Nice review.
Just wanted to correct one thiing, though...

You said that you got the memory on the TI 4600 running at 720 mhz. This isn't exactly 720 mhz at all. The Nvidia spec for the GF2U+ cards call for a 13.6 mhz granuality (or something like that, I don't know the techincal explanation), but the person from Entechtaiwan explained it for me. Basically, if you set the card for 720 mhz, you are really running it at 715 mhz, which is what the RAM is "spec'd" to do, at 2.8 ns. The next higher step is 729 mhz, which I am sure, you saw was giving artifacts.

Usually, the next higher step will be set if you are more than halfway there. 720 mhz should round down to 715, and 722 should round up to 729.

That's why some people say they got their card to 710-720 (715), while others say they got theirs to 730 (729).

Rivatuner has an option to set "fine" clock adjustments on a low level, but I don't know how reliable this is, or even if it's supported on a GF4.

That sounds pretty interesting, and actually I had no idea of such thing (in case your info is correct). I guess you learn new stuff everyday

I know that both Rivatuner and Powerstrip act in this fashion, which is a documented feature of the clock generators. Rivatuner does have extra settings to allow fine-tuning, but they are riskier to use (as they might not work).

You can check the Rivatuner docs for more information, or even contact entech taiwan about it.

I don't know exactly how coolbits works, though, in relation to this.
You can always use coolbits, then check the actual frequency using the clock controls (just checking, not adjusting), using Rivatuner......

That might tell you if coolbits actually set that exact clock or not (or if internally, the values are rounded to 13 mhz increments).

Will probably check that out later with the MX videocard I'm testing right now. BTW, how come you don't stop by the forums more often, it would be nice you have your around.

Thanks for the kind words.
I usually go around forum hopping...and if I see an interesting post, then might reply, or help someone if i'm familiar with a bug. There are a lot of forums, though, and I'm somewhat lazy about having to keep up with all of them. I'll be sure to keep this one bookmarked...
(meaning, I'll visit it more often, of course)

Sounds good... welcome to TS Forums then

Riva Tuner on my GF2 MX400:

stock clock: 200.5 core .......... 167 memory (x2)

1st 'step' : 204.0 core .......... 168.2 memory (x2)

2nd 'step' : 205.2 core .......... 169.4 memory (x2)

3rd --------- 207.6 --------------- 171.8

4th --------- 210.0 --------------- 175.4

5th --------- 211.2 --------------- 179.0

6th --------- 214.8 --------------- 182.6

7th ---------- ???? ---------------- 186.1

8th ---------- ???? ---------------- 189.7

9th ---------- ???? ---------------- 193.3

10th -------- ???? ----------------- 196.9

11th -------- ???? ----------------- 200.5

Except for the 1st core jump of 3.5, all the rest are divisible by 1.2. On the memory side, all are divisible by 1.2 except for the jump from 182.6 to 186.1 (3.5). Hmmm.

I stopped at 214.8 on the core because it acted like it wanted to hang & I'm doing this as I type. I also didn't 'apply' the 200.5 on the memory because I have experienced it hang at that speed.

Riva, GFTU & Cool Bits all say the memory speed is being dbl'd (ie; stock is 334). Fresh Diagnose benchmarks 3rd & 4th on the core as 'best' coupled w/ 8th or 9th on the memory. I run low core (207.6) & high (193.3 x2 or 386.6) memory. Works well for me.

I'll update this w/my memory brand & ns, if anyone is interested in compiling this info.

YMMV (Your Mileage May Vary),


***Update 5/16/02: I brought the memory down to 372 (186.1 x2) because the edges of the monitor got 'jaggy'. I figured it was the memory after I pulled the card & saw it is running @ 6ns. That is kinda slow & won't (evidently) take to a lot of oc'ing. It is VisionTek brand memory too. ***

HTH,