07 — Game Benchmarks
Most tests at 1080p — the sweet spot for this hardware (~RX 6600 level).
Performance Expectations
| Resolution | Quality | Typical FPS Range | Notes |
|---|---|---|---|
| 1080p | Low | 100–144+ | Esports, older titles |
| 1080p | Medium | 80–120+ | Sweet spot for most games (elektricM docs) |
| 1080p | High | 60–100+ | Most titles (elektricM docs) |
| 1080p + FSR Quality | High + FSR | 70–100+ | Free performance boost | [confirmed: @1_gec, 04/12/2025]
| 1440p | Medium + FSR | 50–80 | Playable with upscaling (elektricM docs) |
|---|---|---|---|
| 4K | Low + FSR | 30–40 | Older/less demanding titles only |
AAA Titles
Cyberpunk 2077
| Settings | FPS | Notes |
|---|---|---|
| 1080p High + FSR, no RT | 70–90 | elektricM docs |
| 1080p High + FSR + RT (lighting only) | 50–60 | elektricM docs |
| 1080p Ultra + FSR3.1 | 100+ | elektricM docs |
| 1080p CPU-bound areas | <60 even at lowest (Discord user) | CPU bottleneck in dense areas |
| Power draw | Up to 235W | Most demanding game in the library (elektricM docs) |
Benchmark scores:
mitigations=off: +18 FPS boost — elektricM docs6-core vs 8-core unlock (40 CU, 1900 MHz GPU @ 880 mV, 3900 MHz CPU @ 1150 mV, FHD, _kierownik, 30/07/2026):
| Settings | 6-core FPS | 8-core FPS | Gain |
|---|---|---|---|
| Low, no upscale | 80 (170W) | 89 (180W) | +11% |
| Low, FSR2 | 85 (160W) | 93 (173W) | +9% |
| Low, FSR3 | 83 (160W) | 93 (162W) | +12% |
| Medium, no upscale | 72 (178W) | 76 (187W) | +6% |
| Medium, FSR2 | 79 (170W) | 90 (178W) | +14% |
| High, no upscale | 63 (180W) | 66 (189W) | +5% |
| High, FSR2 | 73 (170W) | 80 (180W) | +10% |
| Ultra, no upscale | 55 (180W) | 58 (189W) | +5% |
| Ultra, FSR2 | 63 (175W) | 70 (185W) | +11% |
| Ultra, FSR3 | 64 (175W) | 70 (184W) | +9% |
8 cores give a consistent +5-14% FPS in Cyberpunk with ~8-10W higher power draw. See 02-BIOS & Firmware for core unlock methods.
8-core 40CU Cyberpunk results (Jul 30 2026):
| User | Config | Settings | FPS | Notes |
|---|---|---|---|---|
| dbkretro | 8 core + 40 CU | 1080p | ~60 | Up from ~52 with 6 core — "8 core 40cu gang" |
| qwert9811 | 8 core unlocked | Dogtown (CPU-heavy area) | Early 50s | Up from early 40s — ~10 FPS gain in most demanding area |
Tips: Enable FSR Quality for a significant boost. DLSS/FSR Frame Generation works well. (elektricM docs)
Black Myth: Wukong
| Settings | FPS | Notes |
|---|---|---|
| 40 CU, 1080p Low (FSR off) | 61 avg | Old Lamer YouTube benchmark (June 2026) |
| RX 6700 (reference) | 59 avg | Same test conditions for comparison |
| RX 7600 (reference) | 71 avg | Same test conditions for comparison |
Source: Old Lamer YouTube benchmark (40CU BC-250 vs RX6700 & RX7600). The BC-250 at 40 CU slightly edges the RX 6700 in this title. See 11-community-and-resources for link.
Hogwarts Legacy
| Settings | FPS | Notes |
|---|---|---|
| 1080p Medium | ~60 (Discord user) | Needs zram enabled (16 GB RAM is tight) |
| With FSR4 on Proton GE | Playable [confirmed: @lonewolf05849, 16/01/2026]
zram-size = ram x 0.75 and close background apps. [confirmed: @hojnikb, 08/03/2026]Use 6 GB static VRAM allocation to avoid OOM crashes. [confirmed: @big_trov, 28/02/2026]
Red Dead Redemption 2
| Settings | FPS | Notes |
|---|---|---|
| Full graphics (DX11) | Smooth (45+ FPS min — elektricM docs) | Heatsink barely warm with 120mm fan |
| 2230 MHz GPU | Crashes (Discord user) | Reduce to 2150 MHz for stability (Discord user) |
| 10/6 VRAM split | Fixed crash (Discord user) | Static allocation avoids ZRAM conflicts | [confirmed: @xseol, 12/07/2025]
Launch flag: -useMaximumSettings — elektricM docs
Adapter fix: May detect as software rendering — change adapter in graphics settings to match vulkaninfo --summary output (elektricM docs)
Temps: ~75C during gameplay. [confirmed: @.strykur, 15/01/2026]
8-core 40CU RDR2 (dbkretro, Jul 30 2026): 1080p, decent quality settings — near 60 FPS with main dip during snow scenes. GPU bound in most scenarios but less stutter in city areas with extra cores.
8-core 38CU RDR2 benchmark (sho.ta, Aug 13 2026): ~65 FPS in benchmark, 55–85 real FPS, High-Ultra 1080p on Bazzite with 0.5/15.5 memory split (38 CU @ 1900 MHz 900 mV, 8 cores @ 3.85 GHz 1150 mV). "Timegraph is smooth most of the time" — FPS drops significantly in Saint Denis (~40–45 FPS) with spiky timegraph due to poor CPU performance.
Spider-Man 2
| Settings | FPS | Notes |
|---|
| Medium, 6 GB VRAM | ~60 | GPU/CPU around 65C | [confirmed: community reports]
| 1080p native AA, Medium | 55–60 | Textures Medium, rest Low |
|---|
| 1440p FSR3 Quality | ~55–60 | Similar to 1080p native | [confirmed: @nexgen3d, 09/12/2025]
| Auto VRAM | Crashes after 5–10 min | Must use 6 GB+ static allocation | [confirmed: @bigmedi, 02/05/2026]
Fix: Add kernel params: ttm.pages_limit=3959290 ttm.page_pool_size=3959290
Elden Ring
| Settings | FPS | Notes |
|---|---|---|
| Any settings | 45–51 (Discord user) | CPU-bound — elektricM docs report expected 60 FPS with settings adjustments |
| 4K | 30–40 | Playable but choppy | [confirmed: @imeden, 09/12/2025]
Fix skybox artifacts:
RADV_DEBUG=nohiz in Steam launch options.8-core GPU unlock boost (glide_2026, Aug 3 2026): Temporary unlock of extra cores gave ~10 more FPS ("definitely added 10ish frames"), though still fluctuates around 60 — gains are CPU-dependent and vary by scene.
The Crew Motorfest
| Settings | FPS | Notes |
|---|---|---|
| 1080p High | 60 capped | GPU OC 2100 MHz, CachyOS with Proton-Cachy |
Borderlands 3 [Discord user]
| Settings | FPS | Notes |
|---|---|---|
| 1080p Medium, 8x AA | ~62 | Eden-6 benchmark; inventory has black box background (cosmetic only) |
Starfield [Discord user]
| Settings | FPS | Notes |
|---|---|---|
| 1080p Medium | 48–60 | 60 FPS capped with frame gen |
| 1080p High | 39–60 | FG: 58–60 FPS |
| 1080p Ultra | 35–60 | FG: 58–60 FPS, temps 64C max |
| 1080p Medium (New Atlantis) | 48–53 | Most demanding location |
Alan Wake 2
| Settings | FPS | Notes |
|---|
| 1080p | ~100 with FG | Reported playable, frame gen recommended | [confirmed: @chiribayashepherd, 10/06/2026]
Shooters & Action
Doom Eternal
| Settings | FPS | Notes |
|---|
| Ultra (not max VRAM) | 100 | Loves Vulkan | [confirmed: @spitko, 03/02/2025]
| 512 MB VRAM | Heavy stuttering | Needs higher VRAM allocation | [confirmed: @fforduck, 02/05/2026]
Doom: The Dark Ages (Update 2+)
| Settings | FPS | Notes |
|---|---|---|
| Low / Handheld preset, 1080p, FSR Quality | ~60 | GPU 2230 MHz, CPU stock |
Fix: VRS crash — use Vulkan_NullVRS ↗ layer (bangstk, May 2026).
Crimson Desert
| Settings | FPS | Notes |
|---|---|---|
| 1080p Medium, no scaling | 38–42 |
Fix: Use Proton Experimental Bleeding-Edge branch with VKD3D RDNA1 fix. Version 1.02 works. [confirmed: @nohanmv, 01/04/2026]
Space Marine 2
| Settings | FPS | Notes |
|---|---|---|
| 8 cores unlocked | ~2x vs 6 cores | smcelrea, Aug 10 2026 — "almost double the perf in space marine 2, in the rippers section of the campaign, with 8 cores unlocked" |
Pragmata
| Settings | FPS | Notes |
|---|---|---|
| 8 cores + governor tweaks | High | lovelifetrustfaith, Aug 10 2026 — "pragmata ran pretty good before the cpu core unlock and additional tweakings"; later "pragmata now on stock governor values, just insane..." |
Death Stranding 2
| Settings | FPS | Notes |
|---|---|---|
| 36 CU, ultrawide 1440p High + frame gen | 60 locked | dartzon, CachyOS, Thermalright PA120, GPU backplate cooler, <72C |
| 36 CU, 1440p High, no frame gen | Dips under 60 | CPU bottleneck (dartzon) |
| 1080p Low, FSR 3.1.5 | ~60 dips to 45 | CPU limited |
PICO upscaler: PlayStation's FSR equivalent was ported to PC as "PICO" — works on Decima engine games (Horizon, Death Stranding). User reports it's superior to AMD FSR3 (dartzon, May 2026).
Cooling: dartzon used Thermalright Peerless Assassin 120 + GPU backplate cooler with fans for VRAM chips. Temps never exceeded 72C with 36 CU unlocked.
Monster Hunter Wilds
| Settings | FPS | Notes |
|---|---|---|
| 1080p Medium | +12% with mitigations=off |
Fix: sudo rpm-ostree kargs --append='mitigations=off' [confirmed: @filippor, 09/12/2025]
Doom 2016
| Settings | FPS | Notes |
|---|---|---|
| Max graphics | 100 | Spanish community benchmark |
Forza Horizon 6 [Discord user, May 2026]
| Settings | FPS | Notes |
|---|---|---|
| 1080p High + FSR 3.1.5 | 40–60 | Preset High, FSR helped fix pixelated textures |
Genshin Impact
| Settings | FPS | Notes |
|---|---|---|
| 1080p Ultra | 60 locked (Discord user) | 65–68C after 4 hours (Discord user). Runs great. |
Competitive / Esports
CS2 (Counter-Strike 2)
| Aspect | Detail |
|---|---|
| Performance (elektricM docs) | 100+ FPS expected at 1080p |
| Community test | 60–80 FPS with stuttering in some configurations | [confirmed: @nonu0038, 14/05/2025]
| Stability test | Unstable OC crashes CS2 first — add +30 mV if crashes occur | [confirmed: @jayrule, 03/12/2025]
Rocket League
| Aspect | Detail |
|---|---|
| Performance (elektricM docs) | 120+ FPS expected at 1080p |
| Community test | 60 FPS locked at 1080p — settings maxed | [confirmed: @codyrainy, 23/04/2026]
Valorant
Expected: Technical challenges — anti-cheat may have issues on Linux (elektricM docs).
Single-Player / Story Games
| Game | Performance | Notes |
|---|
| Half-Life: Alyx | ~80 FPS | CachyOS | [confirmed: @nataliezaki, 27/05/2026]
| Hellblade: Senua's Sacrifice | ~180 FPS | High FPS, well-optimized |
|---|---|---|
| Hellblade II: Senua's Saga | 60fps FSR4 Quality / 65 balance / 73 performance | Medium settings, 1080p. 60fps on FSR4 Quality requires the gfx1013 compute queue patch. [lonewolf05849, Aug 8 2026] |
| Mortal Kombat 1 | Struggles to stay locked at 60 FPS | GPU refuses to fully boost (felingreenleaf, Aug 6 2026). [verified: Steam appid 1971870] |
| Resident Evil 9 | Solid 60 FPS (dbkretro, Aug 9 2026) | "Solid 60fps on RE4 and RE9" (dbkretro, Aug 9 2026); appears in Old Lamer BC-250 benchmark video. |
| Assassin's Creed IV: Black Flag – Resynced | Playable (dmoraza, Aug 5 2026) | 2K FSR Quality + FG, max details no RT. Tested with broken RAM at 1600 MHz. [Steam appid 242050] |
| Resident Evil 4 (2023 Remake) | ~60 FPS stable — dbkretro, Aug 9 2026 | Bazzite + governor + 8-core/40 CU at 1750 MHz. Previously reported crashes (May 2026) no longer reproducible. [Steam appid 2050650] |
| Arc Raiders | 60+ (Discord user) | Medium, FSR Quality — 60+ FPS, ~69C |
| Ghost of Tsushima [Discord user] | 45–60 at 1080p Low | Crashes without game update v1053.5+; runs at 1.7–1.9 GHz GPU OC. Check ProtonDB for AMD GPU fixes. |
| Final Fantasy VII Remake | Playable | Rebirth broken: "DX12 is not supported on your system" — game checks for specific GPU compatibility (elektricM docs) | [confirmed: @dwtoledo, 12/10/2025]
| Horizon: Zero Dawn | Great at 1080p High | No upscaling needed | [confirmed: @nexgen3d, 09/12/2025]
| Horizon: Forbidden West | 45–60 / 70–90 with FG | FSR + frame gen, low settings | [confirmed: @_nk10, 15/12/2025]
| Hunt: Showdown 1896 | 90–120 with FSR / 20–40 without | |
|---|---|---|
| Forza Horizon 5 [Discord user] | 40–100 FPS | Varies heavily by settings |
| Stellar Blade [fforduck, Discord user] | 50–80 FPS at 1440p | Medium settings, FSR4 |
| Helldivers 2 [Discord user] | 40–60 FPS | |
| Valheim | 40–80 FPS | 80 FPS with mitigations=off (Discord user) |
| GTA V Enhanced (RT) [Discord user] | Smooth on Mesa 26 | Went from 3-5fps crash to smooth with Mesa 26 (CachyOS ships Mesa 26) |
| Oblivion Remaster [Discord user] | 30–75 FPS at 3440x1440 | With/without frame gen |
| Marvel Rivals [Discord user] | 100–190 FPS |
| Warframe | 75 FPS at 1080p | V-Sync ON, no FSR | [confirmed: @whomstdv, 02/12/2025]
| War Thunder | Playable at 1080p High | Max GPU OC, no RT |
|---|---|---|
| The Last of Us Part I | 60 FPS locked, 1080p Medium-High | elektricM docs |
| The Callisto Protocol | 60–85 at 1080p Medium (Discord user) | 60 locked, hits 85 frequently |
| Tomb Raider (2013) | 100–140 FPS at 1080p Max |
| Death Stranding | 40–50 FPS at 1080p Max | | [confirmed: @pijuli., 24/03/2026]
| Zenless Zone Zero | Crashes with "Memory shortage" error (Discord user) | May need workaround |
|---|---|---|
| Diablo IV | Playable | Medium-high settings |
| Baldur's Gate 3 | Playable at 1080p | Lower settings in cities |
| Detroit: Become Human | 60 FPS capped, 1080p Medium | elektricM docs |
| Devil May Cry 5 | 100 FPS, 1080p High | elektricM docs |
S.T.A.L.K.E.R. 2 (May 2026)
| Config | FPS | Notes |
|---|---|---|
| Stock 24 CU, 2000 MHz | ~55 (drops to high 40s) | Area after opening cinematic |
| Stock + FSR frame gen | High 90s | |
| 36 CU, 2000 MHz | Mostly 60 (drops to high 50s) | |
| 36 CU + FSR frame gen | 110-120 |
Subnautica 2 (May 2026)
Runs on CachyOS with Proton Experimental, 40 CU, lower settings (biohazardv2.0).
Cities: Skylines 2 (200k population)
20–30 FPS — CPU limited (simulation-heavy).
Emulation
| System / Game | FPS | Notes |
|---|---|---|
| Ryujinx (Switch) — TOTK | 20 FPS consistent | Appears to be board limitation (elektricM docs) |
| Ratchet & Clank (RPCS3) | 45–60 | Playable | [confirmed: @whomstdv, 25/11/2025]
| Breath of the Wild (Cemu) | Works | |
|---|---|---|
| Xenia (Xbox 360) | Does not work — freezes system | |
| PCSX2 (PS2) | Excellent | elektricM docs |
| Dolphin (GameCube/Wii) | Excellent | elektricM docs |
| RPCS3 (PS3) | Good for lighter titles | elektricM docs |
Newly Tested Games (Late May 2026)
| Game | Performance | Notes |
|---|---|---|
| Hitman 2 (40 CU, 1500 MHz) | 160 FPS vs 120 FPS stock | 1.33x CU scaling (itsanarse) |
| Fatal Frame 2 (40 CU) | 60 FPS at 1400-1500 MHz | 24 CU needed 1850-2000 MHz for same -- lower temps/power (maskofsin) |
| MGS3 Delta (40 CU) | 66% FPS boost over 24 CU | big_trov |
| Forza Horizon 6 | Playable via Proton (CachyOS) | Low memory warning after prologue; try 512MB split + zswap (antmagl, jeffr7814); menu FPS drops to 15 (capt.cat_13); works on Proton CachyOS |
| Crimson Desert (38-40 CU) | ~55 FPS FHD, no scaling | +10-15 FPS over 24 CU (pijuli.); CPU-limited in some areas (vfxmz) |
| Marvel Rivals | Playable | Season 8 perf mod on NexusMods (graytl); up to 190 FPS |
| Returnal | Heavy artifacts on marginal 40 CU boards | Good test game for CU health (capt.cat_13) |
| Death Stranding 2 | 36 CU, ultrawide 1440p High + FG: 60 | Well-optimized; CPU bottleneck without FG. See detailed entry above. |
| S.T.A.L.K.E.R. 2 | 24 CU stock: ~55 FPS; 36 CU: ~60; +FG: 110-120 | Big uplift from more CUs. See detailed entry above. |
| Subnautica 2 | Playable at lower settings | Proton Experimental, CachyOS, 40 CU (biohazardv2.0) |
| New Batman (2026) | Runs, GPU bound | 40 CU (codyrainy) |
Games That Don't Work
| Game | Reason | Source |
|---|---|---|
| Fortnite | Easy Anti-Cheat on Linux -- cannot run | elektricM docs |
| Final Fantasy VII Rebirth | "DX12 is not supported on your system" -- game checks for specific GPU compatibility, no fix for BC-250 yet | elektricM docs |
| Spider-Man 2 | Out-of-memory crash with 512MB VRAM. Fixes (help-thread): set 6GB static VRAM in BIOS (_nk10), add TTM kernel params (hojnikb), run 32GB swap script from NexGen3D repo, lower in-game settings (zerosumpr), or add DXVK config overrides (newgbaxl) |
| Expedition 33 (Clair Obscur) | Crashes with 512 MB VRAM -- use 6 GB static allocation or RADV_DEBUG=nohiz | Community report | [confirmed: @fforduck, 12/05/2026]
| Palia | Crashes without workaround (swap may help) (Discord user) | Community report |
|---|
Launch Options & Tweaks (Quick Reference)
| Game / Fix | Launch Option | Source |
|---|---|---|
| Fix artifacts (general, Mesa 25.1+) | RADV_DEBUG=nohiz %command% | elektricM docs |
| FPS overlay (any game) | mangohud %command% | elektricM docs |
| CPU optimization | gamemoderun %command% | elektricM docs |
| Combined (Mesa 25.1+) | RADV_DEBUG=nohiz mangohud gamemoderun %command% | elektricM docs |
| Fix compute queue (Mesa < 25.1) | RADV_DEBUG=nocompute %command% | elektricM docs |
| Combined with debug (Mesa < 25.1) | RADV_DEBUG=nohiz DXVK_HUD=fps,gpu MANGOHUD=1 %command% | elektricM docs |
| Force Vulkan driver | VK_ICD_FILENAMES=/usr/share/vulkan/icd.d/radeon_icd.x86_64.json %command% | elektricM docs |
| Steam Deck compat (Fallout 4, Skyrim SE) | SteamDeck=0 %command% | Community |
| FSR4 upgrade for FSR3 | FSR4_UPGRADE=1 %command% | Community |
| VRS crash fix (Doom TDA) | Use Vulkan_NullVRS ↗ layer (bangstk) | bangstk, May 2026 |
| CPU performance boost | mitigations=off (kernel param) | elektricM docs |
| Larger shader cache | __GL_SHADER_DISK_CACHE_SIZE=10737418240 in /etc/environment | elektricM docs |
Environment Variables (Add to `/etc/environment`)
VK_ICD_FILENAMES=/usr/share/vulkan/icd.d/radeon_icd.x86_64.json:/usr/share/vulkan/icd.d/radeon_icd.i686.json
RADV_DEBUG=nohiz
__GL_SHADER_DISK_CACHE_SIZE=10737418240
Performance Optimization Tips
1. 1080p is the sweet spot — 1440p works with FSR, 4K only for older titles (elektricM docs)
2. CPU is the main bottleneck in most modern games (GDDR6 shared memory latency) [confirmed: @corbanitevevo, 08/12/2025]
3. VRAM: 4 GB for most games, 6 GB for demanding AAA titles (elektricM docs: 4 GB recommended; 6 GB from community)
4. Use FSR for free performance boost (elektricM docs)
5. Update to Mesa 26.x+ for best compatibility and performance (elektricM docs)
6. Try Proton-GE for better compatibility (elektricM docs)
7. Kernel 6.19.x for VRR and DP audio fixes (gennro); 6.18 LTS as stable fallback
8. Mesa 26.x recommended — significant RT and performance improvements; 25.1+ minimum
9. mitigations=off gives ~10–15% FPS boost in CPU-bound games (elektricM docs: "mitigations=off for +10-15% FPS")
10. Keep GPU under 85C for long-term stability (elektricM docs)
11. Disable Handheld Daemon if using Bazzite for gaming (elektricM docs):
sudo systemctl disable --now hhd && sudo systemctl mask hhd
12. CachyOS may be ~5–10% faster than Bazzite in raw benchmarks [confirmed: @.captainwasabi, 23/05/2026]
13. 40 CU unlock: more CUs at lower clocks match higher clocks at stock 24 CU — cooler and less power (big_trov: 40 CU at 1200 MHz = 60 FPS at 73C, 30W less than 24 CU at 2000 MHz achieving same FPS). See 02-BIOS.
40 CU Unlock — Gaming Benchmarks
Community-tested by big_trov and essdee4336 (May 2026). All runs with P12 Pro fan, opened mid fins, PTM7950, 80mm back fan unless noted.
Furmark (Vulkan, 1080p)
| Config | FPS | Temp | Power | User |
|---|---|---|---|---|
| 24 CU, 2000 MHz stock | 57 | 77C | — | big_trov |
| 40 CU, 2000 MHz | 91 | 90C | — | big_trov |
| 40 CU, 1850 MHz / 910 mV | 137 | 71C | — | essdee4336 |
| 40 CU, 2000 MHz / 950 mV | 145 | 75C | — | essdee4336 |
| 40 CU, 2150 MHz / 990 mV | 153 | 79C | ~200W | essdee4336 |
| 40 CU, 2200 MHz | — | 98C (instant) | — | big_trov |
| 40 CU, 2230 MHz / 1050 mV | — | — | — | mrfrakes |
| 40 CU, 2300 MHz | 150 | 85C | ~288W | big_trov |
| 40 CU, 1850 MHz / 960 mV (2x 120mm fans) | — | — | — | soulygenius |
| 40 CU, 2000 MHz / 1000 mV (SMU_OC 78C tctl) | — | — | — | stevounit |
| 40 CU, 2300 MHz / 4100 MHz CPU | — | — | — | adixd90 |
| 40 CU, 1920 MHz / 960 mV (MX-7, triple fan) | — | throttled | — | linepanda (May 2026) |
| 38/40 CU, 1900 MHz | 130 | 84C | 336W wall | pijuli. |
| 24/40 CU, 2130 MHz (same board) | 95 | 84C | 320W wall | pijuli. |
pijuli. tested a 38/40 CU board (2 harvested in SE1 SH0). At 1900 MHz with 38 CUs: 130 FPS, 84C, 336W from wall. Same board at 24 CU/2130 MHz: 95 FPS, 84C, 320W. 35% FPS increase at equivalent temps with only 16W more from wall. Cooling: P12 Max, middle fins removed, PTM7950, new thermal pads, no cage/no back fan.
Superposition (40 CU @ 2200 MHz)
| Preset | Score | GPU Clock | CPU Clock | Notes | User |
|---|---|---|---|---|---|
| Medium | 13507 | 2200 MHz | stock | — | big_trov |
| High | 12491 | 2200 MHz | stock | — | big_trov |
| Medium (4 GHz CPU) | 14004 | 2200 MHz | 4000 MHz | Meager boost from CPU OC | big_trov |
| Extreme (2300 MHz) | 5759 | 2300 MHz | 3500 MHz UV | ~250W at wall | big_trov |
| Extreme (2230 MHz) | — | 2230 MHz | — | 24CU was 235W at 2230 | big_trov |
| Extreme (2100 MHz) | — | 2100 MHz | 4000 MHz | 1020 mV, 40CU | codyrainy |
Superposition Leaderboard (Extreme, 1080p)
24 CU (Stock, Unharvest Disabled)
| Rank | User | Score | GPU Clock | CPU Clock | mV | Date |
|---|---|---|---|---|---|---|
| 1 | nexgen3d | 4713 | 2530 MHz | 4175 MHz | 1165 | Jan 2026 |
| 2 | nexgen3d | 4690 | 2530 MHz | 4150 MHz | 1150 | Jan 2026 |
| 3 | nexgen3d | 4668 | 2500 MHz | — | — | Jan 2026 |
| 4 | nexgen3d | 4576 | 2400 MHz | 3850 MHz | — | Mar 2026 |
| 5 | nexgen3d | 4329 | — | — | — | Jan 2026 |
| 6 | nexgen3d | 4317 | — | — | — | Dec 2025 |
| 7 | nexgen3d | 4280 | — | — | — | Dec 2025 |
| 8 | big_trov | ~4200 | 2200 MHz | stock | — | Feb 2026 |
| 9 | big_trov | 3975 | 2000 MHz | 3500 MHz | — | Apr 2026 |
| 10 | .captainwasabi | ~3700 | — | — | — | May 2026 |
nexgen3d runs liquid cooling (MSI AIO), CachyOS, SMU governor. 24 CU community target is 5000 (uncracked).
40 CU (Unharvest Enabled)
| Rank | User | Score | GPU Clock | CPU Clock | mV | Notes | Date |
|---|---|---|---|---|---|---|---|
| 1 | gennro | ~5900 | — | — | — | — | May 2026 |
| 2 | big_trov | 5759 | 2300 MHz | 3500 MHz UV | — | — | May 2026 |
| 3 | pm_me_kitsunemimi | 5320 | 2250 MHz | 4000 MHz | — | 36 CU, 8 cores, Bazzite, OpenGL; "76 max iirc" | Aug 2026 |
| 4 | codyrainy | ~5400 | 2100 MHz | 4000 MHz | 1020 | 40 CU | May 2026 |
| 5 | cralant | ~5400 | 2150 MHz | 3800 MHz -15 | 1035 | 40 CU | May 2026 |
| 6 | dznuts | 5300 | 2270 MHz | 4000 MHz | — | 38 CU | May 2026 |
| 7 | dznuts | 5300 | 2200 MHz | — | 1060 | 38 CU, CachyOS | May 2026 |
| 8 | mitchthepreacher | 5150 | — | — | — | Redux case; "Temps are 75 but its a jet engine" | Aug 2026 |
| 9 | land_and_air | — | 2000 MHz | stock | — | — | May 2026 |
40 CU Extreme already surpasses the 24 CU record (4713) by 22%+ at lower clocks (2300 vs 2530 MHz). Theoretically should reach ~6500+ at equivalent clocks. More scores expected as community adopts the unlock. Post your results in the Discord #benchmarks channel.
Aug 2026 updates: big_trov confirmed the current record is "5800 or so" (Aug 12 2026). pm_me_kitsunemimi hit 5320 @ 2250 MHz GPU / 4 GHz CPU on Bazzite with 8 cores unlocked — "Not sure what the average score is but 5320 is good" (rocksalt_, Aug 12 2026), max temps "76 max iirc". mitchthepreacher reached 5150 after a CachyOS update with no unlock/OC changes, in a Redux case, "Temps are 75 but its a jet engine" (Aug 13 2026).
38 CU = 36 CU (Aug 12 2026): a "38 CU" config is effectively 36 CU — the Shader Engines (SE0/SE1) must have symmetric WGP counts. fforduck: "You have 36 unlocked 😉 SE0 and SE01 need the same amount. If you have disabled one in SE0 it will automatically disable one in SE1" — and per hashtagoctothorp, disabling another WGP in SE1 gives the same score. The only meaningful CU counts are 24/28/32/36/40 (counting full pairs); the dznuts "38 CU" runs above are the same silicon configuration as 36 CU.
Max clock findings: 2400 MHz at 40 CU causes hard OCP lockup requiring power cable pull (reset/power buttons unresponsive) across multiple boards (big_trov, codyrainy, cralant). 2100-2300 MHz is the stable range for most boards. Trimming CUs for higher clocks is not advantageous: big_trov's 32CU @ 2400 MHz scored worse than 40CU @ 2300 MHz, with similar power draw (~270W).
38 CU voltage findings (May 2026): 38 CU at 2200 MHz requires ~1050-1060 mV (dznuts). 1085 mV not enough for 2230 MHz with 38 CU. At 2200 MHz, 995 mV is stable for some boards (codyrainy). Power limit and voltage ceiling intersect at ~2200 MHz for most boards, creating a hard stability ceiling.
Memory OC (May 2026): dznuts tested memory OC at 1975 MHz CL26 using RobinMemTiming. Only +80 points in Superposition and +1 FPS in Cyberpunk — not worth the instability risk. Memory OC causes 1-in-20 boot failures (no POST).
Gaming (40 CU)
| Game | Config | FPS | User |
|---|---|---|---|
| Furmark VK 1080p | 1200 MHz | 60 FPS at 73C | big_trov |
| Furmark VK 1440p | 2000 MHz | 91 FPS at 90C | big_trov |
| Doom TDA | 1440p, 40 CU | Same FPS as 1080p before unlock | Community report |
| 007 First Light | 1440p Medium + FSR | 72 FPS | big_trov, paul_lionking (May 2026) |
| 007 First Light | 1440p Medium no FSR | 48 FPS | big_trov (May 2026) |
| 007 First Light | 1080p High no FSR | 60 FPS locked | paul_lionking (May 2026) |
| Forza Horizon 6 | 3840x1600 | ~60 FPS | paul_lionking (May 2026) |
| Forza Horizon 6 | 1080p Ultra, 40 CU | ~60 FPS | Community report |
| Monster Hunter Wilds | 1080p Medium + FSR Frame Gen | 50-60 FPS | irontsuki (Jun 2026) |
| Death Stranding 2 | High settings | 40-60 FPS | gaboggamer (Jun 2026) |
| Death Stranding 2 | 3440x1440 Medium-High + PICO + FSR FG | 60 FPS | dartzon (Jun 2026) |
| Counter-Strike 2 | 1440p High, 40CU @ 2GHz | <100 FPS, stutter | weabchan (Jun 2026) |
| Arc Raiders | 1080p Medium no FSR, 40CU @ 2GHz | ~65 FPS | marine1067 (Jun 2026) |
| Shadow of the Tomb Raider | 1080p Highest, 40CU @ 2GHz | 90 FPS avg | bytepond (Jun 2026) |
| Shadow of the Tomb Raider | 4K Highest + CAS (50% scale) | ~63 FPS | bytepond (Jun 2026) |
| FF7 Remake | 1080p High | 80-90 FPS | nataliezaki (Jun 2026) |
| FF7 Remake | 1440p, 120 FPS cap | Solid, rare stutter to 50 | zerosumpr (Jun 2026) |
| Helldivers 2 | 40 CU, 2350 MHz | Stable | Community report |
| RE4 Remake (2023) | 40 CU, 1750 MHz | ~60 FPS stable (dbkretro, Aug 9 2026 — Bazzite + governor + 8-core/40 CU unlock). |
OCP hard lockup (big_trov, codyrainy, cralant): 2400 MHz at 40 CU causes hard lockup where reset and power buttons do nothing -- requires pulling power cable. Likely Over Current Protection triggering. Consistent across multiple boards regardless of cooling. One user with AIO reported 2400 MHz stable at 1120 mV.
CPU OC affects GPU stability (big_trov, hojnikb): Increasing CPU from 3500 to 4000 MHz lowered the GPU voltage threshold for hard lockup. Total system power draw matters -- undervolt CPU when pushing GPU limits.
Game-specific instability (May 2026): RE4 Remake crashes even with stable stress tests (nataliezaki, May 21 2026). Games need more voltage on GPU than synthetic benchmarks to be stable. If benchmarks pass but games crash, increase voltage by 10-15 mV.
Voltage wall at 40 CU (big_trov, May 2026): Two limit curves govern 40 CU stability -- a voltage ceiling and a power limit. These curves intersect at approximately 2200 MHz, creating a hard stability ceiling. Above this point, diminishing returns are severe regardless of cooling. Power consumption: ~250W from wall during gaming, ~350W during Furmark at 40 CU (bytepond, May 2026).
Hard limit: 2300 MHz at 40 CU = ~288W. Stay at or below 2200 MHz for safety with 40 CU.
Games Mentioned in Community (Limited Data)
Games with verified community mentions but limited sample size — use with caution. Sources cross-checked against export scans (2025-11 to 2026-08). "perf" = number of performance-context mentions.
| Game | Report | Notes |
|---|---|---|
| Genshin Impact | 60 FPS constant, ultra, 65-68C (4h session) | |
| Warframe | 120 FPS @ 1440p | |
| Borderlands 3 | 68 → 82 FPS @ 1080p Ultra with 2 GHz OC | |
| Starfield | Stable 60 FPS on ultra with frame gen | Better mid-high without FG |
| GTA V Enhanced | 1440p High FSR3 Quality, 65C, 40 CU @ 1500 MHz | |
| Tomb Raider (2013) | High FPS | |
| Lies of P | 90 FPS @ 1440p on high | |
| RoboCop: Rogue City | Picked up FPS after tweaks | |
| Overwatch | Drops to ~30 FPS on low with heavy effects | |
| Rocket League | 1080p locked 60 FPS max settings | |
| Forza Horizon 4 | Suspected memory leak; VRAM never frees | |
| Diablo IV | 1440p max + FSR rolling FPS | |
| Resident Evil 9 | 60 FPS w/ 1080p high manual + hair strands, FSR quality (dbkretro) | Frame gen crashes on Bazzite with REFramework |
| Resident Evil 7 | Runs very well, everything high, ~64C | |
| Mortal Kombat 1 | GPU refuses to fully boost (felingreenleaf, Aug 6 2026) | Struggles to stay locked at 60 |
| Assassin's Creed IV: Black Flag – Resynced | 45-50 FPS city / 60+ sea, TAA Native 1080p | Also 2K FSR Quality + FG max details (dmoraza) |
| Hellblade II: Senua's Saga | 60 FPS @ 1080p FSR4 Quality w/ compute queue patch | 65 balance / 73 performance (lonewolf05849) |
| Final Fantasy VII Remake | Crashing reports on 512MB BIOS allocation | Match entry in main table |
| The Last of Us Part I | FSR caps GPU clock at 1000 MHz | Workaround in elektricM docs |
| Horizon Forbidden West | Low GPU usage @ 1080p; settings changes didn't help | |
| Ghost of Tsushima | Crashes fixed via game update v1053.5+ | |
| Stardew Valley | Runs fine at 4K 60 FPS | |
| Hollow Knight | 4K playable (lighter game) | |
| Half-Life: Alyx | VR via Monado — runs well but not at max settings (ithinkibrokeit_, Apr 2026) | Also tested at ~120W TDP cap (kilrah) |
| Star Wars Battlefront II | 80-85 → 120-130 FPS after SMU governor + kernel patch (juancarlos24691, Aug 2025) |