Game: FSR 4 runs slower than native 4K on RX 7900 XTX #1319
Labels
No labels
area/accessibility
area/audio
area/input
area/install
area/menu
area/renderer
area/saves
area/tooling
area/upscaling
beanstalk: approved
beanstalk: failed
beanstalk: needs approval
beanstalk: running
beanstalk: tested
ci: approved
good first issue
gpu/amd
gpu/igpu
gpu/intel
gpu/nvidia
help wanted
kind
bug
kind
chore
kind
docs
kind
feature
kind
performance
kind
question
needs measurements
platform/linux
platform/macos
platform/steam-deck
platform/windows
preview
priority
critical
priority
high
priority
low
regression
risk
high
risk
low
risk
medium
status
blocked
status
confirmed
status
duplicate
status
invalid
status
needs-info
status
wontfix
ci: approved
No milestone
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set
Reference
MagicBeanDev/Longshot#1319
Loading…
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Reported on Discord
Longshot version
nightly-397-8e9cfc3a (source a136e5bc), nightly; subsequent captures are nightly-402-6d72ca93 (source 20bcde68).
What happened
The reporter compared FSR 4 at 67% scale with native 4K/no AA and observed lower FPS with FSR 4. Initial screenshots show the FPS values. The comparison was measured while standing on a roof in Kakariko Village, described as an especially intensive area; the reporter says the problem also persists elsewhere. They state native 4K averages about 51 FPS, while FSR 4 lowers this to about 40 FPS. They were testing uncapped. Later paired reports show FSR 3.1 Quality at 67% (2560x1440) versus native 4K with upscaling off on the same RX 7900 XTX. They report that upscaling appears to perform incorrectly with this hardware/settings combination.
Expected behavior
Upscaling at 67% render resolution should not result in lower FPS than native 4K, with all other conditions comparable.
Graphics card and driver version
AMD Radeon RX 7900 XTX, AMD proprietary driver 26.8.1 (LLPC).
CPU, RAM and operating system
AMD Ryzen 7 7800X3D (16 threads), 31 GB RAM, Windows 11 build 26200.
Settings and profiling
The nightly-402 reports record output 3840x2160, HDR Expanded, uncapped FPS, V-Sync On, Anti-Lag 2 On, threaded rendering On, ASTC as RGBA8 On, and other settings in the attached report files. The FSR 3.1 run uses Quality 67% (2560x1440); the native run has upscaler Off and render resolution 100%. Both reports have HOS_FPS_LOG=1, HOS_THREAD_TIMES=5, RX_VK_PASS_TIMING enabled, and RX_VK_SLOW_LOG=50. PsychoV35 is not established in the supplied reports.
Where in the game
KakarikoVillage_Adult; standing on a roof in Kakariko Village for the comparison. Reporter says it also happens elsewhere.
Logs and screenshots
Discord attachments
2026-11-16_20-58-14.zipand2026-11-16_20-58-31.zipcontain the original reports; initial screenshots compare FSR 4 at 67% with native 4K/no AA. Additional profiled nightly-402 captures are in2026-11-16_22-05-48.zip(native) and2026-11-16_22-06-15.zip(FSR 4), and2026-11-16_22-15-31.zip(FSR 3.1) and2026-11-16_22-15-56.zip(native). The last pair containreport.txt,longshot_log.txt,longshot_log.1.txt, andpicture.png.Still needed
Measurement method and activity during each run; confirm other settings and test conditions were comparable. Exact FPS values for the FSR 3.1 comparison are not provided in text. The stated ~51 vs ~40 FPS refers to native 4K and FSR 4 respectively.
Related issues and fixes
#744 reports similar FSR 4 slowdown as render resolution is lowered, but on an RX 6600; this RX 7900 XTX report is distinct. The FSR 3.1 image-lifetime fix in PR #556, first in nightly-285-a8a30c22, addresses retained images after leaving FSR 3.1, not this reported active-upscaling performance comparison.
PR #247, first in nightly-171-b80a68f1, fixes FSR 4 routing with Native HDR and the PsychoV35 tone mapper. The supplied captures report HDR Expanded, and do not establish Native HDR + PsychoV35, so that fix is not known to apply to this report.
Files from the Discord report:
From Discord:
Reporter asks where
longshot_env.txtbelongs for a portable install, and whether it should be next toLongshotLauncher-nightly-100-5f30eea3-windows-x64-setup. Clarified that it belongs in the game folder beside the actual game executablelongshot.exe, not beside the launcher setup executable. They have not yet provided a profile or additional FPS comparison details.From Discord:
New captures from nightly-402-6d72ca93 were submitted:
2026-11-16_22-05-48.zipis native 4K/upscaler off;2026-11-16_22-06-15.zipis FSR 4 Quality 67% (2560x1440) to 4K. Both report RX 7900 XTX / driver 26.8.1 (LLPC), Ryzen 7 7800X3D, 31 GB RAM, Windows 11 build 26200, KakarikoVillage_Adult. Profiling variables are enabled in both logs; FSR log confirms HOS_THREAD_TIMES=5 and GPU pass timing active. Other captured settings include HDR Expanded, V-Sync On, uncapped FPS, Anti-Lag 2 On, Threaded rendering On. The two archives are attached to this Discord message; exact FPS values/measurement and activity per run remain unknown.From Discord:
From Discord: the comparison was measured while standing on a roof in Kakariko Village, an especially intensive area; reporter says the issue persists elsewhere too. FPS values are in the initial screenshots attached to the Discord report. Exact values and measurement method are not transcribed, and comparability of other settings remains unconfirmed.
Updated with more from Discord: Added profiled nightly-402 FSR 3.1 versus native captures, settings, and report that slowdown is also seen outside the Kakariko roof; noted PR #556's distinct FSR 3.1 memory-release fix.
Files:
Updated with more from Discord: Added explicit FPS values and documented the relevant merged HDR/FSR 4 fix as not established to apply to the reported HDR Expanded setup.
From Discord:
confirms the comparison issue persists with HDR disabled. Settings were held constant except for upscaling; tests began at approximately the same in-game time after loading a save, and FPS figures were read from Longshot's overlay.
From Discord:
From in Discord: reports the same issue on an RX 7900 XTX regardless of HDR. Their Longshot version/channel, driver, test conditions, FPS values, and profiling logs are not yet known.
From Discord:
reports native FPS exceeds FSR 3, FSR 4, and XeSS on an RX 7900 XTX regardless of HDR. No version/channel, driver, FPS figures, test location/settings, or performance profile were provided. This is another reporter's corroboration, not a new report from the issue's original reporter.
From Discord:
From in Discord: confirms the same upscaling performance issue on an RX 7900 XTX regardless of HDR; reports native runs faster than FSR 3, FSR 4, and XeSS. 's Longshot version/channel, driver, FPS comparison, and profiling capture are not provided.