I tried with code from gtx520, and it works!!

vesatest shows correct image and page 1/2 indicator

but...
duke3d runs, but it seems that it using vsync , and i get lower fps , compared to mskvbef7
need your feedback

LSS10999 wrote on 2023-09-01, 13:38:

Tried your second PSAVE on my system. Looks like it's producing something useful. I'm getting a 395-byte PMTABLE.BIN file with more contents.

Yes, that is
Thanks!

It slightly different from that i get from gtx520 , will look at it later

Can you test dike3d on your card with mskvbef7(from 1st message in this thread) and without?
Compare fps pls (type 'dnrate' while in game to show)

an64 wrote on 2023-09-01, 16:07:

I tried with code from gtx520, and it works!! […]
Show full quote

I tried with code from gtx520, and it works!!

vesatest shows correct image and page 1/2 indicator

but...
duke3d runs, but it seems that it using vsync , and i get lower fps , compared to mskvbef7
need your feedback

Yes, it is using vsynch when page flipping is active but I think it is normal. Both the real-mode and the protected mode interface provide explicit vsynched HW page flipping mode (Function 07: BL = 80h Set Display Start during Vertical Retrace).
Build engine games and Quake most likely use this mode to prevent tearing.

an64 wrote on 2023-09-01, 16:09:

Can you test dike3d on your card with mskvbef7(from 1st message in this thread) and without?
Compare fps pls (type 'dnrate' while in game to show)

It seems MSKVBEF7 can significantly improve performance even on my Quadro 2000, and fixes (not frequent) flicker issue when playing DUKE3D on 800x600. Player movement is also much smoother after using MSKVBEF7.

Averages without MSKVBEF7:
Starting point: 139
Opening menu: 70
Entering the room with mirror: 24

With MSKVBEF7:
Starting point: 170
Opening menu: 137
Entering the room with mirror: 187

Is anybody know dos utility to set mtrr's on intel ivy bridge cpu?

New version , now you can disable and force vsync

Fps in duke with novsync mode anyways lower then with mskvbef7 and i see some flickering
Anyway , it interesting to test and compare

an64 wrote on 2023-09-02, 08:27:

Is anybody know dos utility to set mtrr's on intel ivy bridge cpu?

My mtrrlfbe should work.

New and final (i hope) version
Universal NVIDIA vbe function 0A fix for gtx 9xx and newer

RayeR wrote on 2023-09-02, 11:52:

an64 wrote on 2023-09-02, 08:27:

Is anybody know dos utility to set mtrr's on intel ivy bridge cpu?

My mtrrlfbe should work.

Thanks , ver 1.6 works for LFB
For vga it doing some strange things , not working at all

Falcosoft wrote on 2023-09-01, 21:12:

Yes, it is using vsynch when page flipping is active but I think it is normal. Both the real-mode and the protected mode interface provide explicit vsynched HW page flipping mode (Function 07: BL = 80h Set Display Start during Vertical Retrace).
Build engine games and Quake most likely use this mode to prevent tearing.

Ok , now we have 2 alternate ways
1: MSKVBEF7 , shows 300+ fps here 😀 I not shure thats right, but impressing
2: My NVUVAFIX . Shows 60/30 (mostly 60) , but with vsync

Can you test it with NVUVAFIX.COM /5 /9 on your 960?
And especially interesting on 1060

[/quote]
For vga it doing some strange things , not working at all
[/quote]

Maybe, I never used it for VGA as the VGA FB is only 64kB and it doesn't have much sense to use it in such modes like 320x200 as they are fast enough anyway...

RayeR wrote on 2023-09-04, 16:41:

Maybe, I never used it for VGA as the VGA FB is only 64kB and it doesn't have much sense to use it in such modes like 320x200 as they are fast enough anyway...

The 64KB VGA FB is used also by all high resolution VESA 1.2 modes (i.e. all modes that do not use LFB but banked mode).
E.g. if you use NOLFB then you will always use VGA FB even for high resolution modes. Also real mode games that use VESA modes and utils written with Turbo Pascal/C that use SVGA BGI drivers use VGA FB for high res graphics.
In the above cases speed could be somewhat relevant.

Yes but window switching bios calls would probably make it much slower regardless setting WC...

RayeR wrote on 2023-09-04, 19:55:

Yes but window switching bios calls would probably make it much slower regardless setting WC...

According to my tests when WC flag is working for VGA FB it can speed up rendering by 2-15%.
Re: MTRRLFBE and AGP/PCIe cards in DOS
On Athlon64/Phenom CPUs you have to fine tune MTRR's since System Management Mode related settings interfere with banked VGA write combining by default. Maybe there is a similar problem with newer Intels also.
Re: MTRRLFBE and AGP/PCIe cards in DOS

Falcosoft wrote on 2023-09-04, 20:07:

According to my tests when WC flag is working for VGA FB it can speed up rendering by 2-15%.

That's not much worth (I guess that bottleneck is calling vbios and PM/RM switching)...
on LFB the speed up may be 50x on modern systems with PCI-E...

RayeR wrote on 2023-09-04, 20:17:

Falcosoft wrote on 2023-09-04, 20:07:

According to my tests when WC flag is working for VGA FB it can speed up rendering by 2-15%.

That's not much worth (I guess that bottleneck is calling vbios and PM/RM switching)...
on LFB the speed up may be 50x on modern systems with PCI-E...

50x speed up sounds to be a rather theoretical result. With real world programs the speed up is expected to be at least 1 magnitude less.
With the same testing methods LFB+WC resulted in 25%-250% speed up on the same tested systems (interestingly the system with dedicated PCI-E card produced the worse speed up in both VGA+WC and LFB+WC cases).

RayeR wrote on 2023-09-04, 19:55:

Yes but window switching bios calls would probably make it much slower regardless setting WC...

Now my NVUVAFIX have a /5 switch for using protected mode code taken from gtx520
I compared vbetest perfomance in 1280x1024x32 banked mode - my code ,calling rm int10 for func 05 and pm code:
9 vs 25 fps and in 1 test i can see the diffence by eye
never think, that pm-rm transitions takes so much time even on fast cpu...

Ps /5 tested only on maxwell 980ti, please someone test it on newer nvidia
Universal NVIDIA vbe function 0A fix for gtx 9xx and newer

Thanks for info, it's really noticable speed up. I will test on my GTX970, I expect it's affected the same as 980.

I recall there was a notice for removal of VBE from AMD cards, here's a third party link:

https://www.scribd.com/document/578511765/Rem … -8-bbb-graphics

BEEN_Nath_58 wrote on 2023-12-11, 02:00:

I recall there was a notice for removal of VBE from AMD cards, here's a third party link:

https://www.scribd.com/document/578511765/Rem … -8-bbb-graphics

This is near 10 year old, what's the current status on AMD cards? Nvidia even broke VGA compatability at IO regs level on RTX, does at least VGA modes still works on current AMD cards? And VBE?