A20 Benchmarks 2 systems AMD, Intel

[Naz]

A bit late to the punch at this point but i haven't had the time to get these done until now, so better fashionably late than never xd. Today we have a host of data to go through, looking at a20.0 vs a19.0, older system vs modern high end, a20.0 vs a19.5, a20.0 status of proven a19 tweaks, new exclusive full screen mode and a note on 30 series nvidia sli.

 

My Other 7DTD Benchmarks

 

Spoiler

Alpha 20
A20.0 Trident benchmarks

A20.0 2 systems benchmarks
Alpha 19

Spoiler

A19 Various Performance Tweaks Benchmarked

A19.3>A19.4>A19.5 Hotbox Benchmarks

Resizable Bar Trident Benchmarks

A19 RTX 3090 Trident Benchmarks

A19.2 B3 Hotbox Benchmarks

A19.0 Benchmarks 2 Systems Intel,AMD,SLI,Core Count And SMT Tested

A15 Trident Benchmarks 

 

 

Benchmark Notes & Disclaimers

 

 

Spoiler

1. These figures should be taken as ball park figures and not absolute values.
7DTD is a very difficult game to benchmark accurately. It's in alpha and sometimes does weird things. Also some things are difficult to account for such as the AI, they will behave differently and spawn differently every run, which leads to #2.

2. What's controlled for and what's not
I haven't controlled for driver versions & windows versions. My goal was just to get "close enough" results instead of 100% accurate as possible. That said i have controlled for memory leaks, restarting the game after every run. Time of day is reset from the console after every run. When changing resolution i changed ingame before shutting down then validated with the unity screen selector on next launch. No zombies/animals where killed so areas weren't "cleared" until x amount of days, to insure that on following runs they would respawn (Although i can't do much if zombies kill animals or vise versa etc). I only did 1 pass on each test, however any result that didn't look right or didn't make sense was discarded and retested.

 

3. 2 systems isn't enough to draw definitive conclusions for every configuration. The conclusions found here may or may not apply to your own systems, these results are only really comparable if you have similar hardware. But they will however give an idea of where the current performance is at with the hardware that was tested.

 

4. It's Alpha

Any update could change any conclusions drawn from these tests. Also as a work in progress things are always improving and getting worse, by the time 7DTD goes gold these results with be obsolete and invalid. Things are always changing in alpha.

 

5. Console Options used
all tests are run with increased view distance (sg optionsgfxviewdistance 12). 

6. Tests on different resolutions were done on monitors native aspect ratio
16:9 Resolutions were tested on an Asus PB287Q 4K 3840x2160 60HZ Monitor
21:9 Resolution was tested on an LG 38GL950G 3840x1600 175Hz Monitor

 

7. The benchmark Run
I've been using the same run since A14 for all tests, you can find more details on the exact run in my a15 benchmarks i did ages ago Here

8. Hotbox benchmarks were run without assigning affinity and Trident results were with assigned affinity. In the future i think i'll move to ditching the applied affinity and increased view distance so the benchmarks represent a more "out of box" experience and more representative of what most users can expect.

9. The system Trident has had a CPU upgrade since the last published results so aren't comparable with my previous results, the a19 comparisons here are with the new component.

 

Original benchmark excel and results files

 

System Tested

 

Hotbox

CPU	Intel I7 6700K 4 Core 8 Threads
Motherboard	ASUS STRIX Z270I Mini ITX Motherboard
Ram	Team Group Vulcan T-Force 16GB (2x8GB) DDR4 PC4-19200C14
Storage Save Data	Samsung SM951 512Gb M.2 SSD
Storage Game Data	Samsung SM951 512Gb M.2 SSD
GPU	Evga Sc2 Gaming GTX 1080ti Black Edition

 

System Overclock Notes

 

CPU: 4.5Ghz, 1.3V, Cache ratio 43   

Ram: Stock: Default JEDEC(2133Mhz) , Overclock: XMP Enabled (2400Mhz)   
GPU: Core:+50 Mem:+325     

Trident

CPU	AMD R9 5950x
Motherboard	Gigabyte X570 Aorus Xtreme
Ram	G.SKILL Trident Z Neo 64GB (4 x 16GB) DDR4 3600Mhz CL16-19-19-39
Storage 1 Save Data	1TB Sabrent Rocket NVMe PCIe 4 M.2 SSD
Storage 2 Game Data	2TB Sabrent Rocket NVMe PCIe 4 M.2 SSD
GPU	MSI Suprim X RTX 3090 24GB

 

System Overclock Notes

 

CPU: PBO2 OC
EDC:230 TDC:130 PPT:1300
Curve voltage offset
Best Cores: -15 (4 cores) Other 12 cores -30

Boost Override: +125Mhz

Ram: Stock: Default JEDEC(2133Mhz) , Overclock: XMP Enabled (3600Mhz)    
GPU Core:+160 Mem:+550      
GPU Bios Flashed with a 500 Watt Power Target   

 

Tested With 2 Video Settings

*Settings 1 & 2 were both the same for Ultra and Lower Settings

"Ultra Settings"

"Settings 1"

"Settings 2"

"Lower Settings"

 

Hotbox Benchmarks

First up are the system "Hotbox" results looking at A20.0 VS A19.0

A20.0 VS A19.0 Ultra Settings

A20.0 VS A19.0 Lower Settings

 

Conclusions

I didn't have GPU Utilisation and clocks to compare since i started logging those after i did those initial a19.0 benchmarks. Theses results at first glance look pretty encouraging. We see a modest but consistent plus in all but 1 performance category. However 0.1% lows are arguable the most important metric and that is the one that trails behind the a19 results by a quite substantial 25-40%. 0.1% lows represent what the lowest frame rate is 0.1% of the run, these quite clearly demonstrate the stutters, pauses, hitches people have been talking about, especially in the new RWG city's. TFP were spared the rod so to speak since these results were obtained in my nav run were the city of dirsevile has not yet been updated with the new tile system. I can certainly confirm first hand from my experience performance in the new citys can be quite rough even on high end hardware optimally configured.

 

A20.0 VS A19.5 Ultra Settings

 

 

A20.0 VS A19.5 Lower Settings

 

 Conclusions

Comparing to A19.5 The gap of improved performance decreases somewhat while the differences in the 0.1% lows are very evident. This isn't unexpected however, usually the fist major alpha is about getting a acceptable version out for people to get their first taste of the new content. While the Axx.a, b, c etc updates fine tune and optimise. it would be nice if the gains made in the smaller updates of a previous alpha are maintained for the next major version, but as far as i understand it, that's just not how it works. We can only hope TFP continue their pattern of improving performance in later updates and historically that has always been the case.

 

A20.0 VS A19.0 Tweak Benchmarks

 

These benchmarks will compare past tweaks that have proven effective in improving performance in past versions. I've also added comparisons for the new Exclusive Fullscreen Option. Testing at 1440p ultra settings for all tests except 1 addition test run at 4k for the Exclusive Fullscreen Option.
Tweaks tested
Assigning 4 Core Affinity
Boot Config Tweak

Exclusive Fullscreen Option

 

 

 Conclusions

All Tweaks tested in a19 continue to provide a benefit. The Assigned Affinity tweak seems to not provide the same boost as a19 and as you will see in Tridents results great i9mprovements seem to have been made possible implementing this tweak natively or something similar in that regard. The boot config tweak continues to provide the same boost. For those not aware the boot config tweak refers to changing some lines in a file called boot config, it comes with the caveat that you cannot run it with EAC so not ideal if you play on a server. I believe Fatal has said that enable gpu jobs will be included in A20.1 so if that comes to be we could see a nice bump in performance EAC or no EAC in 20.1 I tried the new Exclusive Fullscreen Option and it didnt seem to make any difference at 1440p (CPU Bound), the 4k test (GPU-ish Bound) saw a small loss in performance oddly, so i'd just leave it off for now. It could have some importance in sli for the 20 series nvidia cards and older however.

 

Trident Benchmarks

I do have finally a second 3090 in Trident, however 30 series sli as we know it is dead. The type of sli required for 7dtd is locked down in the driver, it can't be forced like you could with 20 series and earlier. That type of sli is called Implicit SLI and isn't included in the drivers for the 30 series. I tried pre 30 series drivers, launch drivers, hours of profile inspector tuning. Tried everything and as far as i can tell it can't be done. The type of sli for the 30 series is explicit SLI which requires native game integration by developers. 7DTD does have Vulkan which in theory does support this, but i couldn't get it to work. It's either a version of Vulkan that doesn't support it or it really does have to be made compatible by developers. In which case given the 3090 is the only 30 series card with nvlink fingers the userbase for such a feature can be counted with 1 hand easily and if AAA studios aren't bothering with it why would a smaller inde dev. So there are no sli performance benchmarks and it will probably be my last dance with muti gpu unfortunately, luckily i have use cases for the second card but it's sad to see the muti gpu space for gaming defunct RIP.

 

Spoiler

A20.0 VS A19.0 Ultra Settings

 

A20.0 VS A19.0 Lower Settings

 

Conclusions

Here we have tridents results pretty much as good as it gets and still, it's not immune to the frequent stutters and freezes in a20 with 0.01% lows of 22fps in cases and frequent sub 30fps. When your avg framerate is over 100-160fps dropping to 22 even briefly stand out like a sore thumb and despite the hardware and overclocks and performance tweaks, it's not enough to brute force the stutter into submission.

 

A20.0 VS A19.0 Tweak Benchmarks

 

Conclusions

Here we see some encouraging results for users with many core cpu's. The affinity tweak here did nothing in comparison to what it did in a19 which suggests 7dtd is doing something similar out of the box which is great for everyone. No more missing out on massive performance gains and messing with affinity shortcuts. There is still a small gain to be had however, i think this can be attributed to locking the game down to the cpu's best cores that boost higher than the others. The gain is much more modest however it does help the 0.1% lows and any help in that area is a huge advantage atm. Boot config and exclusive fullscreen paints the same story as hotboxes results.

 

Conclusions

Overall A20 is technically better than A19 very marginally, however the losses in the 0.1% fall far behind a19.0 and illustrate the hitches and stutters people have been experiencing regardless to hardware. On the brighter side however 7DTD is more friendly to many core cpu's than it has ever been and users will get close to optimal performance without having to do anything. This doesn't negate the stuttering issues however, but i believe A20.1 could possibly make a solid dent in addressing those issues. So i'm optimistic despite the issues currently some pain today for some gain tomorrow. I look forward to taking a20.1 for a run around navs 😛

If you want to chase those small improvements in performance by assigning 7DTD cores, most of the gains i'm pretty sure come from the increased clock speed of the "best cores" of my cpu, this is a relatively recent feature of cpu's so check your model and see if it has cores that are better than the others first. To find out which cores are the best ones for AMD you can use Ryzen Master which tells you. For Intel i believe Intel Extreme Tuning Utility may tell you though i'm not 100% as i don't have an Intel CPU new enough with that feature. they will be different cores depending on your CPU so i can't tell you the exact cores to lock down too. You can however look at my example of how a windows shortcut that assigns 4 core affinity Here and combine that with a guide on how to figure out the hexadecimal value you need Here

 

 

 

Edited January 27, 2022 by Naz (see edit history)