"Today the review embargo lifts on the first AMD-powered Framework laptop. There's one of the AMD Framework laptops in the lab for Linux testing and benchmarking but unfortunately no review for launch day due to being held up by a BIOS regression and thus unable to properly utilize accelerated graphics until a new BIOS revision is made available in the coming days."
"Long story short, in a week or two should be the Linux review and benchmarks for the new AMD Framework Laptop."
These 7840Us[0] on the Framework 13 are performing surprisingly well compared to other laptops carrying the same CPU on Geekbench. Almost as good as the 7940HS[1].
If you read the whitepaper[0], you'd notice quite some weird scenarios. Take the compression (it uses zstd, which a major prop), however it compresses some 10k small files of a total 75MB. Compressing small files individually is just pointless, ztsd is not an archiver (I suppose the compressor/decompressors are re-created for each file, too). The total size is just tiny as well. Then it uses a virtual file system with AES in memory, no idea how the latter is implemented but I suppose it lays entirely in the userland.
Due to small sizes, the workload would be very L2/L3 sensitive - however it doesn't represent any common way that compression is used in practice, esp. zstd.
I don't see why a CPU test should be only about number crunching, and not about RAM and cache access efficiency, provided that the test is not incurring gratuitous cache trashing.
Also, working with a ton of small files is a typical real-world task.
>Also, working with a ton of small files is a typical real-world task.
Yes, but not compressing them individually, and while staying in the userland (and having SHA1 on each, just for kicks). The real world scenarios tend to be make a big archive (e.g. tar) and compress that thing.
>and not about RAM and cache access efficiency
It's about the size of the cache, workloads that fit L2 vs such that don't, exhibit an amazing performance boost. Pretty much, the performance drops off a cliff when it doesn't fit the L2.
Overall microbenchmarks are extremely/notoriously difficult to get right, and more often than not, they are gamed. However the compression/decompression of geekbench is just bad.
I agree with all your points and also agree that honestly, these scenarios aren't far off from real world tasks.
I get the main issue, which is you could adjust the workload by 10% and achieve a 50% performance loss when you do this at the point where we cross the cache threshold and whatnot.
However I see CPUs unique in that I rank them _for_ these scenarios. A particular might be ranked unfairly, but as long as the test is equal, the better one is infact better, just not by the 50% the test might show but it's still going to be 5% better. I expect my GPU to be idle when it isn't training AI or rendering frames, but for the CPU, it's general purpose in real life, and anything goes.
>just not by the 50% the test might show but it's still going to be 5% better.
Sort of, indeed. Yet, when you see any promotional/marketing material - you see all those phallic bar graphs, and how much bigger it is. Other than that - heavy cache utilization hides inferior memory subsystem (latency/throughput), the latter tends to be quite important in the real world. Overall benchmarks/tests that feature handful of MB as datasets, and run in 100s of ms - should not be used as representative... for most use cases.
Well, they usually ship unlocked CPUs, where other vendors often power-throttle the CPU.
I am not a "fan" of the fan noise, so I checked and the EC firmware was open source. It was possible to raise the fan tipping point to run the laptop hotter and more silent.
People have also tried to set the power clamping later. That works and is equivalent to limit it in the BIOS.
Key point is: this does not mean you have to run it that way. Add a power profile, l configure the fan curve, enjoy.
Fan noise and short battery duration were my reasons to return my Framework 13 12th Gen. I believed they were related: CPU running on max (without any reason), thus fan spinning high and short battery. On Windows I couldn't find any way to solve this issue, response from support "works as expected". Is there now any way on Windows to limit CPU such that the fan does not need to run?
Those power figures make no sense, I'm not sure I get it. How could it run for more minutes than the battery size divided by average power draw would imply?
That is be power draw at the outlet [0]. There is probably loss that occurs when drawing from an outlet that does not occur when drawing from the battery.
There is loss to convert AC/DC, likely low 90s efficiency. Having an in-line DC power meter should be a very simple task, measuring the draw power from the outlet is rather pointless.
On the bright side as it gets old, it will be easy to clean out the fan or re-do the thermal paste on the Framework!
We have the first gen framework and while we haven't noticed any fan noise issue, its used to write a novel mostly, so we wouldn't be heating it up anyway.
The smallest 12th gen intel here. Yes, the fan is annoying. Reviews seem to suggest the 12th gen intel is the worst of the offerings, but overall I still like the laptop.
Switching power modes (in windows, haven't ran linux on it yet) does wonders for the fan, but can also destroy certain workloads. I often run DJ software and I can't use Rekordbox in eco modes. Have to be in ballanced or performance. Note that none of this is surprising. DJ software is not lightweight. With it running fan spin is painful in performance mode, but doesn't bother me in balanced.
I can't use some of the more power hungry features in Rekordbox without fan noise issues. Stem separation for example (newer features that separate vocals, drums, melody etc). I never expected to, those need beefy laptops, and I went for the weakest model (for unrelated reasons). It works in performance mode but the fan is really killing me.
Yes. I know most people are unfamiliar with DJ software, but it's the best example I have. And yes, since we're talking DJ software, the music obviously overpowers the fans when there are speakers and a crowd around. I notice the fans at home when practicing.
[Edit/tldr]: Battery saver / eco modes = no fan issues. Balanced mode = fan, but ok. Performance mode = fan is among the worst I've heard on a laptop. I still recommend the laptop overall.
I have a 1st batch framework (the first batch of the first laptop version) which I haven't upgraded or changed at all yet. (Hinges are on the list tho!) The fan engages sometimes but it doesn't bother me in practice FWIW. I'm a computer musician who uses the framework for DSP processing, but I'm also someone who is maybe more tolerant of noise than most so YMMV.
My base model Macbook Air M1 have a Twitch stream, a few dozen browser tabs, VS Code, and docker container running simultaneously with no fan (and no noise obviously), but I try the same tasks on my Framework and it's blasting the fan and the performance goes to shit as it starts thermal throttling. :/
> Framework laptops seem great but the fan is a tad too noisy.
Half that statement is true. They're not great, for a variety of reasons, and the fan is quite noisy.
The main thing I learned buying a Framework laptop and a Macbook Air M2 within a year of one another is how low I rank repairability among my priorities.
Don't forget PTM thermal pads! They have a respectable 8 W/m•K conductivity and don't pump themselves out or dry out like thermal paste. They've become popular with gaming laptop companies and users.
8W/m.K is pitiful as they are much 'thicker' than thermal paste; good pastes should have higher thermal conductivity to boot as well. If a solid pad is needed in place of thermal paste - carbon pads, e.g. "Carbonaut" is a lot better solution. Carbon sheets have higher base thermal conductivity than copper, of course not so high when polymerized.
I don't know anything about it, however the power consumption is non trivial: 5.25 W (net 4.25 W) Maximum noise inside device at 50 cm 21 dBA. Maximum power consumption 1 W. A laptop fan is around 2.25W or so for the whole operation and it can definitely dissipate more than 10W produced. A laptop will need 5 or so, more if it has any discreet GPU.
Funny note of the site - everything is metric (even pressure is Pascals, distances for the noise is cm - 50), but the airflow is in cfm (cubic feet per minute)
What countries in Europe do they even deliver to? I'm in EU, had to buy mine through a friend in Canada and pay double VAT because there is no logistics.
That's kind of strange too, the notebookcheck article shows the Ryzen version having a much larger runtime in their benchmark compared to the i7-1370P version.
Search for "Battery life", you'll see WIFI runtime at 726 vs 527 (I assume minutes) for Ryzen vs Intel respectively.
Ah, it was already mentioned in the previous thread too, that Ars review team had the ports configured incorrectly which resulted in higher power draw.
The alleged proof that they configured the ports wrong is a picture from their Intel framework review. "We're using the same pictures as a previous review because it's the exact same laptop." You can see the same picture here: https://arstechnica.com/gadgets/2023/05/review-framework-lap...
The Ars review also even mentions this problem: "Framework also says the rear ports enter a "high-power mode" when USB-A modules are connected to them, which can reduce battery life."
Either way, glad we have a new product we can blame the user for holding wrong.
The notebook check power figures make no sense when you consider that their measured runtime was longer than what would be possible if the laptop was at idle the whole time. Their average idle power draw was at 6.5 watts, the battery is 65wh and the overall runtime is... 700+ minutes? What am I missing ?
Even if we assume that the test was completely at idle, Intel seems to draw less power than AMD when idling so I'm still puzzled by the run times.
As per Wikipedia article, it is not the same thing. The on-die ECC is needed on DDR5 to reach the reliability of previous non-ECC ram. Real ECC-DDR5 additionally exists for higher reliability.
I think you're getting the reliability even from on-die ECC if there's memory corruption. What you're not getting is warnings that the RAM is bad and that you should replace it. So it's just as resilient as officially ECC ram but you don't know if error correction is happening or not.
"AMD Ryzen Powered Framework Laptop Linux Testing Held Up By BIOS Issue"
https://www.phoronix.com/news/AMD-Framework-Linux-Hold
"Today the review embargo lifts on the first AMD-powered Framework laptop. There's one of the AMD Framework laptops in the lab for Linux testing and benchmarking but unfortunately no review for launch day due to being held up by a BIOS regression and thus unable to properly utilize accelerated graphics until a new BIOS revision is made available in the coming days."
"Long story short, in a week or two should be the Linux review and benchmarks for the new AMD Framework Laptop."