Yeah that's the thing, you can find odd scores for all CPUs that are clearly outliers, but this is something different. So many of the scores are low - even seemingly the majority of them so far. We just don't know what the issue is yet and why - is everyone testing on low power mode? they imported spotlight indexing from the Mac just for the fun? a firmware bug stopping the clocks from ramping up? If these are review units then hopefully the reviewers will tell us what is going on and, maybe, why.
Oh yes I remember the Anandtech article on that.
Aye.
So when I look through the list, one thing to keep in mind is a majority of units (or plurality anyway) are going to ship with the 3.4GHz max clock, which has a GB6 ST of like 2425. So the 2350+- 100 is very easily “explained” by that and probably throwing people off assuming all the SKUs are the 4GHz ones (or 4.2 tho I suspect that one is rare).
———
X1E-84 = 4.2GHz ST/dual-core boost, 12 core, but 3.8GHz all-core permitting thermals probably.
—————-
X1E-80 = 4GHz ST/dual-core boost, 12 core
X1E-78 = 3.4GHz max, 12 core
X1P-64 = 3.4GHz max, 10 core
These three ^ all having 3.4GHz all-core.
The X1E-84 with a 4.2GHz Dual and a 3.8GHz all-core I don’t expect to be that common because demand for the X1E-80 will usurp it and I suspect they’ll just shift some dies down, the tradeoff from 3.4 to 3.8GHz for the all-core mode really adds too much power whereas 4GHz for dual-core could be considered “worth it” for many and still for typical 15-45W parts. New XPS for instance uses it (the X1E-80).
But anyway, that divergence between 3.4 and 4.2GHz, and Samsung’s results with the 1800 stuff explain a lot of this from the search.
Like most of these for their respective SKUs make sense — unless they are totally off the rails. But you don’t see like a 4GHz one performing exactly at 3.4GHz ST, and the scores seem to cluster well.
You basically have either “total firmware mess/testing” scores — almost invariably below 2000 it seems and near 1700-1800 or even worse — or you just have the variation from the different SKUs.
If you sift them into those two categories with everything above 2200 being working, suddenly binning them into X1E78, X1E80, X1E84 makes complete sense and they cluster remarkably well and you don’t actually see (IMHO) even as much wacky stuff as you do with Ryzen or Intel CPUs. I’m going to bet this is bimodal between working/non-working, and the confusion about the working ones is just a misconception about the frequencies.
Actually yeah, the more I look through these, the results *after* the firmware thing (dominated by Samsung but Asus and Lenovo have some doing that too) are extremely easy to tell which result is from which SKU of the 3 on average (or 4 if you include the 3.4GHz plus).
x.com
