DOS ain't dead

Hi,
I got a new PC at work. It's some HP model with intel core i5 750, 8GB RAM, intel H57 chipset and nVidia GT230 VGA. Performance is nice (for singlethread apps my home overcloked E8400 is much better :) but I'm not much happy with mobo peripherals. It has only 4 PCIe x1 + one x16 slots - no classic PCI. It also lacks COM/LPT (we had to buy extra PCIE x1 card), no IDE, no floppy port. So near completly purified of legacy stuff except the BIOS (with poor settings - as usual on Dell/HP mobos). So now I have a chance to test various DOS SW on this machine. Someone here may be interested in my short review to be prepared for future HW upgrade.

I repartitioned HDD and installed FreeDOS 2038 with JEMMEX on FAT32 partition without problem. SATA has BIOS options: AHCI, RAID, IDE. DOS can boot with all this settings becauseof INT13h extension. It reported about 3,3GB available XMS. BIOS provide USB keyboard and mouse emulation, ctmouse detected it as normal PS/2 mouse. I tried to run few utils. MPXplay detected HDA but I didn't have any mp3 there neither phones attached. My CPUID crashed when tried to read core temperature through MSRs so it will need some update. My VESATEST run but it shown very poor graphics performance in VESA BS and even LFB mode (tested 1024x768/32). I tested without and with write combining set by my MTRRLFBE. Results:

BankSwitch, MTRR unset: 28 MB/s (I got same value under WinXP)
BankSwitch, MTRR set to WC: 41 MB/s
LFB, MTRR unset: 29 MB/s
LFB, MTRR set to WC: 29 MB/s

As you can see it helped only for BankSwitched mode but LFB mode has no performance gain, WHY?

On my home system with intel P31 and GF7900GT (it seems that VGA type is not important) I got about 2600 MB/s, it's near 100x more. And good AGP system still make thousands of MB/s. So I'm disapponted but it may be specific only for this system. If someone else can test it and compare on different i3/5/7 system it would be helpful.

And here's PCI devices log

---
DOS gives me freedom to unlimited HW access.

Today I also tested my old VGA GeForce 7600GS in this machine and got exactly the same result. Just to be sure I also tried Japheth's VESAMTRR but I think we use same alg. to set same MTRRs. So now I can say this issue is not releated to VGA. But it still remains to distinguish the cause between CPU, chipset and possibly poor HP BIOS.

Anybody else have access to core i3/5/7 machine? If so, please test it, to know if it's generic or specific issue.

---
DOS gives me freedom to unlimited HW access.

I don't have access to any core i3/5/7 machine.
However you could test the graphics performance also under Linux with VESA output driver. It could show slowdown caused by some bad default ACPI setting for graphics subsystem.

---
DOS-u-akbar!

> I don't have access to any core i3/5/7 machine.
> However you could test the graphics performance also under Linux with VESA
> output driver. It could show slowdown caused by some bad default ACPI
> setting for graphics subsystem.

I could, but I don't know about any VESA benchmark for linux. If there's some live CD with it please tell me :) BTW as I write, I got same bad performance also under WinXP running program in NTVDM fullscreen as under DOS (I tested only bank-switch because LFB is not accesible from DOS under NT-based). I would expect that Windows always do the MTRRs and other needed settings for me...

---
DOS gives me freedom to unlimited HW access.

Maybe you find something related in this document.

ftp://download.intel.com/design/processor/specupdt/320836.pdf

---
DOS-u-akbar!

> Anybody else have access to core i3/5/7 machine? If so, please test it, to
> know if it's generic or specific issue.

Surprisingly, I did gain access to a Lenovo Core i5 650 (Westmere, Intel gfx) desktop recently. But if you want me to test something, you'll have to be a LOT more specific (what .EXEs, settings, etc).

P.S. Did you see me mention "unrestricted guest execution" (real mode, big real mode) under VT-X? Would be interesting if VirtualBox supports it, but I don't have any decent FreeDOS images here yet (and 1.0 is just plain old by now).