VOGONS


First post, by vvbee

User metadata
Rank Oldbie
Rank
Oldbie

On a K6-2 300,* the Voodoo 3 2000's performance is bottlenecked, as is common knowledge. At resolutions up to 800x600, the game Need for Speed 5 (below) at medium settings is capping at an average of about 35 FPS running a 120-second replay of a race on the game's first track regardless of whether the Voodoo 3's core and memory are clocked at the default 143 MHz or underclocked to 100 MHz.

nfs5.png

In that game, at 640x480, a paired t-test suggests that performance at 143 MHz does differ from that at 100 MHz (p = .022); but the figure below (left) shows that this framerate, sampled once per second over the duration of the replay (n = 120), isn't meaningfully distinguished between the two operating frequencies, differing on average by 1 FPS (triangles = FPS samples at 100 MHz, circles = FPS samples at 143 MHz). At 800x600 (middle), there's a slightly more tangible benefit for 143 MHz (highest 10% FPS average = 46) over 100 MHz (highest 10% FPS average = 44), but given that the latter is a 30% underclock, a 5% drop in performance isn't notable. At 1024x768 (right), there's a more clear-cut benefit, mostly across the FPS spectrum, for running the card at 143 MHz rather than 100 MHz. The framerates given here are truncated averages of two runs each.

640x480_100vs143_.png 800x600_100vs143_.png 1024x768_100vs143_.png

A similar trend is found running Quake 3 (DEMO001; default settings; left graph below) and 3DMark 99 MAX (race test; right graph below). Up to and including 800x600, there's on average no framerate benefit to running the Voodoo 3 at 143 MHz: even at 92 MHz, the lowest underclock my card would do without memory issues,** no real differentiation in speed is seen. The framerates given here are truncated averages of two runs each.

q3arena.png 3dmark99.png

The benefit of underclocking is a reduction in core - and possibly RAM - temperature. Underclocking the Voodoo 3 from 143 to 100 MHz, I found a 3 °C reduction in core idle temperature, from 43 to 40 °C (against 24 °C in-case ambient),*** and the same result was had with a repeat some hours later. At the end of a 10-minute loop of DEMO001 and DEMO002 in Quake 3, the temperature was 46 °C at 143 MHz and 42 °C at 100 MHz. The temperatures were measured running at 143 MHz first, then at 100 MHz, which by design may have given a slight edge to the former. It's likely that the GPU core was running at notably higher temperatures than those measured at the surface of its heatsink, as done here, which means that the relative differences in temperature would likely be found larger if measured directly from the core. Quake 3 had a resolution of 800x600, here, i.e. running bottlenecked.

temps.png

In short, when using resolutions of 800x600 or less, and certainly 640x480, on this system with the programs tested here, there's in general no performance penalty for underclocking the Voodoo 3 from 143 MHz to 100 or even 92 MHz, and doing so will reduce the card's operating temperature and possibly increase its lifetime.

* Test system specs: AMD K6-2 300 CPU (100 MHz FSB), PC-Chips M577 motherboard (Csongrádi/Steunebrink 990309J2 BIOS), 64 MB PC-100 RAM, Voodoo 3 2000 AGP GPU, IBM Deskstar 25GP harddrive, Windows 98 SE.
** The card's RAM chips are labeled SEC KM416S1020CT-G7. I can't decipher from their datasheet what the minimum frequency is, but below 92 MHz, I get garbage on screen.
*** As measured with a digital contact thermometer from roughly the middle of the top of the core's heatsink, after the PC had spent 10 minutes idling on the Windows desktop. Throughout, the heatsink received ambient-temperature airflow from a 45 CFM case fan ~25 cm away.

Reply 1 of 13, by clueless1

User metadata
Rank l33t
Rank
l33t

Nicely presented. I'm all about efficiency and agree with you completely!

The more I learn, the more I realize how much I don't know.
OPL3 FM vs. Roland MT-32 vs. General MIDI DOS Game Comparison
Let's benchmark our systems with cache disabled
DOS PCI Graphics Card Benchmarks

Reply 3 of 13, by vvbee

User metadata
Rank Oldbie
Rank
Oldbie

Can't desync core and ram frequency on the voodoo 3 as far as I know.

3dfx's own overclocking utility, at least the one with v. 1.07 drivers, won't let you go below 100 mhz, not sure whether that's arbitrary or a general limitation of the hardware.

Reply 5 of 13, by vvbee

User metadata
Rank Oldbie
Rank
Oldbie

As to what limits the underclock, I'm only guessing, and with a sample size of one, it's especially hard to guess. This particular sample says that below 92 mhz (reported as 93 mhz by the 3dfx info screen), screen corruption occurs, and below 85 mhz or so, a blank screen.

Reply 7 of 13, by F2bnp

User metadata
Rank l33t
Rank
l33t
vvbee wrote:

As to what limits the underclock, I'm only guessing, and with a sample size of one, it's especially hard to guess. This particular sample says that below 92 mhz (reported as 93 mhz by the 3dfx info screen), screen corruption occurs, and below 85 mhz or so, a blank screen.

I'm thinking it could be memory timing? Let's say, that after a certain low point, the timings are lowered (so as to increase speed), and because the memory can't really cope with the reduced timings, artifacts start appearing.

Reply 8 of 13, by gerwin

User metadata
Rank l33t
Rank
l33t

Here is another sample:

VOODOO 3 PLLCtrl  -- 19-04-2011 GB, INFO FROM BIOS FILES:
TYPE 1000/U.C. 1000 2000 3000 3500 3000/O.C.
MHz 100,227 125,090 143,182 166,091 183,750 200,455
K 1 1 1 1 1 1
M 11 11 1 3 1 3
N 180 225 58 114 75 138

trying 1, 3, 54 = 80 MHz ..some garbled output
trying 1,11,180 = 100 MHz ..some garbled output
trying 1, 1, 40 = 100 MHz ..some garbled output (banshee settings)
trying 1,11,198 = 110 MHz ..some garbled output
trying 1, 3, 78 = 114 MHz ..OK
trying 1, 3, 82 = 120 MHz ..OK
trying 1,11,225 = 125 MHz ..OK

--> ISA Soundcard Overview // Doom MBF 2.04 // SetMul

Reply 10 of 13, by dr_st

User metadata
Rank l33t
Rank
l33t

No.

It's a good piece of research, and very nicely presented indeed, but has no practical value. Like a lot of the things we (including yours truly) do and talk about here on VOGONS. 😀

https://cloakedthargoid.wordpress.com/ - Random content on hardware, software, games and toys

Reply 11 of 13, by vvbee

User metadata
Rank Oldbie
Rank
Oldbie

The practical value of underclocking the voodoo 3 here is a ca. 10% reduction in temperature at no cost in performance, given certain hardware and resolutions. You can relate this to the 10% or so increase in performance many people are content with from overclocking their hardware, something that in their view is, in the short term, a benefit at no cost.

The reading of 46 c isn't necessarily low when you translate it from what it is, the surface temperature of the heatsink, to what it represents, the temperature of the core as measured off the surface of its heatsink. Given 46 c on the heatsink, the core may be at 60+ c, and by underclocking you may take it to the 50s. I might tape on some ds18b20 sensors and see about the effect of the fan on the temperature, and I'm curious about the ram chips as well.

Too bad if there's large variation in how low the voodoo 3 can be underclocked, though.

Reply 12 of 13, by gerwin

User metadata
Rank l33t
Rank
l33t
vvbee wrote:

Too bad if there's large variation in how low the voodoo 3 can be underclocked, though.

I just checked my cards, and the frequency they have been running at since 2011 (by BIOS setting):
AGP Voodoo 3 3000: 114 MHz
AGP Voodoo 3 2000: 100 MHz
PCI Voodoo 3: 95 MHz

Increased frequency setting of these cards is of no importance to me, as it is of no benefit in the old games that I run with them. I rather have a cooler card. The AGP Voodoo 3 3000 has a stock heatsink that is OK, but on the Voodoo 3 2000 AGP I put a bigger heatsink. The PCI version has a VRM on the card that gets so hot; it will hurt you when you accidentally touch it. There is now an additional heatsink attached to that VRM.

Going with even lower frequencies gave me increased risk of artifacts or a wobbly image. It depends, because a low motherboard FSB speed + low resolution video mode both increase the risk too.
Here is an older topic that covers some of this: Voodoo3, shaky picture and snow

--> ISA Soundcard Overview // Doom MBF 2.04 // SetMul

Reply 13 of 13, by vvbee

User metadata
Rank Oldbie
Rank
Oldbie

Booting the pc after a few days gave a blank screen on windows' desktop, rebooting gave a cursor and some corrupt icons. This was running the v3 at 92 mhz. I suspect an inverse relationship between temperature and minimum mhz here. Didn't perceive errors once it was booted at 100 mhz.

Measured some temperatures. The setup was like this, the v3 is the top expansion card on the left, with an sb live below it. The white fan, 140x140, on the right was active, the black fan behind it was not.
v3_illust_web.jpg

Had temperature sensors, not shown in the image, on the pcb at the back of the card right where the core is, on the vrm, and on one of the ram chips. No decent way to attach a sensor to the heatsink, so didn't. The others were mounted with electrical tape, not ideal but did the job.

PC first idled for 5 mins and ran quake 3 for 5 mins at 143 mhz, then idled for 10 mins and ran quake 3 for 5 mins at 100 mhz. Below, the temperatures of that wrt. time. The first peak is running q3 at 143 mhz, the second at 100 mhz.
ds_temps.png

The vrm didn't get all that hot either way. Ram, cool. Pcb, ok. The card's surroundings were well ventilated, so add 5+ c otherwise. That the temp was about the same off the pcb as off the heatsink earlier may suggest poor thermal transfer to the heatsink.

At 143 mhz, pcb min = 44 c, pcb max = 46 c. At 100 mhz, pcb min = 40 c, pcb max = 42 c. The sensors reported 23 c ambient before first bootup.