25 FPS

CPU AMD Ryzen 7 5800x 3D

RAM 32GB (2x16) @ 4000 MT/s

RAM Timings: 16-16-16-36

GPU: 6900XT

It appears, as in other benchmarks of the 5800x3d that memory speed/bandwidth is less important on this chip because the large L3 is largely whats responsible for the increased performance.

[Saturnus]

4 hours ago, Billio said:

[Edit] One more addition after a quick and dirty increase in memory speed.

As far as I can tell from reviews is that 3800 is the maximum MT/s setting you should attempt with the 5800X3D.
To increase performance beyond that you have to tweak memory timing. The easiest and safest way to do that is to use the DRAM Calcuator but note that voltage settings are somewhat optimistic unless you have a really high-end motherboard. I usually allow 0.05V more than the DRAM Calc suggests to get it fully stable in all use cases.

[Billio]

41 minutes ago, Saturnus said:

As far as I can tell from reviews is that 3800 is the maximum MT/s setting you should attempt with the 5800X3D.
To increase performance beyond that you have to tweak memory timing. The easiest and safest way to do that is to use the DRAM Calcuator but note that voltage settings are somewhat optimistic unless you have a really high-end motherboard. I usually allow 0.05V more than the DRAM Calc suggests to get it fully stable in all use cases.

Previously I ran 3800 MT/s on my 3600x because I couldn't boost my fclk higher than 1900 MHz (and that should match ram speed for best performance). I assumed the same would be true of the 5800X3D, but just for kicks I tried boosting the fclk to 2 GHz and it worked, this allowed me to boost my DRAM to 4000 MT/s without a penalty and I was able to keep the same timings as 3800.

That aside, you're likely right in that the raw memory bandwidth just isn't going to do much to uplift performance (at least in the cases that allow the 5800x3d to shine). I think my RAM is a decent bin of B die, so I may try for c14 timings and see where that takes me. Thanks for the discussion!

[gabberworld]

13 hours ago, Billio said:

I don't think I tracked the time closely enough to qualify it as a record since my reaction time on the stop watch would add some variance, but I ran 2 tests that clocked in at approximately 102s and 103s respectively so I'm fairly confident the result isn't an anomaly.

[Edit] As was mentioned in other benchmark runs, I notice occasional periods where the entire game loop freezes for up to a second or 2, and I think these hitches probably would prevent any CPU from getting a perfect score.

One thing I'm not sure about is whether any game updates may have contributed to overall runtime since even the score for my old system seemed to be on the higher end for that CPU.

Would love to see more data points, but its good to know that the 5800x 3d seems to do well here as in other simulation games (since that's the main reason I decided to grab it).

 

[Edit] One more addition after a quick and dirty increase in memory speed.

Time 103s @ ~25 FPS

CPU AMD Ryzen 7 5800x 3D

RAM 32GB (2x16) @ 4000 MT/s

RAM Timings: 16-16-16-36

GPU: 6900XT

It appears, as in other benchmarks of the 5800x3d that memory speed/bandwidth is less important on this chip because the large L3 is largely whats responsible for the increased performance.

thanks for info. at least your test shows that its not better than intel 12 series cpu

yes, game does have freeze issue, soo times jump somewhere 102-103 sec,  

with this mod [BETA] Fast Track - Performance Mod for Oxygen Not Included - Page 3 - [Oxygen Not Included] - Mods and Tools - Klei Entertainment Forums 

you should able hit 100 sec, it minimizes that freezing issue

 

[Saturnus]

12 hours ago, Billio said:

That aside, you're likely right in that the raw memory bandwidth just isn't going to do much to uplift performance (at least in the cases that allow the 5800x3d to shine). I think my RAM is a decent bin of B die, so I may try for c14 timings and see where that takes me. Thanks for the discussion!

Yeah, AMD already proved with Infinity Cache on their graphics cards that having an extremely large cache basically eliminates the need for ultra high bandwidth. Memory timings will still be somewhat relevant but it means that one doesn't have to invest in fast memory to get at least most of the benefits from the CPU upgrade.
Furthermore, other tests have shown that since the 5800X3D runs at significantly lower voltage and uses significantly lower power than the standard 5800X, you also don't have to pair it with a higher end motherboard. Most low midrange and up B550 boards is fine, or even a good quality B450. It's the perfect upgrade CPU to almost anyone that already have an AMD system, and plays sim games.

[Billio]

4 hours ago, Saturnus said:

Yeah, AMD already proved with Infinity Cache on their graphics cards that having an extremely large cache basically eliminates the need for ultra high bandwidth. Memory timings will still be somewhat relevant but it means that one doesn't have to invest in fast memory to get at least most of the benefits from the CPU upgrade.
Furthermore, other tests have shown that since the 5800X3D runs at significantly lower voltage and uses significantly lower power than the standard 5800X, you also don't have to pair it with a higher end motherboard. Most low midrange and up B550 boards is fine, or even a good quality B450. It's the perfect upgrade CPU to almost anyone that already have an AMD system, and plays sim games.

My thoughts exactly! It would actually be interesting to get a tougher version of this map to benchmark on to see if the 5800x 3D ties with or could actually exceed the 12900k/s since it looks like the higher end CPUs are bumping up against the limits of this test.

[gabberworld]

1 hour ago, Billio said:

My thoughts exactly! It would actually be interesting to get a tougher version of this map to benchmark on to see if the 5800x 3D ties with or could actually exceed the 12900k/s since it looks like the higher end CPUs are bumping up against the limits of this test.

limit is 100 sec soo yes

[Nedix]

02:02:50 = 122s

CPU  = Intel I5 11500

RAM = 2 x 8GB DDR4-2133 (XMP 2.0)

RAM timings = 14-14-35-50

GPU = Asus GeForce GTX970

HDD = Samsung 970 EVO Plus NVMe™ M.2 SSD - 1 TB

 

PS: I still need to upgrade the RAM and GPU

[gabberworld]

2 hours ago, Nedix said:

02:02:50 = 122s

CPU  = Intel I5 11500

RAM = 2 x 8GB DDR4-2133 (XMP 2.0)

RAM timings = 14-14-35-50

GPU = Asus GeForce GTX970

HDD = Samsung 970 EVO Plus NVMe™ M.2 SSD - 1 TB

 

PS: I still need to upgrade the RAM and GPU

yeah, by default 2133 not good ddr4 

[NardoLoopa]

02:31 = 151s

CPU  = Intel i5 6600 3.3GHz

RAM = Corsair Vengeance LPX 8 GB (2 x 4 GB) DDR4-2133 CL15 (Non-ECC)

RAM timings = 15-15-15-36

GPU = Gigabyte GeForce RTX 2060 6GB OC Pro

HDD = WD_BLACK SN750 2 TB M.2-2280 NVME (Game Files)

Samsung 850 Evo 500 GB 2.5" SSD (OS / SaveGame file)

OS = Windows10

 

[Raukgorth]

100 seconds is not the limit of this test.

When we are getting 100 seconds, we can start looking at average FPS and score based on that.

[gabberworld]

30 minutes ago, Raukgorth said:

100 seconds is not the limit of this test.

When we are getting 100 seconds, we can start looking at average FPS and score based on that.

well yes that is true but for time that test is for100 sec

problem with fps is that its jump too much, scroll to some other location you may get totally differed results 

[fritozipper]

New high(ist)score!!

Surface pro 3

FPS ≤1 

Time : 15:55 or 955 seconds 
CPU : i7-4650U  2.3GHz
Graphics card : intel HD 5000
RAM : 8GB 1600 MHz

CAS : 12 15 15 34
HDD : 256gb samsung ssd

[Bazlish]

CPU  = Intel i7 10700KF @ stock 

RAM = 16GB DDR4 CL16 @ 3600Mhz

GPU = RTX 3090 

HDD = Kingston M.2 1TB. 

Thought it would be a good way to test the new "Performance" Patch coming out also that is in the beta branch currently, also wanted to see if ultrawide resolutions had any impact.

Current stable release patch times: @ 3440 x 1440: 113 seconds @ 2560 x 1440: 111 seconds 

"Fast Friends" Beta patch: @ 3440 x 1440: 104 seconds @ 2560 x 1440: 102.5 seconds. 

So the new patch seems to defintely have a nice lift in performance, the guys with the 12th Gen Intel with DDR5 and the 58xx3D's should be seeing 100 seconds now I'd say on the new patch.

[gabberworld]

1 hour ago, Bazlish said:

So the new patch seems to defintely have a nice lift in performance, the guys with the 12th Gen Intel with DDR5 and the 58xx3D's should be seeing 100 seconds now I'd say on the new patch.

i hope soo because that freezing bug is main issue currently for hit 100 sec, if they fixed that then yes

that is huge performance jump tho, from 111 sec to the 102 sec. that what i call they actually change something big at game

[Bazlish]

4 hours ago, gabberworld said:

i hope soo because that freezing bug is main issue currently for hit 100 sec, if they fixed that then yes

that is huge performance jump tho, from 111 sec to the 102 sec. that what i call they actually change something big at game

Yeah well I think that is roughly an 8% increase? I have a somewhat basic laptop that has an 11th gen I5 & no dedicated gpu, 4c/8t I'm thinking about testing it on also, I'm imagining the performance increases on slower systems will probably be greater

 

[cpy]

7 hours ago, Bazlish said:

CPU  = Intel i7 10700KF @ stock 

RAM = 16GB DDR4 CL16 @ 3600Mhz

GPU = RTX 3090 

HDD = Kingston M.2 1TB. 

Thought it would be a good way to test the new "Performance" Patch coming out also that is in the beta branch currently, also wanted to see if ultrawide resolutions had any impact.

Current stable release patch times: @ 3440 x 1440: 113 seconds @ 2560 x 1440: 111 seconds 

"Fast Friends" Beta patch: @ 3440 x 1440: 104 seconds @ 2560 x 1440: 102.5 seconds. 

So the new patch seems to defintely have a nice lift in performance, the guys with the 12th Gen Intel with DDR5 and the 58xx3D's should be seeing 100 seconds now I'd say on the new patch.

Can you also test this against fast track mod?

[Bazlish]

2 hours ago, cpy said:

Can you also test this against fast track mod?

Hey yeah just finsihed it then. 
@ 3440 x 1440 100.6 @ 2560 x 1440 100.3

[Billio]

On 5/7/2022 at 11:29 AM, gabberworld said:

well yes that is true but for time that test is for100 sec

problem with fps is that its jump too much, scroll to some other location you may get totally differed results 

The other problem is that while FPS can fluctuate it doesn't really have much of an impact in this game. Losing simulation time on the other hand actually translates into wasting time playing the game on late/large colonies because it cannot run at full speed.

That said I'm sure that FPS is probably a reasonable proxy for this within a given system (i.e. once FPS goes below X on a given configuration you can be fairly certain that you're also losing sim time).

On 5/9/2022 at 3:02 AM, gabberworld said:

i hope soo because that freezing bug is main issue currently for hit 100 sec, if they fixed that then yes

that is huge performance jump tho, from 111 sec to the 102 sec. that what i call they actually change something big at game

I just tried re-running the test save and I still get occasional hitching. I'm still right at that 102-103s barrier due to that. I did notice that FPS increased relative to what i documented previously: going from ~25 FPS to 35 FPS or so (except for the hitches). This was all done at 3440x1440 so I might've seen slightly better times if I were further away from the limit.

I hope Klei can work out what it is in the mod that fixes the hitching and work to implement it, if the mod is successful in doing that.

[cpy]

For me losing simulation time is not as bad as framerate going apeshift crazy! I'd rather have FPS to be completely decoupled from any game function. So you can plan without freeze and stutters.

I'd rather have UI run at 30-60FPS than going 0-200 all the time.

[Bazlish]

10 hours ago, Billio said:

The other problem is that while FPS can fluctuate it doesn't really have much of an impact in this game. Losing simulation time on the other hand actually translates into wasting time playing the game on late/large colonies because it cannot run at full speed.

That said I'm sure that FPS is probably a reasonable proxy for this within a given system (i.e. once FPS goes below X on a given configuration you can be fairly certain that you're also losing sim time).

I just tried re-running the test save and I still get occasional hitching. I'm still right at that 102-103s barrier due to that. I did notice that FPS increased relative to what i documented previously: going from ~25 FPS to 35 FPS or so (except for the hitches). This was all done at 3440x1440 so I might've seen slightly better times if I were further away from the limit.

I hope Klei can work out what it is in the mod that fixes the hitching and work to implement it, if the mod is successful in doing that.

Yeah so from what I understand the game reduces its calculations when you are below 30fps, the fact you're now back above 30fps and still the same time... Your game is now running at 100% calculations + a better frame rate with still just that little freeze.
I did notice when testing after a few times in a row that the freeze was present almost every run, but if I restarted my game the first few runs it was gone.

[magei]

Just decided to try this my self, Just finished putting my system together last week, I didn't reinstall windows though because I hate doing that lol, no overclocks applied except using memory xmp profile, everything else in bios is stock to asus tuf gaming x690 plus wifi d4 (yeah I'm staying to ddr4 since I got 64gb)

I did a test before and after restarting, Before I restarted I had plenty of programs open along with chrome and many tabs, plus hardware monitoring software, etc, The difference was 1 second

Also note, I did get a few freezes during tests before and after restart

Beta branch didn't really make a difference to performance, 1 second difference, Also browser is open

To note, HWMonitor reported cpu 0 and 2 were between 80-90%, Core 4 between 40-60%, Core 6 30-50%, Core 8 would stay mostly at 0% and occasionally run to 10%, All other cores were 0% Note all those cores are P cores, All E cores sit around 10-20%

 

Time : 1:43 or 103 seconds (BETA: 1:42 or 102 seconds)
CPU : 12th Gen Intel(R) Core(TM) i9-12900KF Base speed of 3.19Ghz (P core clock of 4.9-5ghz and E core 3.7ghz according to HW monitor when running test) 
Graphics card : RTX 3080 ti
RAM : 64 GB DDR4 3200 MHz

CAS : 16-18-18-36
HDD : Samsung SSD 980 PRO 2TB NVMe on pcie4

 

[gabberworld]

11 hours ago, magei said:

Just decided to try this my self, Just finished putting my system together last week, I didn't reinstall windows though because I hate doing that lol, no overclocks applied except using memory xmp profile, everything else in bios is stock to asus tuf gaming x690 plus wifi d4 (yeah I'm staying to ddr4 since I got 64gb)

I did a test before and after restarting, Before I restarted I had plenty of programs open along with chrome and many tabs, plus hardware monitoring software, etc, The difference was 1 second

Also note, I did get a few freezes during tests before and after restart

Beta branch didn't really make a difference to performance, 1 second difference, Also browser is open

To note, HWMonitor reported cpu 0 and 2 were between 80-90%, Core 4 between 40-60%, Core 6 30-50%, Core 8 would stay mostly at 0% and occasionally run to 10%, All other cores were 0% Note all those cores are P cores, All E cores sit around 10-20%

 

Time : 1:43 or 103 seconds (BETA: 1:42 or 102 seconds)
CPU : 12th Gen Intel(R) Core(TM) i9-12900KF Base speed of 3.19Ghz (P core clock of 4.9-5ghz and E core 3.7ghz according to HW monitor when running test) 
Graphics card : RTX 3080 ti
RAM : 64 GB DDR4 3200 MHz

CAS : 16-18-18-36
HDD : Samsung SSD 980 PRO 2TB NVMe on pcie4

 

Yeah . Freeze is that what kills  2-3 seconds and it really pisses me off. Because its possible see with eye.