VOGONS


Reply 20 of 26, by The Serpent Rider

User metadata
Rank l33t++
Rank
l33t++

but FX 5800 Ultra runs at 500 MHz

Also it's practically does not exist and overclocked to the max from the factory (no headroom left).

I must be some kind of standard: the anonymous gangbanger of the 21st century.

Reply 21 of 26, by havli

User metadata
Rank Oldbie
Rank
Oldbie

It was direct competitor to R9700 Pro. And it does exist, I have one right here. 🤣
OC headroom is not the point... only stock clock is because it is the official specification. Btw - R9700 Pro is just as poor overclocker as FX 5800 Ultra. Mine Ultra can do +7% on the GPU and +16% on the MEM. And 9700P is +14% GPU and +6% MEM. So more or less the same.

HW museum.cz - my collection of PC hardware

Reply 22 of 26, by swaaye

User metadata
Rank l33t++
Rank
l33t++

Yeah I don't remember them being rare at the time but they didn't build many. 5900 was almost ready when 5800 finally launched.

500 MHz was probably a stretch but they pulled it off. And I wouldn't be surprised if the awful cooler actually made them sales because of its looks.

Reply 23 of 26, by Putas

User metadata
Rank Oldbie
Rank
Oldbie
agent_x007 wrote:
swaaye wrote:

By the way, Putas has put together a comprehensive table. Tries to abstract everything to an extreme so you can kinda compare all the cards.
http://vintage3d.org/dbn.php#sthash.Il6om7Lq.dpbs

How he get to 48 Pixel Pipes number compared to 96 ROPs on newer NV cards (Maxwell/Pascal/Turing) ?
Does he have a specsheet with number of pixels/rasteriser (or per SM) ?

8 pixels per rasterizer.

Reply 24 of 26, by agent_x007

User metadata
Rank Oldbie
Rank
Oldbie
Putas wrote:

8 pixels per rasterizer.

Hello 😀
Sadly, it's wrong value. I base this on test from here : LINK
Here's explenation of the above on TechReport : LINK
Also, my GM200 card gets over 2x performance bump on 3DMark Vantage Pixel Fillrate test vs. GK110b, at similar clock speed (ie. within 10% GPU clock).
Unless those "Rasterisers" in Big Maxwell are simply mindblowingly more efficient compared to those on Big Kepler's, it's impossible to have this kind of jump without doubling "pixel pipes".

PS. Why no Titan V/Titan RTX ?

157143230295.png

Reply 25 of 26, by Putas

User metadata
Rank Oldbie
Rank
Oldbie

Most probably you are correct and rasterizer : ROP output ratio remains near 1. However, from those articles I cannot tell whether Nvidia confirmed Damien's conclusion. Since the fillrate is still significantly behind theoretical maximum of 8 pixels/rasterizer, there is a chance other changes caused the uplift.

Scratch that, the numbers from b3d suit broke the barrier, I will also go with 16 pixels.

agent_x007 wrote:

PS. Why no Titan V/Titan RTX ?

I limited the database to discrete 3d accelerators usable for games. Even if missing Titans can do that, I think their price excludes them from consideration as cards for games.

Reply 26 of 26, by Ozzuneoj

User metadata
Rank l33t
Rank
l33t

Got my hands on a QuadroFX 3000 (FX 5900 Ultra equivalent).

Here is what everest says.

QuadroFX3000_everest.jpg
Filename
QuadroFX3000_everest.jpg
File size
485.9 KiB
Views
421 views
File license
Fair use/fair dealing exception

Yup. 4 pipelines, 8 TMUs, as expected. It also shows 3 vertex and 1 pixel shader, for what that's worth. The specifics of those seem a bit hard to figure out based on other info online.

I will say, I was surprised to see that this card downclocks when not under load. When the fan kicks up to a higher RPM the card is at 399mhz core. When the fan drops to a lower speed the card is running at 300Mhz core. I had no idea the FX series did this. Is this a Quadro feature?

Also, I'm sure most of these cards have a 400Mhz RAMDAC, but I just realized that this is the same as my GTX 970, which is one of the last VGA cards released. Pretty crazy.

Now for some blitting from the back buffer.