[Bug Report] CPU Detection
#11
Interesting, CPU-Z also doesn't report SSE4.1. That would imply a system bug of some kind.

[Image: cpuz.png]

I'll have a look at my system and see if I can figure out why it isn't reporting SSE 4.1, since by all accounts, including official Intel documentation, this CPU does have it. This is a retail Q9550, for those wondering.

At this point I wouldn't spend time looking at your emulator as it is highly likely a problem on my system.
Reply

Sponsored links

#12
Very well. Moved for now to general forum.
Reply
#13
One thing that jumps out at me is that the intended multiplier for your CPU is x8.5, not x8.0. You're running at 2.66ghz instead of 2.83 ghz, effectively making your CPU the same as a Q9450. Dunno if that's a limitation of your motherboard, or if your CPU is also reporting it's multiplier information incorrectly to the Bios (yikes!).

Additionally, I've added a new double-checking mechanism to Pcsx2's cpu instruction set identification, which is more reliable than using Gsdx SSE4 builds (since those seem to work on non-SSE4 machines more often than not). So the next beta of Pcsx2 (or compilation of trunk/ if you have the ability to compile the binary yourself) will tell you explicitly:

* if your cpu really does support SSE4.1 (buggy CPUID)
* .. or if your CPU is somehow unfortunately without SSE4.1 (wth?)

In either case, I'd hope it's still under warranty! Because neither behavior is correct, and in both cases Intel should replace it with a non-defective one.
Jake Stine (Air) - Programmer - PCSX2 Dev Team
Reply
#14
And the solution has been found: Reflash the BIOS. Notice I didn't say update the BIOS, I was running the latest version and the Intel Flash utility reminded me that I was flashing the version I already had.

The actual problem was the Intel Management Engine firmware, believe it or not. Something had bugged out with its configuration and was causing a problem. A reflash reset the configuration, and thus has cleared up the problem. Both PCSX2 and CPU-Z correctly note the presence of SSE4.1 now. That was also the problem for the misreporting of the multiplier. The CPU was operating at 2.8GHz, at least it was producing Prime95 speeds consistent with other Q9550s, it was just reporting things wrong. It now reports a correct 8.5x multiplier.

My apologies to all for wasting your time, it never occurred to me this was a system error, especially since it seems that SSE4.1 was working fine (since I was using the SSE 4.1 graphics plugin). In the future I'll be more careful to audit for system errors before making a report.

Thanks for all your help and suggestions, especially since it wasn't a problem with your code.
Reply
#15
Well, glad that you solved your problem AND that we didn't have some lingering bug Smile
Reply




Users browsing this thread: 1 Guest(s)