Here is a cheap quad core system that I have found runs Haiku R1 Alpha 4.
I have added it to he Haikuware Hardware List.
Mine is the Acer Aspire AX1300 Quad Core Phenom running at 2.3GHz.
AX1300-U1802A is the actual model
Video is only VESA but this system does offer two slots so I may pick up a different video card and see if I can get better video. But VESA does 1600x1200 on VGA port. HDMI port only 1280 x 720.
The NVidia Sound did not work direct from Haiku so I had to download OpenSound. That does work with the NVidia sound on this system.
Network port works fine.
Overall for a sub $100 computer, it works nicely and is Quad Core!
TJ
Thank you for the details, TJ.
I don’t think so but not really sure.
I would think if hyperthreading was happening, I would see 8 pulses on my Pulse applet rather than just the 4 I see now.
TJ
Yes, you would. Have you checked your BIOS settings?
Some systems let you turn HT on and off. And while you do not always see a big speedup with HT, the support for more threads makes the machine run smoother under heavy loads.
It’s got a Phenom, that doesn’t have SMT of any kind. AMD introduced their approach to SMT in the Bulldozer design for the FX series of CPUs.
So what happens if you have a dual core CPU with hyperthreading and you run two CPU-heavy single-threaded apps like Teapot?
Does the scheduler know to assign the threads to the two “real” CPUs? Or will they bounce around between the four virtual CPUs, spending half of their time actually running on the same physical core with the associated performance hit?
[quote=ASoftwareHatingFurry]
Does the scheduler know to assign the threads to the two “real” CPUs? Or will they bounce around between the four virtual CPUs, spending half of their time actually running on the same physical core with the associated performance hit?[/quote]
The latter, although statistically it’s actually about one third of their time. Haiku doesn’t have any CPU topology awareness and very little soft affinity so the threads will “bounce around” but in fact the chance of both ending up on the same physical core is only 1-in-3.
It would be interesting to benchmark that to see what the real-world performance impact is on typical software.
One can imagine a scenario where an app is using just two busy threads (say a Web browser with its shadow thread in app_server) which might run faster with hyperthreading turned off. But on the other paw, I suppose the “virtual” CPUs share cache memories which could outweigh the performance hit.
It seems like the best stop-gap would be to modify the multiprocessor initialization code to detect a multi-core system with hyperthreading and disable the virtual CPUs.
One clear benefit would be allowing Haiku to work correctly on today’s CPUs without waiting for a new API call to be implemented. The 8 CPU limit is already breached by an inexpensive single-chip 6 core system! Perverse multithreading indeed.