Haven’t seen anyone post this yet. Notebookcheck.net reviews are some of the best in my opinion. With very detailed testing, providing hard numbers and easy comparisons versus other laptops.
As expected, much better battery life with AMD. Plus better multi-thread performance and twice the graphics performance.
Runtimes are longer than the Intel-based configurations by significant margins due to the lower power consumption levels mentioned above. We’re able to browse the web for over 3 hours longer on our AMD model versus the Core i7-1370P model when both are set to Balanced mode and 150 nit brightness.
Don’t be fooled by the fancy charts! Notebookcheck is good all-around, and especially thorough, but their battery life test is just scripted web browsing over wifi.
Ars ran the PCMark modern office battery test which includes writing, browsing, and video conference. It’s a real benchmark from a company that makes benchmarks. Not in any way less legit than notebookcheck’s in-house test, and it tests a more diverse use pattern that might better reflect real world performance.
Probably the big difference is video conference. I’d guess the Intel video hardware is either more efficient or better utilized by the software in the test. Depending on which of those is true, the PCMark test may be a much better indicator of battery performance for many users. It might also be that the e-cores pay off in office tasks, specifically. Someone would have to follow up in more detail to really get to the bottom of this.
On one picture in the Arstechnica review there seems to be an USB-A card in the top right slot. I don’t know if the test was performed that way or if this can explain the big difference, but it is a possibility that this caused higher than necessary power draw.
That’s the RAM they’ll be shipping in preassembled machines and with DIY orders that include RAM. So far the only DDR5-5600 sticks that the manifacturer unambiguously declares can run at CL40 without XMP/EXPO/PnP (which these CPUs don’t understand) seem to be the G.Skill ones, and they’re not on the compatibility list in the KB (hardly any non-Framework-branded modules you can actually find retailed are on that list, though, so they’re not special in that respect).
Marginal at best. We’re talking a 2ns difference in first word latency between CL40 (14.29ns) and CL46 (16.23ns) while keeping the same bandwidth and speed. For the vast majority of use cases the 2ns is virtually insignificant other than scoring a few extra points at some unrealistic (for type of device) synthetic benchmark workload.
Of course, there exist legitimate workloads with memory access patterns where the lower latency would be advantageous. But the counterargument here is if one’s workload is really that sensitive to your memory latency, you probably wouldn’t be doing it on a laptop, certainly not an ultrabook.
If you look at consumer desktop DDR5, most modules on the market today tend to default to 4800MT/s if not using their XMP/EXPO profile. Even at CL40 that’s 16.67ns latency, with higher speeds and CL ratings usually achieved with XMP/EXPO profiles (e.g. 6000 CL30). Of course, this might change as DDR5 becomes more ubiquitous and technology improves.
In my view, Framework’s choice of DDR5-5600 allows for a good balance between speed and latency, be that CL40 or CL46.
If you’re curious, Crucial have a very good blog post on speed vs latency when it comes to DDR5, as well as the main advantages of DDR5 over DDR4 and why the former can perform better even at comparatively higher latency.
This is interesting - I expected generally more balanced performance overall, including in more GPU heavy tasks.
We’re currently working towards looking at alternatives to Dell for general purpose CAD work, and the AMD Framework boards offer a plausible and attractive alternative to the relatively power hungry and heat inefficient combo of Intel + Nvidia DGPU (yes, the 13th gen Intel chips are heat efficient, but the Nvidia mobile GPUs really aren’t).
There are a few barriers to this…some bureaucratic, some budgetary (Dell offer cost effective RTX workstations that are great on paper, but terrible in real world performance), but Framework are a good option in terms of the ease of deployment and sustainability feel-goods.
The big one for my lot though is TB4 support: we have a tonne of Dell thunderbolt 4 docks that may or may not work with the Framework AMD boards. They work okay with my 12th gen Intel Framework, that’s TB4 compliant…but has anyone had any joy with TB4 docks getting dual display outputs and PCIe data channels through the USB-C ports on the AMD boards?
I’m so torn right now. Lenovo at the moment has a ThinkPad P14s with the same 7840U (Ryzen 7 PRO), 64GB RAM, 1TB SSD, 2.8k OLED panel on sale for CAD 1,529 (but the battery is only 52.5Wh). There’s no FL13 AMD end-user reviews yet. Which should I choose?
If I were to get the FL13 AMD 7840U DIY, no RAM, no SSD, no OS, no USB-PD adapter, with 4 USB-C cards…it would still come to CAD 1,627. Hum…
On one hand, I’ve been waiting for the FL13 AMD for so long, and it’s repairable. On the other hand, I know I can (almost certainly) count on a ThinkPad.
Reliability, on-site support (one year…can extend up to 5 years on-site if needed) globally, chassis / lid rigidity, only 1 x USB 4 port though, built-in ethernet OLED is nice to have…not a must. Good track record with BIOS update frequency (in general).
Framework has great support over email and marketplace, and I can potentially re-use the chassis going forward. But the hinge bounce is a concern, bottom chassis is a tad bit on the soft side as well.