AMD Ryzen 7 7840HS Linux Performance With The TUXEDO Pulse 14 Gen 3

[qarium]

interesting...
Zooming in on the power graphs, it looks like the Ryzens were very consistent. Each test ran three times. And the three power spikes were almost exactly the same.
Power on the 155H was inconsistent on some tests.
The 155H often had a super aggressive power spike on the first run.
I'm guessing the 155H often hit a turbo/thermal/power limit and throttled? To the point that on some tests you can see the graphs for the 155H are much longer. i.e. the 155H had to re-run the tests multiple times to calculate an accurate average with all the turbo/power limit/throttle dance going on under the hood?

i am in no favor of intel... but outside of benchmarks and workloads who constandly drain all power budgets this is something really good.

people do not open software 3 times and run a task 3 times. it is a fact people open software and expect it to open as fast as possible.

what intel does here is very good it just makes sure it opens the app as fast as possible but this performance has power consumtion over the level of the cooling solution to the cpu has to throttle after that short period of time.

if you ask forum users like coder this is just cheating and bad for the consumer but i have different opinion.

intels 10nm process node of the CPU is horrible outdated and can not compete with TSMC 4nm for the AMD CPUs

if intel would not do such stuff who does not benefit benchmarks because it do throttle but it does benefit real life workloads like open up a app and the system just feels much more responsive for the user.

as i said the user does not open the app 3 times to force the throttle instead the user open the app 1 times and has the subjective feeling of a faster system.

as you see overall intel CPU on 10nm is 33% slower than AMD on 4nm if intel would not do such stuff they would become totally irrelevant and no one would buy them.

[mwelss]

in another forum threat people reported that ROCm 6.0 works with RDNA3 780M iGPU

the orginal post is here:



https://www.phoronix.com/forums/foru...e4#post1436182

excellent! thanks a lot, qarium!

[geerge]

Cool. So arc was worth it even if currently just to get intel's iGPUs up to par. And intel's CPU is power sucking garbage.

[Eudyptula]

Michael I want to suggest to always make sure Intel blue on charts/diagrams and AMD red, orange etc. And of course, greens for Nvidia. This review is inconsistent in regards to this as the line charts and bar charts are flipped in confusing manner (I've noticed before too). The bar chart is blue for Intel but the line charts are red for Intel (hopefully the results aren't wrongly represented). For multiple SKU of the same brand, there's always light, dark and different hues to differentiate them (as well as solid vs dotted lines) .

It really helps legibility and a lot of sites does this. It also helps with people with ADHD (I one of those) and dyslexia etc. I actually read the blue as Intel and then realized I had to go through the article again to re-absorb the data again for it not to wrritten to neurons incorrectly.

---

Does not seem like a good part for a laptop due to power efficiency, but for a desktop it could be an acceptable solution for higher performance.

It's not that simple.

There are other analyses telling a complete different story (hopefully the Phoronix chart inconsistencies doesn't mean the results are incorrect). The Intel part is also newer (just released) and a year went by before they caught up to where they are, which is usually what Intel does in the mobile chip space (specially in terms of GPU performance, video decoding capabilities, multi-threaded performance and power-efficiency at load). So we shouldn't necessarily give too much credit to Intel here. It's just a matter of time before they are left to catch up again, unless they are to change a long-lasting norm (I hope they do change the norm).

155H is a 115 W "peak TDP" chip with 16 cores and a very different configuration. It seems to have issues with idle consumption as well. The 7840U and 7840HS are 8-core chips sitting at respectively 28 W and 35 W for TDP (although TDP aren't necessarily apples-to-apples across different chips or brands). There are both newer chips from AMD (albeit on the same architecture) and different SKUs that would be a more appropriate comparison.

Different laptops with the same chip can produce very different results and it's also dependent on power profile and tuning. So the very different results from this article to Notebookcheck's analysis makes me wonder whether the factor to blame is the laptops used being badly designed, a difference due to Linux, incorrect results due to a mix-up, unfavorable power profile configuration or a combination of them all.

---

I disagree. The Intel chip had generally less performance for more power draw, and higher power spikes.

While I'd agree the 7840U in Michael's test was clocking at a point on its volt-frequency curve that was slightly more efficient, the HS part could provide better absolute performance in a chassis with sufficient power and cooling. I'd also be curious how the BIOS is configured between the two machines and if they limit power on battery (showing a bigger difference in perf or battery life).

For this methodology and these products, best performance to power ratio goes to Intel.

But the article can only cover the given scope. There seem to be issues with the products used leading to this outcome. Or there might be a different explanation. At least there's a disparity here left unexplained.

[Michael]

Michael I want to suggest being mindful of making Intel blue on charts/diagrams and AMD red, orange etc. And of course, greens for Nvidia. This review is inconsistent in regards to this (which I've noticed before too). The bar chart is blue for Intel but charts are red for Intel (I hope the results are correct). When there are different SKUs, there's always light, dark and different hues.

It really helps legibility and a lot of sites does this. It also helps with people with ADHD (I one of those) and dyslexia etc. I actually read the blue as Intel and then realized I had to go through the article again because my mind had gathered the wrong information.

Yes Intel is typically blue. The exception -- as with this article -- is on line graphs. Due to my own API limitations and there potentially being the possibility of multiple "Intel" (or multiple "AMD" line graphs with a legend), they really can't all just be 'blue'. Just as the two AMD systems in this article for the line graphs need to be represented in different colors.

So for bar graphs and others without a key/legend needing to correlate, it's a lot easier to do the "color branding" but short of some great shading color algorithm for different blues or reds (I've tried but haven't come up with a good one), it's really difficult to do effective color branding for line graphs.

Edit: In case you're unaware, all of my graphs are 100% fully automated along with my testing... so it's not like I am manually picking random colors myself for graphs.

[Eudyptula]

Yes Intel is typically blue. The exception -- as with this article -- is on line graphs. Due to my own API limitations and there potentially being the possibility of multiple "Intel" (or multiple "AMD" line graphs with a legend), they really can't all just be 'blue'. Just as the two AMD systems in this article for the line graphs need to be represented in different colors.

So for bar graphs and others without a key/legend needing to correlate, it's a lot easier to do the "color branding" but short of some great shading color algorithm for different blues or reds (I've tried but haven't come up with a good one), it's really difficult to do effective color branding for line graphs.

Edit: In case you're unaware, all of my graphs are 100% fully automated along with my testing... so it's not like I am manually picking random colors myself for graphs.

Thanks for the clarification. Well, like I said, there are many different blues by adjusting hue and saturation. Even blue's cousin, cold-purple (whereas pinkish-purple would be red's cousin).

Have you thought about having sets of predefined color values for each respective brand, which are sorted in order (by most to least preferred)? The most "brand-authentic" and the most contrasting value sorted highest. Then for each "test subject" to be assigned a color value, one by one. At the point of list exhaustion you, of course, have no choice but resorting to other hues with low or no brand-correlation.

The color sets could also be defined as a range of hues (from value to value) that a certain branded "test subject" is in correlation with. You can also have some two-tone striped color pairs to make the color values list longer.

I am indeed unfamiliar with the benchmark tools. There's only so far my raft can travel on my lake of assumptions. I don't know whether test results are accumulated into a file or database and what limitations there are. But I assume it needs to be able to distinguish between different "test subjects" anyway (by which it groups all test results "owned by" each given "test subject"), so as far as I can row, I don't see an issue with the predefined color set solution. Though, my lake could be covered in too much fog for me to see clearly.

[darkbasic]

It would be interesting to know why my HP Elitebook 865 G10 sucks so much in comparison: https://openbenchmarking.org/result/...DARK-240119240