Update GPU architectures

[fwyzard]

NVIDIA CUDA

• drop support for Pascal (sm 6.0);
• add support for Blackwell (sm 10.0 and 12.0).

AMD ROCm

• drop support for Instinct MI100 (gfx908) and Radeon Pro W6800 (gfx1030).

[fwyzard]

enable gpu

[fwyzard]

please test

[cmsbuild]

A new Pull Request was created by @fwyzard for branch IB/CMSSW_17_0_X/master.

@akritkbehera, @iarspider, @raoatifshad, @smuzaffar can you please review it and eventually sign? Thanks.
@ftenchini, @mandrenguyen, @sextonkennedy you are the release manager for this.
cms-bot commands are listed here

[cmsbuild]

cms-bot internal usage

[fwyzard]

assign heterogeneous

[cmsbuild]

New categories assigned: heterogeneous

@fwyzard,@makortel you have been requested to review this Pull request/Issue and eventually sign? Thanks

[cmsbuild]

-1

Summary: https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-bd38d3/52733/summary.html
COMMIT: d121459
CMSSW: CMSSW_17_0_X_2026-04-17-2300/el8_amd64_gcc13
Additional Tests: GPU,AMD_MI300X,AMD_W7900,NVIDIA_H100,NVIDIA_L40S
User test area: For local testing, you can use /cvmfs/cms-ci.cern.ch/week1/cms-sw/cmsdist/10493/52733/install.sh to create a dev area with all the needed externals and cmssw changes.

Failed External Build

I found compilation error when building:

WARNING: Target pattern parsing failed.
ERROR: no such package '@rules_java//java': java.io.IOException: Error downloading [https://github.com/bazelbuild/rules_java/releases/download/5.3.5/rules_java-5.3.5.tar.gz] to /data/cmsbld/jenkins/workspace/ib-run-pr-tests/testBuildDir/BUILD/el8_amd64_gcc13/external/tensorflow-sources_x86-64-v2/2.17.0-8889fad576923fd9a5d67c315d6f4715/build/86a8fa780a46a4b96ff787d7a11dc98a/external/rules_java/temp2800132884451372752/rules_java-5.3.5.tar.gz: GET returned 504 Gateway Time-out
INFO: Elapsed time: 27.949s
INFO: 0 processes.
FAILED: Build did NOT complete successfully (0 packages loaded)
error: Bad exit status from /data/cmsbld/jenkins/workspace/ib-run-pr-tests/testBuildDir/tmp/rpm-tmp.pL9mow (%build)

RPM build errors:
Bad exit status from /data/cmsbld/jenkins/workspace/ib-run-pr-tests/testBuildDir/tmp/rpm-tmp.pL9mow (%build)

* The action "build-install-external+tensorflow-sources_x86-64-v2+2.17.0-8889fad576923fd9a5d67c315d6f4715" was not completed successfully because The following dependencies could not complete:


[fwyzard]

The error

WARNING: Download from https://github.com/bazelbuild/rules_java/releases/download/5.3.5/rules_java-5.3.5.tar.gz failed: class java.io.IOException GET returned 504 Gateway Time-out

looks like a transient issue, and pretty much unrelated to these changes.

[fwyzard]

please test

[fwyzard]

The RunInfo failures should be unrelated.

[cmsbuild]

Pull request #10493 was updated.

[fwyzard]

please test

[makortel]

I know we need to do this eventually, but at this moment is there something specific that motivates dropping Volta? I see some number of 7.0 GPUs in the global pool at the moment (number-wise actually less than 6.X). I suppose we should run this through O&C management (I can do that on Wednesday). Given the small number of resources compared to >= 7.5 I would not expect objections.

[fwyzard]

Nothing urgent: we said we would do it after the end of Run 3 data taking, and the last data taking release should be 16.1.x.

The NVIDIA part could be further postponed.

[cmsbuild]

-1

Failed Tests: RelVals-AMD_W7900
Summary: https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-bd38d3/52756/summary.html
COMMIT: 5929fde
CMSSW: CMSSW_17_0_X_2026-04-19-2300/el8_amd64_gcc13
Additional Tests: GPU,AMD_MI300X,AMD_W7900,NVIDIA_H100,NVIDIA_L40S
User test area: For local testing, you can use /cvmfs/cms-ci.cern.ch/week0/cms-sw/cmsdist/10493/52756/install.sh to create a dev area with all the needed externals and cmssw changes.

Failed RelVals-AMD_W7900

The relvals timed out after 4 hours.

Comparison Summary

Summary:

• You potentially added 13 lines to the logs
• Reco comparison results: 6 differences found in the comparisons
• DQMHistoTests: Total files compared: 53
• DQMHistoTests: Total histograms compared: 4186813
• DQMHistoTests: Total failures: 60
• DQMHistoTests: Total nulls: 0
• DQMHistoTests: Total successes: 4186733
• DQMHistoTests: Total skipped: 20
• DQMHistoTests: Total Missing objects: 0
• DQMHistoSizes: Histogram memory added: 60.999999999999964 KiB( 52 files compared)
• DQMHistoSizes: changed ( 1000.0,... ): 1.525 KiB CSC/Summary
• Checked 227 log files, 197 edm output root files, 53 DQM output files
• TriggerResults: no differences found

NVIDIA_H100 Comparison Summary

Summary:

• You potentially removed 29 lines from the logs
• Reco comparison results: 339 differences found in the comparisons
• DQMHistoTests: Total files compared: 13
• DQMHistoTests: Total histograms compared: 216795
• DQMHistoTests: Total failures: 35456
• DQMHistoTests: Total nulls: 32
• DQMHistoTests: Total successes: 181307
• DQMHistoTests: Total skipped: 0
• DQMHistoTests: Total Missing objects: 0
• DQMHistoSizes: Histogram memory added: 0.0 KiB( 12 files compared)
• Checked 49 log files, 50 edm output root files, 13 DQM output files
• TriggerResults: no differences found

NVIDIA_L40S Comparison Summary

Summary:

• You potentially added 23 lines to the logs
• Reco comparison results: 364 differences found in the comparisons
• DQMHistoTests: Total files compared: 13
• DQMHistoTests: Total histograms compared: 216795
• DQMHistoTests: Total failures: 32544
• DQMHistoTests: Total nulls: 35
• DQMHistoTests: Total successes: 184216
• DQMHistoTests: Total skipped: 0
• DQMHistoTests: Total Missing objects: 0
• DQMHistoSizes: Histogram memory added: 0.0 KiB( 12 files compared)
• Checked 49 log files, 50 edm output root files, 13 DQM output files
• TriggerResults: found differences in 1 / 12 workflows

[makortel]

No clear objections were raised in the O&C meeting, but a question on "why V100 needs to be dropped now" was raised. Apparently there are V100 resources that "would be nice" to be kept supported as long as feasible.

One option would be to drop Pascal now, and Volta later when that is really necessary.

[fwyzard]

I don't mind keeping Volta.

I'm confused because a couple of months ago I was told that it was OK for O&C to drop Pascal and Volta. I suggested keeping them until the end of Run 3, and here we are 🤷🏻‍♂️

If we keep Volta, I would keep also Pascal for the time being. We will need to drop both relatively soon anyway, once we move to CUDA 13.x.

[cmsbuild]

Pull request #10493 was updated.

[fwyzard]

enable gpu

[fwyzard]

please test

[cmsbuild]

+1

Summary: https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-bd38d3/52982/summary.html
COMMIT: db1d090
CMSSW: CMSSW_17_0_X_2026-04-30-1100/el8_amd64_gcc13
Additional Tests: GPU,AMD_MI300X,AMD_W7900,NVIDIA_H100,NVIDIA_L40S
User test area: For local testing, you can use /cvmfs/cms-ci.cern.ch/week1/cms-sw/cmsdist/10493/52982/install.sh to create a dev area with all the needed externals and cmssw changes.

The following merge commits were also included on top of IB + this PR after doing git cms-merge-topic:

• @cms-sw fix icc and icx compiler tools to point to correct base #10521
• @cms-sw fix for gcc-plugin directory in tool file #10520

You can see more details here:
https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-bd38d3/52982/git-recent-commits.json
https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-bd38d3/52982/git-merge-result

Comparison Summary

Summary:

• You potentially removed 4 lines from the logs
• ROOTFileChecks: Some differences in event products or their sizes found
• Reco comparison results: 9 differences found in the comparisons
• DQMHistoTests: Total files compared: 53
• DQMHistoTests: Total histograms compared: 4187168
• DQMHistoTests: Total failures: 33
• DQMHistoTests: Total nulls: 0
• DQMHistoTests: Total successes: 4187115
• DQMHistoTests: Total skipped: 20
• DQMHistoTests: Total Missing objects: 0
• DQMHistoSizes: Histogram memory added: 0.0 KiB( 52 files compared)
• Checked 227 log files, 197 edm output root files, 53 DQM output files
• TriggerResults: found differences in 1 / 51 workflows

AMD_MI300X Comparison Summary

There are some workflows for which there are errors in the baseline:
34634.402 step 2
The results for the comparisons for these workflows could be incomplete
This means most likely that the IB is having errors in the relvals.The error does NOT come from this pull request

Summary:

• You potentially removed 69 lines from the logs
• ROOTFileChecks: Some differences in event products or their sizes found
• Reco comparison results: 284 differences found in the comparisons
• DQMHistoTests: Total files compared: 12
• DQMHistoTests: Total histograms compared: 200550
• DQMHistoTests: Total failures: 33992
• DQMHistoTests: Total nulls: 26
• DQMHistoTests: Total successes: 166532
• DQMHistoTests: Total skipped: 0
• DQMHistoTests: Total Missing objects: 0
• DQMHistoSizes: Histogram memory added: 0.0 KiB( 11 files compared)
• Checked 47 log files, 48 edm output root files, 12 DQM output files
• TriggerResults: no differences found

AMD_W7900 Comparison Summary

Summary:

• You potentially added 67 lines to the logs
• Reco comparison results: 322 differences found in the comparisons
• DQMHistoTests: Total files compared: 13
• DQMHistoTests: Total histograms compared: 216259
• DQMHistoTests: Total failures: 40584
• DQMHistoTests: Total nulls: 31
• DQMHistoTests: Total successes: 175644
• DQMHistoTests: Total skipped: 0
• DQMHistoTests: Total Missing objects: 0
• DQMHistoSizes: Histogram memory added: 0.0 KiB( 12 files compared)
• Checked 49 log files, 50 edm output root files, 13 DQM output files
• TriggerResults: found differences in 6 / 12 workflows

NVIDIA_H100 Comparison Summary

Summary:

• You potentially removed 10 lines from the logs
• Reco comparison results: 340 differences found in the comparisons
• DQMHistoTests: Total files compared: 13
• DQMHistoTests: Total histograms compared: 216259
• DQMHistoTests: Total failures: 32943
• DQMHistoTests: Total nulls: 35
• DQMHistoTests: Total successes: 183281
• DQMHistoTests: Total skipped: 0
• DQMHistoTests: Total Missing objects: 0
• DQMHistoSizes: Histogram memory added: 0.0 KiB( 12 files compared)
• Checked 49 log files, 50 edm output root files, 13 DQM output files
• TriggerResults: no differences found

NVIDIA_L40S Comparison Summary

Summary:

• You potentially removed 13 lines from the logs
• Reco comparison results: 337 differences found in the comparisons
• DQMHistoTests: Total files compared: 13
• DQMHistoTests: Total histograms compared: 216259
• DQMHistoTests: Total failures: 31457
• DQMHistoTests: Total nulls: 26
• DQMHistoTests: Total successes: 184776
• DQMHistoTests: Total skipped: 0
• DQMHistoTests: Total Missing objects: 0
• DQMHistoSizes: Histogram memory added: 0.0 KiB( 12 files compared)
• Checked 49 log files, 50 edm output root files, 13 DQM output files
• TriggerResults: no differences found

[fwyzard]

+heterogeneous

[smuzaffar]

+externals

[cmsbuild]

This pull request is fully signed and it will be integrated in one of the next IB/CMSSW_17_0_X/master IBs (tests are also fine). This pull request will now be reviewed by the release team before it's merged. @mandrenguyen, @sextonkennedy, @ftenchini (and backports should be raised in the release meeting by the corresponding L2)