Mind throwing an explaination on GPU-Z(software) variables??

So this is my GPU-Z screenshot.I need some clarification as to what exactly each variable heading(like ROP's,texture fillrate etc.) denote and what it means and how it contributes IN actual to a gpu's performance!


Cheers !!

 

I don't understand the question. Pixel and texture fillrate should be obvious. Those are the number of pixels and texels that the GPU can fill per second. The ROPs, or Render Output units, output the final pixel values.

As for the effect that it all has on real world performance, the answer is "it depends." On a lot of things.

 

well ofcourse you did understood the question for you wouldn't give such an apt answer without an understanding.BTW rop's?? still a confusion...
Cheers!

 

ROP

http://en.wikipedia.org/wiki/Render_Output_unit

 

Thank you,but i needed a general explaination in simple english from you tech guru's@NBR

 

My understanding of what a ROP does is it takes the output from the GPU and converts it to something that can be displayed by a 2d panel. The ROPs essentially tell the screen how to display what the GPU calculated.

 

All of that information is too specific to determine something as broad as "how will this contribute to performance in a real scenario".

It totally depends. It would probably be more useful to try and get a general understanding of performance of the chip you have than to try and calculate performance contributions of each specification. You can't assign a % of performance to each of these specifications. It's sort of like saying: ok, I've got this car.

1. How many wheels does it have? 4 wheels.
2. What type of engine does it have? V4 engine
3. How many cylinders does the engine have? 4 (this isn't extra information, it's a v4 engine)
4. How many main bearings does it have? 3 (all v4 engines have 3 bearings, doesn't give us any new information about the engine)
5. What's the red line RPM on the engine? 7000 RPMs

Given the performance of the car, how do each of these factors (1-5) contribute? Obviously, it's impossible to say. First off, we don't have nearly enough information about the car. Even if we did, we wouldn't have a consistent performance measurement for that car, because it is dependent on conditions. Tires and road and weather conditions will have a HUGE impact on performance.

Again, we would to know the weight of the car, how it's distributed, and a lot of other information about the wheels and the engine system and all sorts of other things to even think about seeing how the performance really works.

Even then, it wouldn't be meaningful to figure out to what degree each of the qualities of the system contribute towards end-performance, because the car is designed as a single unit, and many of the features of the system are designed to work together to produce a result.

ie- what percentage of performance do we get from using gas in our car vs. if we omitted the gas? if we omit the gas, how detrimental would that be to the performance of the car? Is the gas therefore responsible for 100% of the performance? even if we were talking about better gas vs. worse gas, we would need to either go back to the design to understand what's happening, or do tests, to see how that contributes to performance.

This is why, with car performance, we just look at benchmarks that have some relevance to us. Stopping distance, gas mileage, 0-60 time, that sort of thing. It's the same with graphics chips and other components. A lot of times, we'll look at frame rates in games that matter to us. If they aren't available, you can standardized benchmarks, which aren't as reliable or useful, but may be more available. The situation is, that the graphics system is so complex that the same situation applies as with the car. We would need to know so much more than the few facts on the spec sheet to do ANY type of performance analysis. We would need to know about the design of the GPU. From there, figuring out how each specification on the spec sheet contributes to performance would be a loaded question- the specs are mostly theoretical benchmarks with no meaning at all or VERY BASIC information about the chip, which could be looked anywhere and is akin to information like "number of wheels on car" "seating capacity" "gasoline type" etc.

---

The shaders are specialized processor cores for graphics operations. The ROPs do the final calculations on the pixels to prep them for presentation on the monitor. The pixel fill rate is just the number of ROPs * GPU core frequency. That's the theoretical maximum number of pixels that the graphics card can calculate each second if the ROPs were putting up pixels full time on each clock. This is not useful information, and it's certainly not NEW information - because it's just derived from we already have. Same story with texture fill rate (number of textured pixels processed theoretical maximum per second).

The memory is probably self explanatory, it's the amount of memory on the graphics chip available to the GPU. The bus width (128 bit) and memory type (ddr3) don't matter because they list the bandwidth. The former stats contribute to the bandwidth, which is the rate at which data can be transferred between the GPU's memory and the core / shaders themselves. The core clock is the clock rate of the core, which exerts control over the shaders. The shader clock rate is the rate at which operations are performed by the shaders. etc.

The bus interface determines the maximum bandwidth available between your main processor (and main memory) with the GPU (and it's processors and memory). This is not really relevant.

DX Support and the computing check boxes show the software features supported by the GPU.

 

V4 engine? I didn't know cars used V4? I thought most were inline 4 cylinder...

Well maybe some motorcycles.

 

You are correct, Mostly found in motorcycles now, but used to be in some cars. Inline 4 is what is used now in most cars.

 

That GPU has 96 shader processors, CUDA cores as Nvidia calls them or Stream Processors as ATI called them. The more the better, regardless. Your Graphics card has a maximum pixel and texture fillrate of roughly 13 billion pixels and texels a second. A texel is to a texture as a pixel is to your screen. At least that's the general idea.

Your Core and Shader clock are locked, the Shader clock being double the core for the GTX 400 line and later. These two control the fillrate of your textures and pixels, the higher the better here. However increasing these can cause instability and potentially crash your card, however it's unlikely this will cause any really serious damage since I've caused my GPU to crash several times.

Now, the number of ROPs denotes how high if a pixel fill rate your GPU can achieve, however the number of Texture Address Units, commonly referred to as TMUs, relates to how high of a texture fill rate your GPU can achieve. This number isn't displayed. You can usually tell how many TMUs your card has simply by looking at the ratio between your pixel fill rate and texture fill rate. Since they're the same, you have 16 Raster Operators and 16 TMUs for this card. Again, the more the better.

As a fun fact, you can usually find out how many ROPs a GPU has simply by looking at the fillrates and clock speeds. If the core clock is at 800 mhz, and the pixel fill rate is at 19.2 GPixels/s, while your GPU is at the same core speed but has 12.8 GPixels/s for it's fill rate, you can note that the faster model has 24 ROPs. Relativity is your friend most of this.

The Revision number, Transistor count, Die size, Bios version, etc, are all things you don't have to worry about as that's just generic information about the physical components and software that drives your graphics card.

The Memory Size, commonly referred to as Vram or the GPU's Framebuffer, shows the physical amount of memory that is dedicated to your Graphics Card alone. Meaning your GT 435M has a huge 2GB Vram slab that games can manipulate to whatever means. More is generally better here, however I will tell you how there can be too much. Your Bus-Width and total memory bandwidth matters more than the actual amount of physical Vram your card has too.

If your card doesn't have a high enough flow of bandwidth coming and going from the Vram, the same thing happens there that happens when you're on a slow internet connection or a slow computer, meaning that full 2GB may not be used in games. Performance slows down the more Vram is used on that card because of this lower bandwidth. I would recommend overclocking the memory clock your GPU has to further increase this as you will see a much greater increase in overall performance in games and other programs simply because your GPU will be able to handle using more video ram.


I hope this helps clear things up a bit.

Here's a bit of proof that bandwidth matters a lot. My 3DMark Vantage run with the clocks in the picture below. My 460M is running as fast as a stock GTX 470M, yet my pixel and texture fill rates are lower than a GTX 560M. Bandwidth matters a lot for these cards.

NVIDIA GeForce GTX 460M video card benchmark result - Intel Core i5-460M Processor,TOSHIBA Qosmio X505 score: P8243 3DMarks

 

Epic explaination.Thanks man

 

This should work the other way: ROPs (important) * clock speed (important) = theoretical fill rate (not important).

Technically, you could solve for any of those 3 if you knew the other two.

 

Agreed. Gotta remember it's about relativity...