Hey @supercordo!
I just noticed small error in my 3DM2001SE entry. CPU clock was 2708MHz, not 2806MHz (that was spi1M clock).

forteller wrote on 2026-08-28, 17:56:

Hey @supercordo!
I just noticed small error in my 3DM2001SE entry. CPU clock was 2708MHz, not 2806MHz (that was spi1M clock).

Fixed!!!

I won't risk it... maybe someone else will? 😉

MattRocks wrote on 2026-09-02, 05:53:

I won't risk it... maybe someone else will? 😉

It would be a downgrade.

supercordo wrote on 2026-09-02, 07:06:

MattRocks wrote on 2026-09-02, 05:53:

I won't risk it... maybe someone else will? 😉

It would be a downgrade.

Are you sure? Once the 3DMark assets are loaded, it's over to the GPU.

MattRocks wrote on 2026-09-02, 07:59:

supercordo wrote on 2026-09-02, 07:06:

MattRocks wrote on 2026-09-02, 05:53:

I won't risk it... maybe someone else will? 😉

It would be a downgrade.

Are you sure? Once the 3DMark assets are loaded, it's over to the GPU.

I have one and don't recall performance to be that great.. could give it a shot at some point.. Time is the issue at hand 😁

well, that wasn't easy

The Sandman wrote on 2026-09-02, 16:50:

well, that wasn't easy […]
Show full quote

well, that wasn't easy

Very Nice!!!!! Ill let that stand for a few. Work is busy at the moment. Its also still summer time so im sure we will get more scores once fall comes.

MattRocks wrote on 2026-09-02, 05:53:

I won't risk it... maybe someone else will? 😉

Now you give me ideas.

Pretty sure I already have some of these, maybe with the x16 slot instead of the x1 slot.

However, they will only be running at PCI 33-mhz (or whatever the PCI bus is running at).

The bridge chip does support 66-Mhz operation, but only the server boards that aren't really going to benefit from that due to lower CPU and RAM speed will do that.

Waste of money. The later agp cards are already using a PCIE bridge chip. Putting it on the slowwer pci bus is retarded.

supercordo wrote on 2026-09-02, 22:24:

Waste of money. The later agp cards are already using a PCIE bridge chip. Putting it on the slowwer pci bus is retarded.

But you can only go the 3850 and 4670 as the fastest AGP cards.

The bus bottleneck is of course going to affect performance but I wonder if you could get a higher score with a 3870 or 4890 or something on the Nvidia side such as an 8800 series or faster.

How about pairing an Athlon XP with a Titan Black or ATI 7970?

cyclone3d wrote on 2026-09-02, 22:49:

But you can only go the 3850 and 4670 as the fastest AGP cards. […]
Show full quote

supercordo wrote on 2026-09-02, 22:24:

Waste of money. The later agp cards are already using a PCIE bridge chip. Putting it on the slowwer pci bus is retarded.

But you can only go the 3850 and 4670 as the fastest AGP cards.

The bus bottleneck is of course going to affect performance but I wonder if you could get a higher score with a 3870 or 4890 or something on the Nvidia side such as an 8800 series or faster.

How about pairing an Athlon XP with a Titan Black or ATI 7970?

With only a 133MB/s bus? The drivers will propable crash first thing.

The Sandman wrote on 2026-09-02, 16:50:

well, that wasn't easy […]
Show full quote

well, that wasn't easy

Hell yeah! I'm glad that an actual AGP card won.

supercordo wrote on 2026-09-02, 22:24:

Waste of money. The later agp cards are already using a PCIE bridge chip. Putting it on the slowwer pci bus is retarded.

The value of bandwidth is less time loading textures before the benchmark run begins, and nobody seems to care to measure that particular delay. When comparing AGPx2, AGPx4 and AGPx8, I think we see an FPS improvements of ~2% when increasing bandwidth from 266MB/s to 2133MB/s.

The more valid criticism is that adding any bridge (e.g. PCIe to AGP, or PCIe to PCI) will also add a translation layer with command queue batching. Some A/B testing is needed to know if its really painful or not. It is easier to design a test than to find the time to execute it 😉

Could be a waste of money. The PCIe-PCI bridge probably does not support VBIOS at POST, but any attached PCIe card can be detected by the OS, so.. I don't know!

It's not over yet for the native AGP cards.

It is remarkable.
I cannot follow as my motherboard tops out at 220MHz (in theory, let alone in practice) but do you have any lessons learned on temperatures and temperature control?

Great score.
Why 2-2-2-8, not 2-2-2-11?

The Sandman wrote on 2026-09-04, 09:58:

It's not over yet for the native AGP cards.

Nice, i wont be home for a few weeks to play too.

For the temperature and controlling it I can only provide you with this information: stop benching before shit hits the fan. If you see that the waterblock is getting moist on the outside, shut the system down and remove it from the board. For the tubing I was wrapping around one layer of toilet paper to keep it dry. That's it.

I honestly didn't try any other timings. It's stable in 3DMark01 😉 . Back in the day people said that 2-2-2-11 is the fastest timing for nF2, I know. I don't think that there is much of a difference in 3DMark. Benching SuperPi is another story.

stay safe supercordo

SlowA wrote on 2026-09-04, 12:34:

Great score.
Why 2-2-2-8, not 2-2-2-11?

2-2-2-11 was the general advice back then, my assumption is that this was chosen because the default for tRC and tRFC (both depend on the primary timings) is 13/15. 2-2-2-5 is actually faster, especially when you lower RC and RFC to 7/11 or 9/12. so basically 2-2-2-5-7-11 or 2-2-2-5-9-12. note that not all boards or even ram sticks will handle this.

I’ve binned my BH-5 sticks specifically for low tRC and tRFC on NF2, as some sticks will run with 13/15 at high clocks but fail completely (memtest errors at or below 240Mhz) once you tighten to 9/12. my board and chipset can’t handle 7/11 at all…

In contrast my crucial ballistix (2x512mb 5B G) won’t run 250Mhz at tRAS 5, but 2.5-2-2-9 on NF2. Tweaking drive strength and slew rate to 2/6 is required though.