Have it run something demanding - see if it has errors after doing that for a while, like loop 3D01 Nature or the PS/fill tests, or AQ3 (if the card runs it). I'd try bootable stuff too - see what it does in VGA mode for example. Tbh I've never experienced such issues though (never done 133 on a BX but I've run AGP overclocked on boards without a divider) - just thinking general stress testing ideas. Ofc you could use furmark or something, but I don't like the load that tends to put on cooling. You could also look for games that are notoriously finnicky and see if they have issues - Morrowind and ORB both come to mind offhandedly.

If you can run Aquamark 3 looped for like an hour @ that setting its probably stable.
That bench will crash at the slightest instability.

You need something weaker for lesser cards tho...
Looping the dragon merry-go-round, 1-light & 8-color lighting tests from 3dmark2001 might be good, I used to do that for burn-ins. It crashes pretty easy if the system is unstable.

I think you should at least run one benchmark which taxes the card.

Problem is that 89mhz agp is overclocking so a card can handle it maybe for weeks without crapping out while another can do it for years.

Maybe transferring lots of data over the AGP port? Wasn't there some benchmark with huge textures which shouldn't fit into the card's RAM to test AGP texturing speed?
The question is what exactly is supposed to cause the issues, just the high clock presented to the card, or the high clock during data transfers.
I can image that there might be a interdependency with the AGP mode, ie. AGP4x more sensitive than AGP2x. 440BX only does AGP2x, though. If you find a card that gives trouble you could try to restrict it to AGP1x, if that's at all possible.

Question: Isn't this going to be rather like overclocked CPUs? By that I mean some individual chips will handle a setting while others won't. And the perceived risk is reduced lifetime due to running hotter.

I'm still very interested in this test. If a bunch of the cards are unstable, that'll be something of a revelation. But mixed results or overall stable results won't tell us as much, will they?

I'm maybe being a bit extreme here, but in order to qualify stability at a given speed, I prefer to test at a slightly higher speed. So to qualify a BX133 configuration, I would be in favor of testing at 138-140FSB.
The rationale is that it somewhat compensates for the imperfection of stress testing, and also establishes that a reasonable margin exists, so you know you're not sitting on the edge. I've used this approach when overclocking (not with this application though) and it has always given safe results. I was bit once when I didn't do this, and the person's PC started having BSODs about 3 months later.

One problem with doing this though is that it may just as easily introduce problems that have nothing to do with the video card. BX133 builds are already well above the originally intended speed of a 440BX chipset, so adding a slight bit more isn't inconsequential to it.

I suppose a test that stresses memory transfers across the AGP bus would be good, but I don't know what is best for this. A few laps at peak operating temperature of 3DMark at 138FSB/92AGP would be pretty convincing IMO (but time consuming). POSTing several times could maybe be another test, but as long as the occasional failure to POST doesn't bother the user then this doesn't necessarily matter.

It would be interesting to see a large population of test results from people's different cards all in one place. I have some cards I'd be interested to test, but I don't have a suitable system assembled right now.

pewpewpew wrote:

Question: Isn't this going to be rather like overclocked CPUs? By that I mean some individual chips will handle a setting while others won't. And the perceived risk is reduced lifetime due to running hotter.

I'm not sure if they will run hotter. Video cards have their own clock oscillators, so I'm not sure how much of the functionality would really be overclocked. Signaling would be fast though, and some cards might not be as stable with that.

The thing is, just because it works doesn't mean it will do so over an extended period of time.

The other thing I'm interested in is: What is the actual performance difference?

I use a 100 MHz FSB Pentium III 1.1 GHz in my AOpen slot 1 board with CL2 memory and it's fast as heck. How much faster would a Tualatin in a BX440 machine be? Is it worth the trouble?

Mau1wurf1977 wrote:

The thing is, just because it works doesn't mean it will do so over an extended period of time.

The other thing I'm interested in is: What is the actual performance difference?

I use a 100 MHz FSB Pentium III 1.1 GHz in my AOpen slot 1 board with CL2 memory and it's fast as heck. How much faster would a Tualatin in a BX440 machine be? Is it worth the trouble?

From my benchmarks I believe it's about 2000 3Dmark 2001 points. I'll double check when I get back from work 😀

Good points have been brought up in this thread. Certainly it's a question in how the problems manifest themselves. Nobody that I know have given any details about this. Not all the AGP cards I have can run 3Dmark 2001SE, so on some ATI and S3 cards I maybe have to run 3Dmark 2000. I also think the overclocking to 140FSB is a good idea to test the limit even more. If the card can run 3Dmark 2001SE then it's good to go in my opinion.

I don't have a Tualatin Slot 1 adapter so I can't help out. I could compare a S370 i815E CL3 1.4 GHz with the 1.1 GHz BX 440 CL2 but not sure if I'm eager enough.

Mau1wurf1977 wrote:

I don't have a Tualatin Slot 1 adapter so I can't help out. I could compare a S370 i815E CL3 1.4 GHz with the 1.1 GHz BX 440 CL2 but not sure if I'm eager enough.

No worries, it's not the point of the test. It's the 133MHZ FSB that is up for test, not the performance differences. Many Pentium 3 CPU's used 133MHZ FSB, not just the Tualatin.

One question I'd have on the P3 comparison - 2k pts in 3D01 sounds pretty impressive, however what does that actually translate to in terms of FPS gains on the high detail or advanced tests? 3D01 is pretty CPU weighted, and usually the "huge scores" are the result of the LD tests being allowed to go to the moon - which ofc does show that the CPU performance is improving, but a lot of games are either more equally weighted or GPU weighted (especially higher resolution/IQ enhancement/etc) so it may not be as substantial a factor in terms of gaming performance.

I'm not sure 3DMark01 is the best test for AGP/GPU stability. At least in my experience, AGP issues sometimes don't manifest themselves in synthetic benchmarks. My PIII box is a Tualatin-1585 on a VIA 694 board. The board is nice and stable in AGP 2X mode, but most of that stability goes out the window in 4X mode.

However, I usually don't see AGP-related instability in 3DMark, Aquamark, or even in games. My ATI cards (9800Pro, x800XT, x1950 Pro) will loop 3DMark 2000-2003 for hours without crashing in AGP 4X mode. Yet Firefox will hang the system in a matter of minutes, usually with a bunch of lines/artifacts on the screen. Shader 2.0 assisted WMV-HD decoding will also instantly crash these cards when they're operating in 4X mode.

OP says "I'm planning on testing my AGP cards in a 133mhz FSB system running on the
ASUS P3B-F v1.03 Intel 440BX chipset. There are so many people saying there are issues
with the overclocked AGP bus"...

You didn't say if you had seen this webpage, so..;
http://www.reocities.com/ResearchTriangle/448 … 4%20and%20below

+1 for the 140mhz fsb, very good suggestion.

If it has the headroom to work at 140mhz it will be fine at 133mhz i guess. Also let us know which agp bios settings you use as these can be important as well.

obobskivich wrote:

One question I'd have on the P3 comparison - 2k pts in 3D01 sounds pretty impressive, however what does that actually translate to in terms of FPS gains on the high detail or advanced tests? 3D01 is pretty CPU weighted, and usually the "huge scores" are the result of the LD tests being allowed to go to the moon - which ofc does show that the CPU performance is improving, but a lot of games are either more equally weighted or GPU weighted (especially higher resolution/IQ enhancement/etc) so it may not be as substantial a factor in terms of gaming performance.

Score with 100mhz FSB (1050mhz) = 6892
Score with 133mhz FSB (1400mhz) = 8515

Not entirely 2k, but close.

Also for the first time ever I tried 140mhz FSB and 150mhz FSB on this system. I got a VGA boot errorcode on the speaker. So I guess it boils down to if its able to boot or not.

OP says "I'm planning on testing my AGP cards in a 133mhz FSB system running on the ASUS P3B-F v1.03 Intel 440BX chipset. There […]
Show full quote

OP says "I'm planning on testing my AGP cards in a 133mhz FSB system running on the
ASUS P3B-F v1.03 Intel 440BX chipset. There are so many people saying there are issues
with the overclocked AGP bus"...

You didn't say if you had seen this webpage, so..;
http://www.reocities.com/ResearchTriang ... nd%20below

Not sure how this is relevant as this is for CPU instabilities in Photoshop. I havent noticed any problems like they are describing though.

I'd love to see same frequency CPUs put side by side. I've been thinking of getting a Tualeron instead of a full Tualatin. The Celeron Tualatin features 256kb (like most Tualatin CPUs anyway) and a 100MHz Bus, which should be the most significant factor that hinders performance somewhat.