Per request, moving benchmark/performance discussion here
Previous thread: FW13 Pro Batch 1 Guild - #50 by GreyXor
Thanks
So I was saying, for reference my FW13 HX 370 got 16246 points
Since you are on windows 11, can you increase the power limit using xtu?
My 7840u fw13 can cool 43w sustained and a lot more peak at <95C so the pro doing barely 35 at 100C is a bit concerning. Of course they may just be measuring different temperatures but still.
Also how about some real-ish world testing like some games? Thatâs where pantherlake should really shine.
GB6 tests are not running into power limits on Windows. The PL2 is 60W and is never hit.
Tested some games, mostly on Linux and they are not very great - there are so many issues that needs manual fixing and some games outright wouldnât work despite trying everything
But on Windows games run very well
One factor is XESS on windows vs FSR on linux, and then I think without upscaling many games are just not as performant on Linux or running into scheduling issues.
Never put much value into geekbench results myself, it runs on everything but thatâs about all it has going for it. Have you tried cinebench? Would be less bursty but not quite as brutal as prime95.
Since this is a very new architecture that is unfortunately kinda to be expected, will likely improve pretty rapidly now that hardware is actually entering peoples hands.
Iâd also be curious about some more power draw tests, whatâs idle battery draw on linux at 20% brightness and whatâs the battery draw doing 720p (30) youtube playback in firefox? (for reference my 7480u uses 3.6W and 7.2W respectively).
let me try with cinebench when i got time to try with windows again
Unfortunately I think thereâs no way to really tell - I donât have a power meter and the power draws reported by s-tui is either âpackageâ which I think only accounts for the CPU, or âpsysâ which frequently exceeds 60W which I think is inaccurate for everything combined. We want to look for something like CPU + GPU combined when using hardware decoding.
The one in the battery comes for free, on windows you can use hwinfo64 to look at the reported battery drawn, on linux you can read it using upower.
The reported powers by the soc are incomplete and also likely donât measure exactly the same on amd and intel.
Cinebench:
Also running cinebench MT hits dynamic thermal throttling down from 35W PL1 to 32W.
XTU says âunsupported platformâ
Just my observation over the years with various laptops and processors:
Short burst benchmarks tend to really test 1. the TIMâs ability to move heat, and 2. to the heatsink to absorb *. i.e. The fan doesnât come into the equation at this point.
So, thatâs kind of what GB6 is testing (relatively speaking, coming to long running tests like CB)
*But you donât want the heatsink to âholdâ onto too much heat. You want the heatsink to move the heatâŚthermal diffusivity (different from thermal conductivity). Purity of copper used come into play.
HumâŚthe score is low compare to whatâs out there. Between your system, and the Dell XPS 14, on paper, the difference is the 60W vs 69W PL2. In long running benchmarks, the Framework should win out with its 35W PL1 (XPS only has 30W PL1):
Really wonder whatâs going on.
The FW13 doesnât have PL1 advantage as it throttles to 31-32W.
Also FW13 cannot utilize the PL2 at all for the same reason.
Itâs really not surprising, not to mention ram frequency is also much lower.
I wonder whether this is where the pro chassis comes in. IIRC, you have the new board in the old chassis, right?
1 Like
yep, old chassis. Doesnât look like new chassis is thicker with more breathing room and heatsink is the same.
The result and never hitting pl2 allmost makes it sound like they arenât configuring the powerlimits right in the ec, would not be surprised if this thing got a bios update real soon.
Also possible you just got a dud with with bad ptm application, I hope weâll get a few more people with devices in hand to test.
Cinebench usually doesnât care about ram all that much.
I kinda doubt it it uses pretty much the same airflow path as the original case if I see it right.
Pretty sure it only looks thicker because the removed the chamfers to fit the battery and speakers. Donât think there is any more breathing room there.
Much more, then points in some benchmarks, I am interested in real-world-performance. TBH I also donât get the point of those Geekbench benchmarks. Even the exact same machine is there with so totally different results. So I think, without further information these benchmarks donât give much information to me.
Also I donât know why the Watts are so interesting? If the machine does not reach itâs maximum power it might have some unused potential, ok. But more interesting is the performance, not? Best cast is has better performance with lower powerusage. Well, yes⌠It it reaches itâs max power there will be even more performaceâŚ
Do you have comparisons for games or other real-world-workloads? On the same platform and settings of course.
As is everyone else but that is incredibly hard to quantify
On the low end itâs an indicator for efficiency, on the high end it indicates how well the cooling works.
Standardised (ishâŚe.g. excluding environmental factors) testing: Given per $ amount how much performance can I get for a pre-defined / repeatable set of instruction execution.
e.g. X amount of dollars purchased a race horseâŚis it running faster than the next horse on the race track (pre-defined, repeatable metrics).
Also, in conjunction with not all 358H CPUs are equal, not all TIM applications are equal, not every heatsink perform equally well. Every component coming off an assemblyline has itâs own variation distribution. The laptop, assembled from the components has a performance distribution / spread. The tighter the tolerance (performance groupings), the higher certainty / assurance of knowing what performance range one can expect from a particular purchase.
For example, if I told you that each laptop (of the same brand and model) is factory-tested with a performance index, and the index is disclosed to you at no additional cost to you, would you want to know / purchase the particular unit that has the highest scoreâŚor would you go with âDonât understand the need, donât care about the index. Iâll take my chances with real-world execution, and randomly pick whichever one comes to meâ?
@Zack1, think your systemâs wattage is fine if youâre seeing 31-32W. The specs sheets says 30W.
I was thinking about the intake and exhaust grill sizeâŚor are they all the same as well?
Edit: NevermindâŚlooks pretty much identical to the non-pro chassis:
yep, maybe FW will update the bios to more strictly enforce the sustained TDP as currently itâs 35W for both performance and balanced mode 
Just to make sure, what charger are you using? I have dug relatively deep in the ec code of the amd boards as far as power limits are concerned and the charger wattage plays a big roll there.
65W PD (20V 3.25A)
I said âcannot utilize PL2 at allâ because in cinebench the package reaches thermal dynamic throttling very quickly.