or Intel is lying, which wouldn’t be the first time by a long shot.
Isn’t it like desktops where the boost is simply limited by thermal headroom? So if the chip was cool enough it would reach ~60W until TAU expires? When testing I got up to the low 50Ws but immediately thermal throttling resulting in continuous dropping power down to 28W.
2 Likes
I suppose someone may one day test in a cold room or fridge, but clearly the issue is temperature related.
So a 60W ceiling isn’t commonly attainable.
On the bright side knowing it can do 60W means I can hopefully run 32W or 23W . . . .
1 Like
Think there might be a couple of ways to look at this:
- Design and tested to be capable of cooling the heat generated by the theoretical load. Or
- Design & hope to be capable of cooling the heat generated by the theoretical load. But not tested.
[Decent] Mobile workstations & gaming laptops fall into the first category, generally. With consumer laptops falling into the second category. Business laptops, somewhere between the two, depends on specific model and pricing brackets.
I know for a fact that some mobile workstations do boost up to 80w, and doesn’t throttle at all for the full duration of 56 seconds. But they’re way thicker / bulkier. For example, the P15 Gen 2, with 11800H. The 11800H can have a PL2 up to 100w+, but Lenovo limited it 80w OOTB. Different models have different thermal solutions for the thermal headroom and different actual performance expectation.
Thin & light laptops definitely have some compromises, and Framework isn’t immune to this. But the issue doesn’t seem to be around whether it’s powerful or not…but more about whether it can realistically, practically do what’s stated by Framework, “allows the CPU to run up to” 60w boost. Having the 60w expectation is not unreasonable if it’s stated in Framework blogs. IMO, what’s not quite reasonable is that it’s stated in the blogs, but can’t realistically deliver.
How’s that considered as “allows” if it thermal throttle during regular room temperature?
2 Likes
May be a good idea to show where you get the statement from as else it’s just hearsay. But as you point out, it’s disingenuous to say it can due to some other source, when it can’t.
So yes did Framework just pass on some specs without testing, as I imagine most people do most of the time.
So you have no actual quote from Framework that it can run at 60W?
I’ll have a look later.
nrp IS Framework’s CEO / Founder. Did you not click on the links? Or are you saying posts made by nrp with “Framework Team” and “Blog” tags still don’t represent “Framework” (the company)?
Here, screenshot:
1 Like
I saw this, below the link, so though it was not confirmation. Like I said I will look later. Thanks
I know that the DIY kits with the 12th Gen are shipping in August right now. But does anyone know when the mainboard and lid combination will be ready to ship from the markets place? I have seen and read tons of articles about this update but know none of them have mentioned when they will be available.
1 Like
Yeah, that! Also the EU marketplace availability, as I don’t want to pre-order whole laptop.
Worried about battery life, as first generation frame work where not really good at battery life + all the first reviews of 12th gen intel gives bad battery life.
- ZenBook 14 Oled : 8 h 40 min with 75Wh Vs 11th doing 9H46 with 63Wh !
- Galaxy book2 pro : 7H20 with 63Wh (values provided by lesnumerique tests protocol, playing video)
- Foreseen Framework 12th gen : 5H video playback with 55Wh
If the framework retains same battery pack (55Wh), it should loose at least 1H00, but more likely ~2H of battery life. (about 5H video playback in comparison to those 2 laptops)
Please let us know ASAP the expected battery life of the upcoming laptop.
Finally this wall 12th gen intel is weird, Big-littel is supposed to lower power consumption when giving the same workload, as those test gives the same video playing workload they should all see a decrease in power usage…but exactely the opposite happen… kernel optimisation maybe…Bottom line most of those manufacturers have increased batterie size of their models with 12th gen.
1 Like
Both 60W Boost and 30W sustained are indeed achievable with 12th Gen. Here’s an example just now of running Cinebench R23 under the Multi Core test. Make sure you have the power setting set to “Best Performance” in Windows and have power plugged in. How long the system will stay in Boost will vary based on your ambient temperature and system to system, but a few seconds is typical. That is, Intel’s Boost (PL2) is intended to improve performance and responsiveness for short bursts of peak activity, while the sustained performance (PL1) indicates the performance for longer loads.
11 Likes
Awesome. That’s good news. Is this the engineering unit, or pre-production unit? Also, was this while the unit is plugged-in with the Framework 65W power adapter? Or was this (screenshot) during on battery?
What’s the system to system variation if they were all running the same OS, drivers, software… Is there a QC check on the thermal performance of each unit as they come off the assembly line…and what’s the passing criteria?
Notebookcheck has the 1280p scoring 11480 points in Cinebench R23:
https://www.notebookcheck.net/i7-1280P-vs-i7-1260P-vs-i9-12900HK_14059_14064_14041.247596.0.html
The Framework unit seems to be currently around / almost 10% lower. Will the production 12th gen board be any better than what’s shown in the screenshot?
(Granted, the score difference could be the difference between ‘few seconds’ vs, say, ‘48 seconds’ tau capable units)
It’s definitely an improvement over the 11th gen (as Framework Support thinks it’s ‘normal’ not to be able to reach 60W TDP for the 11th gen).
1 Like
Another question, the screenshot shows Windows 10 with a build number of 22000. Does it make a difference?
Windows 10 doesn’t seem to have this build number out yet (Unless it’s Windows 11). Looking at the rounded corners, it does look like it should state Windows 11. Seems like Cinebench is reporting the OS incorrectly?
1 Like
Follow-up question @nrp if I’m reading Intel’s guidance correctly here Tau can be set by the OEM with a minimum of .1 secs and a maximum of nearly 8 mins, with a recommended setting of 28 secs. Exactly how long have you guys set Tau for?
This does not encourage me too much as it implies that even the 28 secs recommended by Intel is not achievable. Instead being limited to what I consider a “few” seconds (like 6-7 secs). Given that the Cinebench run you posted only seems to have just started 5 secs before the screenshot, can you elaborate on this?
1 Like
Just finished lapping the heatsink of the Framework laptop, and it seems like there’s a problem / issue with it from the factory, at least on my particular heatsink.
While lapping the heatsink, it became clear that the contact plate is dished / not flat. This created a poor contact point for 1 of the 4 cores, causing this particular core (physical core) to heat up before the other 3…subsequently causing thermal throttling to kick in.
After lapping (still not completely flat yet…that dish is deep, about 2/10 of a mm it seems), the processor can momentarily reach 59.1W peak for a split second.
Background, I used the Shapton Glass Stone (just a whetstone really) for 1000, 2000, 6000 and 10000 grits.
3 Likes
Tau is 28seconds on the Framework Laptop.
But you get something like this:
The area in red is the potential performance lost due to thermal throttling.
The processor doesn’t reach 60w (my unit, 11th gen), and the wattage will continue to drop as the processor is trying to maintain the temperature to under 100c. The duration of the PL2 is 28secs, then PL1 kicks in…dropping to 28w, sustained.
The most ideal / capable laptops would have the red area in blue, full 60w for the entire duration of tau. But that’s not the Framework laptop.
The fan speed profile doesn’t seem to change in Windows between the different Power Modes. Ideally, if the fan spin-up can be more responsive / aggressive in “Best performance” mode, that could change the slope gradient of the thermal throttling period…to something a bit flatter.
3 Likes
Correctly and incorrectly. The major version of the product as reported through all of the version-related APIs (as far down as RtlGetVersion) is 10. The marketing version is not reported through any APIs.
2 Likes
So Windows 11 is both Windows 11 and Windows 10 at the same time…that’s one way to do quantum computing…
2 Likes