FW13 Benchmarking : Needs Your Results

Hi All,

I’ve been enjoying dialing in my FW13. I’m making this a general post as my sense is we can all benefit from comparative benchmarks independent of OS or cpu :wink:

Testing with Phoronix Test Suite (PTS) since it’s cross-platform.

Your Mission
if you choose to accept it, is to run the same benchmarks on your systems and post the resulting links below.

The first three tests to run are:

  1. openssl - max cpu load
  2. furmark - max gpu load
  3. Unigine Heaven - approx fps for games at a given resolution

Based on JanW’s post, it’s straightforward to run comparative tests including cpu temps & power usage.

I have access to a Ryzen 5700G based desktop running the same OS/kernel and started by running these tests as a baseline.

To do the same on my FW13, on linux ensure system stats are available to std user:
sudo chmod -R a+r /sys/class/powercap/intel-rapl

Next, run each test, respectively:
MONITOR=cpu.power,cpu.temp phoronix-test-suite benchmark 2312137-NE-5700GOPEN86
MONITOR=cpu.power,cpu.temp phoronix-test-suite benchmark 2312134-NE-5700GFURM44
MONITOR=cpu.power,cpu.temp phoronix-test-suite benchmark 2312135-NE-FW137640U42

Based on @lane_ftw’s response, here are more details for Windows users:

  • for system monitoring (also works on linux):
    MONITOR=cpu.power,cpu.temp phoronix-test-suite

  • after checking to confirm the above tests were available on Win, I didn’t realize the versions have to match (my FW13 is w/o Win). Alternate way to run openssl:
    MONITOR=cpu.power,cpu.temp phoronix-test-suite benchmark openssl
    then choose option 1: RSA4096

  • Alternate for Furmark, in case the above doesn’t work due to version mismatch:
    MONITOR=cpu.power,cpu.temp phoronix-test-suite benchmark gputest
    then choose options 1: Furmark, 3: 1280 x 1024, 2: Windowed

  • Alternate for Unigine Heaven, again if versions differ:
    MONITOR=cpu.power,cpu.temp phoronix-test-suite benchmark gputest
    then choose options 1: Furmark, 3: 1280 x 1024

A comparison of the openssl results, first FW13 results are on battery, the second is on mains.

Next comparing the Furmark results.

Then the difference in Unigine Heaven (solely on plugged in).

Of interest:
a) 5700G boosts to 4.6 Ghz for one thread or 4.32 Ghz for all 16 threads
b) 7640u boosts to 4.9 Ghz for one thread or 4.30 Ghz for all 12 threads (observed max 4.23 Ghz on batt)
c) 5700G shows 29-30 fps during Furmark at spec 2000 Mhz gpu clock, while
d) 7640u shows 37-38 fps at 2600 Mhz gpu clock (observed up to 2141 Mhz on batt)
e) The 760m performs nearly as well (same fps, marginally lower test result) on battery at a 21.4% lower clock
f) Unigine Heaven shows 65% higher fps for the 760m. Power draw is quite similar.

The cpu results are 79.3% more sign/sec for the Framework and 33% fewer verify/s. Power usage was 115% lower avg and 48.1% less max.

For Furmark, the 760m uses an avg 16% less W for a 29.3% higher result at 26.7% faster fps on mains and 1/2 the power for 29.1% higher result on battery!

Part of this uplift is attributable to the difference in RAM and gpu clocks:

  • 5700G mem runs at 3600Mhz (so 1800Mhz gpu mem), however the gpu clocks at 2000Mhz
  • 7640u mem runs at 5600Mhz (so 2800Mhz gpu mem), gpu clocks to 2600 Mhz

However it is reassuring to see measurable generational improvements from RDNA3 and such a performant cpu at relatively low power.

Sys config
CPU: 7640u
RAM: 32 GB 5600Mhz (2 x 16 GB) Crucial
OS: Ubuntu 22.04.3 LTS
Kernel: 6.5.0-1008-oem kernel
WM: Xfce 4.16
GPU driver: amdgpu (OSS)
PPD, EPP power pref, scaling_gov: performance
Ambient: 21.5C

Let’s see what these laptops can do!

1 Like

Windows strikes again!

Turns out I had to read some documentation.
.\phoronix-test-suite benchmark openssl seems to work.
Didn’t see an option for furmark in .\phoronix-test-suite benchmark list-all-tests
If anyone knows the equivalent for export MONITOR=cpu.power,cpu.temp ; on windows please post.

OpenSSL 1.0.1g:
SCORE: 213.6

Sys config
CPU: 7840u
RAM: 32 GB 5600Mhz (2 x 16 GB) Crucial
OS: Win11 Pro Build 22631
GPU driver: AMD 23.12.1
PPD, EPP power pref, scaling_gov: Windows Balanced
Charger: 45W
Battery: 85% (limited in bios)
Air: 0C

Temps during run, fan never turned on that I could hear.

20C with other apps running
SCORE: 210


20C after a reboot
SCORE: 212.9


Ok I’m willing to contribute but what are the commands to execute exactly, and what data do you expect me to collect/paste, on top of the phoronix TS report?

@lane_ftw thank you for running these tests!

I chose not to setup my FW13 to dual-boot, so I was able to confirm available tests for Windows but not actually run them.

Based on your reply, I’ve updated the above commands to be OS agnostic.

Will you check to see what version of the compress-7zip test is available for pts on Win11?

It looks like v1.10.0 on linux. You can check by installing via:
phoronix-test-suite install compress-7zip

This could serve as an alternate multi-threaded test.

It’s interesting you were able to access 0C air temp. It appears your FW13 didn’t come close to maxing all of its threads, so it’s unsurprising the results are so different.

Perhaps 7-Zip will offer a more direct comparison…?

Great, thanks @fw13amd!

Updated the OP with concise commands. May need to replace the openssl test with 7-zip for a more direct Windows comparison. We’ll see what @lane_ftw determines.

The Open Benchmarking link should do it :metal:


MONITOR=cpu.power,cpu.temp phoronix-test-suite benchmark openssl

MONITOR=cpu.power,cpu.temp phoronix-test-suite benchmark gputest

MONITOR=cpu.power,cpu.temp phoronix-test-suite benchmark gputest

Hope it’s useful :wink:

I thought there was a way to put a large list in a dropdown on discourse, can’t remember how though. Was going to post the result of list all tests, regardless.

7zip works, actually hits the CPU as well.
Will do more cold weather tests when it gets cold again. Also, I can log more power parameters with hwinfo if you’d like, LMK.
20C test with other stuff running, I’m on lunch right now.

7-Zip Compression 22.01:
Compression: 64417
Decompression: 48235

Thanks @fw13amd!

I certainly find it interesting. Combining the test results of two or more Open Benchmarking results simply requires adding the test identifiers separated by commas as linked below.

For instance, the 7840u in your system ups the ante for openssl sig/s and nearly matches the 5700G in verify/s. Looking at the temps in that test, both of our FW13s hit 100C and may be thermal throttling based on the dips in the graph.

It’s interesting to see the 7640u & 7840u so close to each other in the Furmark results.

I particularly find the Unigine Heaven comparison worthwhile as it likely shows the delta between the 7640u vs 7840u in gaming.

You may wish to re-run that test after turning on UMA_Game_Optimized in your Bios as this will move your available GPU memory from 512MB to 4GB :sunglasses:

The impact will depend on how much gfx ram that test uses.

Overall, worthwhile comparo!

Cool @lane_ftw, this is a great start!

Here’s the comparison to a 7640u running linux.

Did the revised prefixes not work for temp & power on Win?

Will also be interesting to compare graphic test results :grin:

Didn’t try them until just now, sorry, had just run .\phoronix-test-suite benchmark compress-7zip
When running .\MONITOR=cpu.power,cpu.temp phoronix-test-suite benchmark 2312137-NE-5700GOPEN86
I get

I have a feeling MONITOR is a linux thing.

-3*C results. For full transparency I’m letting the laptop sit in ambient temps for 10+mins in s3 sleep before I run each cold weather benchmark.

Sys config
CPU: 7840u
RAM: 32 GB 5600Mhz (2 x 16 GB) Crucial
OS: Win11 Pro Build 22631
GPU driver: AMD 23.12.1
PPD, EPP power pref, scaling_gov: Windows Balanced
Charger: 45W
Battery: 85% (limited in bios)
Air: -3C

OpenSSL 1.0.1g:
SCORE: 215.8


7-Zip Compression 22.01:
Compression: 77495
Decompression: 63821


Just make sure you don’t get condensation inside the laptop

No worries, I’ve lived in the cold long enough to know how to deal with it.

1 Like

Hi @lane_ftw, this is quite interesting as it seems to substantiate thermal throttling during your prior 7-zip benchmarks.

Here’s the side-by-side.

That’s a 16.9% uplift in compression and 24.4% improvement in decompression!

With past laptops, a combination of repasting and undervolting the CPU kept it from throttling. I’ve not done either with the FW13 yet.

I’ve got a line on a windows system that I can use to test monitoring. I’ll keep you posted on what I find out.

It looks like the openssl test is single threaded on Win11 instead of multi.

In the meantime, I encourage running the gputest and/or Unigine Heaven benchmarks for gpu comparison :sunglasses:

Had to edit some video today and connected a Sonnet Breakaway 550 with a RTX3060. The TB3 is similar to a PCIe x4 rather than a x16 when connecting directly to a motherboard.

Still a notable performance boost in Unigine Heaven results.

More to come…