The Last Bout of ‘03 – NVIDIA’s GeForce FX 5700 Ultra
by Derek Wilson on October 23, 2003 9:30 AM EST- Posted in
- GPUs
Architecture
There was a great deal of talk about why architectural decisions were made, but we will concern ourselves more with what exists rather than why this path was chosen. Every architecture will have its advantages and disadvantages, but understanding what lies beneath is a necessary part of the equation for developers to create efficient code for any architecture.
The first thing of note is NVIDIA's confirmation that 3dcenter.de did a very good job of wading through the patents that cover the NV3x architecture. We will be going into the block diagram of the shader/texture core in this description, but we won't be able to take quite as technical a look at the architecture as 3dcenter. Right now, we are more interested in bringing you the scoop on how the NV36 gets its speed.
For our architecture coverage, we will jump right into the block diagram of the Shader/Texture core on NV35:
As we can see from this diagram, the architecture is very complex. The shader/texture core works by operating on "quads" at a time (in a SIMD manner). These quads enter the pipeline via the gatekeeper which handles managing which ones need to go through the pipe next. This includes quads that have come back for a second pass through the shader.
What happens in the center of this pipeline is dependent upon the shader code running or the texturing operations being done on the current set of quads. There are a certain few restrictions on what can be going on in here that go beyond simply the precision of the data. For instance, NV35 has a max of 32 registers (less if higher precision is used), the core texture unit is able to put (at most) two textures on a quad every clock cycle, the shader and combiners cannot all read the same register at the same time, along with limits on the number of triangles and quads that can be in flight at a time. These things have made it necessary for developers to pay more attention to what they are doing with their code than just writing code that produces the desired mathematic result. Of course, NVIDIA is going to try to make this less of a task through their compiler technology (which we will get to in a second).
Let us examine why the 5700 Ultra is able to pull out the performance increases we will be exploring shortly. Looking in the combiner stage of the block diagram, we can see that we are able to either have two combiners per clock or complete two math operations per clock. This was the same as NV31, with a very important exception: pre-NV35 architectures implement the combiner in fx12 (12 bit integer), NV35, NV36, and NV38 all have combiners that operate in full fp32 precision mode. This allows two more floating point operations to be done per clock cycle and is a very large factor in the increase in performance we have seen when we step up from NV30 to NV35 and from NV31 to NV36. In the end, the 5700 Ultra is a reflection of the performance delta between NV30 and NV38 for the midrange cards.
If you want to take a deeper look at this technology, the previously mentioned 3dcenter article is a good place to start. From here, we will touch on NVIDIA's Unified Compiler technology and explain how NVIDIA plans on making code run as efficiently as possible on their hardware with less hand optimization.
Facebook
Twitter
RSS
114 Comments
View All Comments
Anonymous User - Saturday, October 25, 2003 - link
The irregularities ATi's drivers allegedly display in AquaMark 3 and UT2003 require further investigation. Factors such as image quality, driver reliability, and compatibility are hard to convey in a review anyway.this is from tom
so to al the ati lovers here go #"&' yourself
a few weeks/months ago you guys said that nvidia cheated
and now ati does you still have a big mouth
no i am not an nvidia lover
i am de performance lover ( for me ati maybe change with the nv 40) but you guys AAAARRRGGGGHHHH
btw i havent read page 4 and 5 too busy
Anonymous User - Friday, October 24, 2003 - link
Speaking of the Hardocp review. I did notice on the intial review at nvnews that their screen shot of Halo shows the blinking lights in the hallways that Hardocp said were not there. They are using the 5950 and the latest drivers, so it would seem one of the two sites made a booboo.Anonymous User - Friday, October 24, 2003 - link
Excuse me, but if a review without an IQ comparions is ok, why even care about video cards at all? If image quality isn't important because yu cant really notice it in games, go buy yourself a geforce3 and knock yourself out, since it will play everything fine and you dont have to worry about image quality because you can even turn it up, how about that?And you dont need to zoom in anything to see Nvidia's new filtering method which is now, worse than ATI's. The bilinear filtering is ESPECIALLY noticeable in motion and causes the same kind of effect that aliasing does. Go look in the hardocp for yourself, especially in nascar and flight sim
Anonymous User - Friday, October 24, 2003 - link
Wow. Anyone who whines about people demanding actual image quality comparisons is certainly NOT a gamer, or at least not one with decent hardware and eyes.I'll school all of your asses at UT and I'm damn sure not going to do it with dynamic detail reduction(TM), not for $200 or more.
If the IQ differences between the cards are so minimal, why is it readily obvious when playing FS2004 and TR:AOD and UT2003 which card you're playing on? I'll tell you why:
Because the ground textures on FS2004 look like crap, and trilinear filtering DOES NOT WORK AT ALL AFTER THE FIRST STAGE, REGARDLESS OF APPLICATION OR DRIVER CONTROL SETTINGS in D3D with the FX family.
Instead, we get slightly improved bilinear that looks visually inferior to trilinear by a mile.
And you know what?
It's a lot EASIER to see WHEN YOU"RE PLAYING THE GAME, because the texture 'horizon' is always moving.
Not that anyone who's fawning over AT would know. An FX5200 ain't gonna show you son.
Anonymous User - Friday, October 24, 2003 - link
watchu' talkin'bout willis?!watchu' talkin'bout willis?!
watchu' talkin'bout willis?!
WooDaddy - Friday, October 24, 2003 - link
#62, Live, I agree.Derek, thanks for the review. I really liked the fact that the Ti4200 was included. REALLY helpful. I think I can hold out for a while. If not, the ATI 9700PRO will be considered.
Jeff7181 - Friday, October 24, 2003 - link
Ok... I checked out the other reviews... and HardOCP's results differ from AT's... but Tom's look pretty much the same.AT did come up with a different conclusion though, saying the FX5700 is a better buy than the 9600XT. And I agree. I know many of you will get your shorts in a knot about this, but ATI's driver quality still isn't up to par with nVidia's. A friend of mine has had nothing but trouble getting his 9800 Pro to work correctly.
In my opinion, ATI will have to take a hefty lead in the performance area to make up for the driver problems to get their card into my rig.
Jeff7181 - Friday, October 24, 2003 - link
I'll have to take a look at the reviews by other sites... but personally, on my Aopen FX5900 @ 490/950, everything looks great. The quality is better than the 45.23's in my opinion. Taking a look at a still picture that is blown up 400X to compare individual pixels is stupid. What might look worse pixel per pixel may look better at normal size, frame by frame.Anonymous User - Friday, October 24, 2003 - link
ok..5700ultra seems fine... but i bought 5600ultra 1-2 months ago... what will happen to me... ????:(
Anonymous User - Friday, October 24, 2003 - link
i think we need a butt pirate joke right about now