StarCraft

StarCraft 2 Benchmarks – AMD Ryzen 7 3700x vs AMD Ryzen 7 5800x (and also RAM)

So I just upgraded from an AMD 3700x to a 5800x. I wasn't able to find a lot of information on the performance difference between these generations of Ryzen on StarCraft 2, so I did some benchmarking runs to share.

I wanted a 5950x, but the global shortage has made that difficult, so I'll go top of the line in the next generation. Fortunately, this makes the tests I did more direct, as there is no 5700x yet, so the 5800x is basically the successor. Both are 8 core, 16 thread. The 5800x has higher base and boost clocks, and is a lot more expensive, even when you compare by launch price, rather than current price. The 5800x also runs hotter and generates more heat at full load.

I also decided to test the impact of RAM speeds and timings while I did this. I ran my RAM at the stock auto, which set it to 2133 MHz. I also used my manual overclock of 3800 MHz with timings and subtimings tweaked. I freshly installed Windows and drivers before testing, and used a custom video settings combination I've found to be a good balance of visuals and performance (see below).

I ran a replay of Nightscape LE 2v2. This map always seemed to hit performance heavier than a lot of others, so I figured it would be a good test. I started each benchmark as soon as my probes touched the mineral fields, ramped the replay speed to 4x at 5 seconds, and ended the benchmark at 11 minutes, just when I was pushing the enemy main and one AI surrendered.

Test system:

AMD Ryzen 7 3700x / AMD Ryzen 7 5800x

MSI Mag X570 Tomahawk

G.Skill 2x8GB 3600 MHz CL 16 Samsung B Die

ASUS ROG Strix 1080 Ti w/ EKWB water cooling

StarCraft 2 options:

2560×1440

AA on

Vsync off for testing purposes (usually off, g-sync on, vsync forced via Nvidia CP)

Shaders: High

Lighting: High

Shadows: High

Terrain: Ultra

Reflections: Off

Effects: Medium

Texture quality: Ultra

Post-processing: Medium

Physics: Medium

Models: High

Unit portraits: 3D

Read more:  A silly - and possibly cursed - idea for redesign of H&H P3

Results are presented as min, max, average. The first three runs are at 4x replay speed, and only intended to show the relative performance of the processors without me spending all day on this. The last run of each section is played on "faster," or real time, and should be more representative of the gameplay experience.

3700x w/ 2133 MHz RAM

23, 254, 78.214

22, 257, 79.683

22, 256, 79.813

RT: 60, 254, 119.340

3700x w/ 3800 MHz RAM

33, 299, 104.139

34, 298, 104.615

31, 303, 105.650

RT: 81, 309, 153.639

5800x w/ 2133 MHz RAM

30, 425, 152.291

34, 424, 156.001

39, 423, 158.737

RT: 96, 479, 235.533

5800x w/ 3800 MHz RAM

46, 426, 173.865

36, 426, 171.172

33, 423, 163.589

RT: 118, 495, 256.462

So after switching to the 5800x, it was clear it was a massive improvement. What was odd was the very consistent minimum framerates on the 3700x were wildly inconsistent on the 5800x. I thought maybe Windows was doing something odd, since I didn't reinstall between processor swaps. Windows has become so good at handling hardware changes, and the CPU itself doesn't require any specific drivers, that I thought this wouldn't matter. I reinstalled and ran the tests again on the 5800x.

5800x w/ 3800 MHz RAM (fresh Windows install w/ this CPU)

44, 491, 181.216

41, 499, 176.808

36, 501, 175.056

39, 498, 180.750

54, 488, 177.459

RT: 109, 523, 253.703

Making things even weirder, the minimum framerate inconsistency didn't level out, even landing a freakish 54 on one run.. but the maximums jumped considerably. Everything I installed and every step prior to testing was identical, save for installing Windows before testing on the 3700x, but not reinstalling when I swapped in the 5800x for the first batch.

I'm not certain what caused the improvement on the top end, but I think the inconsistency on the low end is a software problem with StarCraft. I don't think the 5800x performs inconsistently, but it appears to have brute forced another bottleneck into play. Sometimes, it lucked out and the 5800x could stretch its legs. Usually, it hovered only slightly above the results the 3700x could achieve.

Read more:  What is actually being done about smurfing?

Unfortunately, silicon will always be limited by the software you throw at it. StarCraft 2 predates the insane modern APIs and such we have today. See Doom Eternal for how well a game can take advantage of the hardware. Brute force IPC can only go so far.

Lastly, I played a quick 4v4 and benched it as I played. Most of the heavy battles showed 60-70 FPS and the minimum here surprised me when I checked the log. I must not have noticed the one time it dipped that low. It was an enjoyable and solid experience, and a significant improvement over my old 4770k, 7820x, and 3700x. You can't go wrong with Ryzen 5000 on StarCraft. 99% of the game modes I actually play should retain 100+ FPS throughout, and that's exciting.

5800x w/ 3800 MHz RAM (Shipwrecked LE 4v4 gameplay)

40, 301, 128.067

TL;DR and Conclusions:

  • Overclock RAM on Ryzen 3000 and 5000 processors to 3600-3800 MHz if you can
    • Most people agree this is the sweet spot for Ryzen RAM speed
    • Not everyone knows how to do this manually, so buying a 3600 MHz kit and running XMP is easier. XMP doesn't always work. Your mileage may vary.
    • Intel chips and other generation products will show different results. My 7820x doesn't care about RAM speeds. Results vary by application as well.
    • This may change for future Ryzen generations

  • The 3700x provides a perfectly adequate gaming experience
    • Other, cheaper Ryzen chips will also give you a solid experience
    • Especially so if 60 FPS is your target. Don't break the bank if you aren't planning on hardcore 4v4 and you aren't a stickler for performance.

  • The 5800x can give you 50-100% more frames than the 3700x in some scenarios
    • This was shocking. I expected 15-25%, more in line with the IPC gains
    • I didn't do frame time analysis graphs to show 1% and .1% frame times, but subjectively, the experience of my gameplay showed massive improvements, in line with the min / max / average results
Read more:  Why do I get so mad when I lose? How do I prevent it?

  • SC2 performance is weird
    • …and will continue to be weird as CPUs beef up. The limitations of the engine are holding the hardware back.
    • This is less apparent in actual gameplay, and more apparent in faster replays

  • This isn't scientific
    • Good benchmarks eliminate a lot more variables than I reasonably can without investing a lot of time into this, and purchasing equipment I don't need.
    • My sample size is small. I may have seen less consistency on the 3700x minimums had I kept at it, but I'm not swapping CPUs again to find out.
    • Your mileage WILL vary. You will see different results in different system and software configs, settings, or gameplay modes. Sometimes worse results.

  • A fresh Windows install never hurts
    • Even though I started with a fresh install before any testing, I saw some improvement after reinstalling again after swapping CPUs.
    • Eliminate bloat. Don't use programs you don't need. Keep Windows clean and it will remain snappy.
    • Keep drivers updated.

Source

Similar Guides


More about StarCraft

Post: "StarCraft 2 Benchmarks – AMD Ryzen 7 3700x vs AMD Ryzen 7 5800x (and also RAM)" specifically for the game StarCraft. Other useful information about this game:





Top 20 NEW Medieval Games of 2021

Swords, dragons, knights, castles - if you love any of this stuff, you might like these games throughout 2021.



10 NEW Shooter Games of 2021 With Over The Top Action

We've been keeping our eye on these crazy action oriented first and third person shooter games releasing this year. What's on your personal list? Let us know!



Top 10 NEW Survival Games of 2021

Survival video games are still going strong in 2021. Here's everything to look forward to on PC, PS5, Xbox Series X, Nintendo Switch, and beyond.



You Might Also Like

Leave a Reply

Your email address will not be published. Required fields are marked *