Howdy fellas. When I keep my eye on CPU tech, I tend to follow the ratings here:
https://www.cpubenchmark.net/singleThread.htmlUp until a week or two ago, the top slots on these charts were a pretty mixed bag between AMD and Intel, but it seems like PassMark upgraded their tests and now Intel is utterly dominant again. Does anyone have some insight into why this might be? Did Intel recently find a fix for the Spectre/Meltdown bugs which allows them to re-enable speculative execution or something (without the performance loss, I mean)? Did a bit of googling and didn't find anything conclusive. Is PassMark just becoming an Intel fanboy?
None.
We can't explain the universe, just describe it; and we don't know whether our theories are true, we just know they're not wrong. >Harald Lesch
I was trying to find some specific explanation and not rely on people posting about made-up bribe money
(unless there was some solid source for Intel bribing passmark but the tweet doesn't seem to provide anything close to that). I saw the passmark forum responses, but again they're kind of vague and while their admin does list some specific differences it doesn't seem very concrete.
None.
We can't explain the universe, just describe it; and we don't know whether our theories are true, we just know they're not wrong. >Harald Lesch
The gist is Passmark changed their method of calculating the CPU score. What exactly changed they won't say, except that AVX-512-instructions are now factored in. AMDs don't have those, but most newer Intel High-End CPUs do. Outside of very specific applications that vectorize the computations there's not much use to this feature. So either they overvalued that specific test to get such crass differences or they did other stuff we don't know about. You connect the dots.
But no, there's no concrete evidence as to why Passmark did it.
EDIT:
And comparing the test results of Threadripper 2990WX and Xeon W-3175X suggests that higher memory bandwidth and faster caches (which the Intel has) are also valued higher now.
Their admin seems to feel it was quite justified:
https://www.passmark.com/forum/pc-hardware-and-benchmarks/46748-cpu-benchmarks-huge-changesAlso, if it is well-justified, one could argue that Passmark has been unintentionally favoring AMD for the past few years for no reason.
None.
We can't explain the universe, just describe it; and we don't know whether our theories are true, we just know they're not wrong. >Harald Lesch
Their admin is paid to feel that way. At no point he explains which of these methods favor Intel and why they decided it was necessary to remove the bias against AMD you're arguing. He basically says: "Look we did some changes, funny how that made Intel win. We didn't even notice, pinky swear!"
So yeah it could be incompetence, and not sponsor money.
At the very least this shows how useless synthetic benchmarks are.
