BFG nVIDIA GeForce 7800GS OC on Win98

[RainyShadow]

AGP status - disabled

Looks like you have a problem with the chipset drivers...

[Goodmaneuver]

Yes sounds like VIA GART driver is not installed. It is in the 4 in 1 drivers. I think I have found where to download the Intel accelerator driver here https://download.cnet.com/Intel-Application-Accelerator/3000-2086_4-10163577.html It was still supplied in a G31 motherboard install disc in 2008. Executable name starts with iaa23. I found VIA drivers on VIA web site. 4 in 1 should be here directly https://d34vhvz8ul1ifj.cloudfront.net/Driver/VIA_HyperionPro_V524A.zip

Edited October 9, 2020 by Goodmaneuver

[Feamane]

Hi guys,

Yes, of course I saw that.  But the strange thing is if I install ForceWare 81.98 or 82.69 instead of 77.72 then the AGP does not show as disabled.  And as it shows in the screenshot I posted the AGP driver is working according to Windows device manager.  It doesn't seem to matter if I install VIA drivers 4.56v, 5.04A, 5.09A, 5.11A, 5.13A, 5.20A or 5.24A--I have tried ALL of those so far because I have been doing my homework searching this forum for anybody who reported using this same motherboard and took note of the VIA drivers they reported using.  So far none of them has been the magic solution to getting 3DMark03 scores like I get with the same exact hardware on XP.

But notice this: the fastest score is with XP, the second fastest is with the 77.72 driver even though it says the AGP is disabled!  And I get a lower score with FW 81.98 or 82.69 even though HWiNFO and Coolbits show the AGP is enabled with those two drivers.

Does it make any sense to you guys that I would get a better 3DMark score with AGP disabled and a worse score with it enabled?  So at this point I do not trust what the utilities are telling me about the AGP.

I think what I am going to do this week is try different combinations of FW and VIA drivers and run the same diagnostics and do some screenshots.  This will be a repeat of what I have already tried but I was just recording the 3DMark03 scores and not the other info.  I will do this again using the different tools and record the results.  This way you and I both will have a clearer picture of what I'm seeing.

Thanks, but I have all the VIA drivers from 4in1_4.28 to Pro_5.24A, and all the ForceWare drivers from 56.64 to 307.83 squirreled away on my server so I can try many different combinations easily.  And I HAVE been trying many different combinations!

BTW, these are my notes about the different VIA driver versions I have:
Version - 4.28v     26 December 2000    AGP 08-10-2000, V4.04        INF V1.20
Version - 4.31v        04 May 2001        AGP 08-10-2000, V4.04        INF V1.20
Version - 4.35v        25 October 2001        AGP 08-10-2000, V4.04        INF V1.40
Version - 4.36a        15 March 2002        AGP 08-10-2000, V4.04        INF V1.40
Version - 4.43v      10 September 2002    AGP 07-24-2002, V4.30        INF V1.70
Version - 4.49v        22 October 2003        AGP 07-02-2003, V4.42        INF V1.90
Version - 4.53        10 August 2004        AGP 07-02-2003, V4.43        INF V2.20
Version - 4.53v      16 August 2004        AGP 07-02-2003, V4.43        INF V2.20
Version - 4.55vp1      25 November 2004    AGP 07-02-2003, V4.43        INF V2.20
Version - 4.56v      12 April 2005        AGP 08-19-2004, V4.43        INF V2.30 (1st to support PT880pro)
Version - 5.04A      11 October 2005        AGP 08-19-2004, V4.43h        INF V2.50A
Version - 5.07A      04 January 2006        AGP 08-19-2004, V4.43 (4.50D?)    INF V2.70A
Version - 5.08A      07 April 2006        AGP 08-19-2004, V4.43 (4.50D?)    INF V2.70B
Version - 5.09A      30 June 2006        AGP 08-19-2004, V4.43 (4.50E?)    INF V2.70B
Version - 5.10A      28 February 2007    AGP 06-29-2006, V4.60A        INF V2.80A
Version - 5.11A      01 February 2007    AGP 06-29-2006, V4.60A        INF V2.90A
Version - 5.12A      30 May 2007        AGP 06-29-2006, V4.60A        INF V2.90A
Version - 5.13A      27 August 2007        AGP 06-29-2006, V4.60A        INF V2.90A
Version - 5.14A      27 November 2007    AGP 06-29-2006, V4.60A        INF V3.00A
Version - 5.15A      07 January 2008        AGP 06-29-2006, V4.60A        INF V3.00A
Version - 5.16A      28 March 2008        AGP 06-29-2006, V4.60A        INF V3.00A
Version - 5.17A      28 April 2008        AGP 06-29-2006, V4.60A        INF V3.00A
Version - 5.18A      16 June 2008        AGP 06-29-2006, V4.60A        INF V3.00A
Version - 5.20A      12 August 2008        AGP 06-29-2006, V4.60A        INF V3.00A
Version - 5.21A      20 October 2008        AGP 06-29-2006, V4.60A        INF V3.00A
Version - 5.22A      30 December 2008    AGP 06-29-2006, V4.60A        INF V3.00A
Version - 5.23A      03 April 2009        AGP 06-29-2006, V4.60A        INF V3.10A
Version - 5.24A      09 June 2009        AGP 06-29-2006, V4.60A        INF V3.10A

Notice that the AGP driver doesn't change between 5.10A and 5.24A!

I downloaded the Intel accelerator driver so I can give that a try.

Thanks,
DJ

[jumper]

What are the single-core/hyperthread XP scores with 77, 81, and 82?

 

[RainyShadow]

On 10/11/2020 at 3:05 AM, Feamane said:

I downloaded the Intel accelerator driver so I can give that a try.

I don't see what does this have to do with graphics performance. It seems to be a replacement SATA driver.

Do you get huge differences in the disk transfer part of the benchmark?

Open side by side the benchmark details of a high-scoring test and a low-score one, and compare the individual scores for each part.

 

[Feamane]

On 10/11/2020 at 8:51 AM, jumper said:

What are the single-core/hyperthread XP scores with 77, 81, and 82?

 

I got some testing done but still a ways to go, maybe by the end of the week.

AFAIK 3DMark03 is single-core only, so the test results should be the same.  I don't think the score should be more than double with the same hardware on XP, should it?

Thanks,

DJ

[Feamane]

On 10/12/2020 at 6:48 AM, RainyShadow said:

I don't see what does this have to do with graphics performance. It seems to be a replacement SATA driver.

Do you get huge differences in the disk transfer part of the benchmark?

Open side by side the benchmark details of a high-scoring test and a low-score one, and compare the individual scores for each part.

 

I don't see disk transfer scores in the 3DMark03 details, just scores for different parts of the test--but none of them say anything about disk transfer bench-marking.  I am I missing something there?

According to DiskSpeed32 I am getting about 72MB/s "Avg speed" with the HDD and about 99MB/s "Avg speed" with the SSD.  In another thread here people seemed to think that was decent real life performance for SATA1/2 on this older hardware.

Thanks,
DJ

 

 

[RainyShadow]

3 hours ago, Feamane said:

I don't see disk transfer scores in the 3DMark03 details, just scores for different parts of the test--but none of them say anything about disk transfer bench-marking.  I am I missing something there?

I was wondering the same... how would changing your SATA driver improve your graphics performance? lol

That "Application Accellerator" name is kinda misleading...

[Feamane]

Are those average speed numbers I'm getting from DiskSpeed32 not good for this rig?  If they are good then I don't think that's the problem and I need to focus on other things like the drivers or other parts of Win98.  If those numbers are well below average for this setup then that area bears looking into further.

Thanks,

DJ

[Feamane]

I tried installing it, but it wouldn't install because I don't have the correct hardware.  Maybe it is only for Intel chipsets and I have VIA?

Thanks,

DJ

[Feamane]

OK, I have not finished my testing, but I think a trend is emerging.

All following tests with 4CoreDual-SATA2 R2.0, DC E6800(3330MHz), 1GB DDR2-667, DX9.0C, & RLoew SATA Patch:

					3DMark	Coolbits	HWiNFO		GPU	Everest	chipset		CPU-Z
Video	OS	VIA(GART)	FW	2003	AGP Tab	OS	GPU		MHz	AGP status		Graphic Int.
====================================================================================================================
GF6800	Win98	4.56v(V4.43)	77.72	6846	3.0 2x FW=0	8x@8x		324	Enabled	8x		AGP 3.0, 8x
GF6800	Win98	5.09A(V4.50E)	77.72	7438	N/A		8x@Disabled	324	Disabled		greyed out
GF6800	Win98	4.56v(V4.43)	81.98	6955	3.0 2x FW=0	8x@8x		324	Enabled	8x		AGP 3.0, 8x
GF6800	Win98	5.09A(V4.50E)	81.98	7594	N/A		8x@Disabled	324	Disabled		greyed out
GF6800	Win98	5.13A(V4.60A)	81.98	7592	N/A		8x@Disabled	324	Disabled		greyed out
GF6800	Win98	4.56v(V4.60A)	81.98	7594	N/A		8x@Disabled	324	Disabled		greyed out

(sorry, no matter what I do the tabs get messed up)

Basically, any version of the VIA drivers besides 4.56 do not install, I have to install them manually--and when I do so all the utilities show AGP disabled.  BUT, it doesn't seem to matter because the 3DMark03 scores end up being pretty close to each other with the 5.xx VIA drivers scoring higher even through they can't be installed by setup.exe and show no AGP enabled.  Asrock provides 5.11A as being the driver for the 4CoreDual-SATA2 R2.0, but doesn't provide any official driver for Win98.  On this forum people have reported successfully using 4.56v, 5.11A, 5.13A, 5.20A and 5.23A.  I have tried all these in the past and none but 4.56v will install by setup.exe, the rest have to be installed manually through Device Manager.

And it should be pointed out that all these 3DMark03 scores are WAY below what I get with the same exact hardware under XP.  So maybe none of these versions of the VIA drivers are installing correctly under Win98?

Thanks,

DJ

[Goodmaneuver]

On Wednesday, 14 October 2020 at 12:37 PM, Feamane said:

Are those average speed numbers I'm getting from DiskSpeed32 not good for this rig?

Some testers test a small area of the drive and all they can end up doing is testing the drive's cache speed. Sustained speed is the test. DriveSpeed32 seems to read consecutive cylinders so it seems to be a continuous sustained speed test. Those figures are good AFAIK.

 

On Wednesday, 14 October 2020 at 11:24 AM, RainyShadow said:

I was wondering the same... how would changing your SATA driver improve your graphics performance? lol

That "Application Accellerator" name is kinda misleading...

There is no INF or VXDs that I can see installed by iaa23. It is applicable for Intel chips 440BX to G31. Intel accelerator 2.2 was known to stop a nVidia card from working. If you take HDD off VIA and install on Intel then iaa23 will install and then swap back to VIA. This is PDF for 2.2 http://tetraedre.ch/public_download/opencv/opencv_tutorial.pdf. 2.3 made my system speed up but might have just been the hard drive accessing. This is for vs 6 https://jp.xlsoft.com/documents/intel/ipp/lin61/ippnotes.pdf .

 

16 hours ago, Feamane said:

I have tried all these in the past and none but 4.56v will install by setup.exe, the rest have to be installed manually through Device Manager.

Is this a new phenomenon? If so then installing the old drivers through the device manager will not over-write the new files if they have the same name but if there is a prompt and you are writing over the later file versions then OK. The reason for not being able to install older versions is because of the registry settings.
 

Edited October 15, 2020 by Goodmaneuver
spelling

[jumper]

On 10/13/2020 at 2:07 PM, Feamane said:

I got some testing done but still a ways to go, maybe by the end of the week.

AFAIK 3DMark03 is single-core only, so the test results should be the same.  I don't think the score should be more than double with the same hardware on XP, should it?

Thanks,

DJ

The drivers likely are multi-core on XP. Features not fully accelerated by the hardware are implemented in the drivers. Check the benchmark details for subtests with extra large result deviations.

Until you test with the same drivers on XP as on 9x, you're comparing apples to an orange. The XP results are not yet relevant.

 

[RainyShadow]

In addition to what @jumper said... check in the BIOS settings if you can disable the extra cores and HT - basically make it a single-core system.

Then test again in both XP and 98.

[Feamane]

6 hours ago, Goodmaneuver said:

Some testers test a small area of the drive and all they can end up doing is testing the drive's cache speed. Sustained speed is the test. DriveSpeed32 seems to read consecutive cylinders so it seems to be a continuous sustained speed test. Those figures are good AFAIK.

 

6 hours ago, Goodmaneuver said:

Is this a new phenomenon? If so then installing the old drivers through the device manager will not over-write the new files if they have the same name but if there is a prompt and you are writing over the later file versions then OK. The reason for not being able to install older versions is because of the registry settings.

No, I don't think it is something new.  Please keep in mind that for each test I'm formatting the HDD and reinstalling fresh, so absolutely no worries at all about possible driver version contamination or registry clutter.

Thanks,

DJ