G80 Pics

wannabedamned · 2 Oct 2006 at 09:38

lol, I'm in love.

For people who dont want to buy due to power consumption, More power..Needs more power. You can't get something for nothing.

Tim Allen would say it best.. MORE POWER... Arrh! Arrh! Arrh!

Stelly · 2 Oct 2006 at 09:55

Thats a bit of a monster

Stelly

The $6m Dan · 2 Oct 2006 at 11:46

As I remember the 6800s had two power connectors when they were in development, and ended up with only one.
Possibly (hopefully) the same will happen here, and once they've got everthing ironed out, 1 power connector will be enough.

zo1d · 2 Oct 2006 at 12:37

Cool & Quiet for GPU's??

Is it not possible for graphics card manufacturer to implement a version of cool & quiet for gpu's, reducing clock speeds and voltage as needed? For the most part unless you are gaming, you only need a small fraction of the cards performance.

Kezmo · 2 Oct 2006 at 12:43

^^ They have done with the x1k's , thats why the cards have a 2d and 3d mode

Arcane · 2 Oct 2006 at 12:44

I'm guessing that's what the different 2D/3D clock speeds/modes are for.

fish99 · 2 Oct 2006 at 13:30

wannabedamned said:
For people who dont want to buy due to power consumption, More power..Needs more power. You can't get something for nothing.

7600 GT is faster than a 6800 Ultra and uses half the power. Even a 7950 GT uses less power than a 6800 Ultra. So it's not set in stone that every generation of cards must use far more power. Core 2 Duo uses less power than a Prescott P4.

Duff-Man · 2 Oct 2006 at 14:19

fish99 said:
7600 GT is faster than a 6800 Ultra and uses half the power. Even a 7950 GT uses less power than a 6800 Ultra. So it's not set in stone that every generation of cards must use far more power. Core 2 Duo uses less power than a Prescott P4.

No, but it is a general rule of thumb.

Performing calculations (whether using a CPU, GPU or whatever else) involves the 'ordering' of electrons, resulting in a local decrease in entropy. Hence, the second law of thermodynamics requires the production of heat. Even if we were able to perform the calculations with 100% efficiency, we would still neccesarily produce heat, and the amount of heat should be proportional to the processing power.

Now, of course we're nowhere near 100% efficiency, and various factors can help increase this efficiency (different materials, design improvements, but mainly the process size - 90nm will generally be more efficient than 130nm etc). However, it's always a losing battle against the fundamental laws of thermodynamics to produce cards (or chips) which consistently out perform their predecessors within a given thermal envelope.

Gashman · 2 Oct 2006 at 14:59

no because hes right, newer generations are starting to become more power efficient, for instance as he said the 7600GT (have two) use chunks less power than the 6800 ultra yet are faster at stock, mine are overclocked quite a lot and still don't use more than 75W (PCI-E, no extra power) same with conroe, uses less power than its northwood/prescott predecessors and has hugely improved processing power, clock speed is the real killer when it comes to power. its like superluminal space travel, according to the advanced theory of relativity its impossible, power requirements would be infinitely large, simple solution...don't go faster than light at all, in star trek, the ship doesn't physically move, it merely distorts space around it, riding on a 'wave' of space/time, so as processors start to do more work for there clocks they'll get faster with negligable power increases, sorry what can i say, im a science geek

Duff-Man · 2 Oct 2006 at 16:19

Gashman said:
no because hes right, newer generations are starting to become more power efficient, for instance as he said the 7600GT (have two) use chunks less power than the 6800 ultra yet are faster at stock, mine are overclocked quite a lot and still don't use more than 75W (PCI-E, no extra power) same with conroe, uses less power than its northwood/prescott predecessors and has hugely improved processing power, clock speed is the real killer when it comes to power. its like superluminal space travel, according to the advanced theory of relativity its impossible, power requirements would be infinitely large, simple solution...don't go faster than light at all, in star trek, the ship doesn't physically move, it merely distorts space around it, riding on a 'wave' of space/time, so as processors start to do more work for there clocks they'll get faster with negligable power increases, sorry what can i say, im a science geek

Newer CPUs / GPUs are using less power for a given performance due to increases in efficiency. Like I said, this is mainly down to more intelligent design and smaller manufacturing processes. However the general principle remains. You could represent the power requirement of a CPU as follows:

Power required = (power required to maintain local entropy drop) / efficiency.

The power required to maintain the local entropy drop is (courtesy of thermodynamics) directly proportional to the amount of ordering performed - ie the theoretical maximum performance of the CPU. Also remember all the energy used by the CPU will be distributed as heat.

So, in order to produce a more powerful CPU which uses the same amount of power (or less) you need to improve the efficiency of the CPU. Now, with CPU / major GPU revisions being done every couple of years in general, the field has often developed enough to allow more intelligent and efficient designs to be created - along with manufacturing process refinements - which sometimes allow this to happen. However, it is always going to be like fighting a losing battle - we cannot go on increasing design efficiency indefinitely at the sme rate as we increase CPU power, and there are quantum-mechanical limits to how small the manufacturing process can be.

Consider the power requirements of a conroe with those of a first-edition P4, and a first edition P3 etc. I think you'll find a steep downward trend, even though each new processor family was more power efficient than the top-of-the-line model it replaced.

A couple more things to bear in mind:

1) About clockspeeds and power requirements - it is more efficient (in terms of the above) to go for a multi-core approach than to increase clockspeed. This is because with current CPU designs, (say) doubling clockspeed requires more than double the power, whereas adding a second core will require almost exactly double the power. Hence it makes sense to focus on a lower speed multi-core approach. This is one of the reasons intel has been able to gain such a lot going from prescott -> conroe.

2) Video cards have been increasing in processing power much more rapidly than have CPUs. As such, you would also expect the power requirements to rise more rapidly also.

Boogle · 2 Oct 2006 at 16:58

Duff-Man said:
Hence it makes sense to focus on a lower speed multi-core approach. This is one of the reasons intel has been able to gain such a lot going from prescott -> conroe.

Intel gained a lot since they went to a completely different architecture. The second core is barely used most of the time in consumer applications. Thats changing, but don't under estimate a new design.

GPUs are currently designed via computer, CPUs are designed by hand. If GPU refreshes could be extended to at least 3 years then maybe GPUs could be designed by hand too.

Sure adding more means more heat, but thats a very brute force approach. Lets use a basic example:

If you increase the size of an internal combustion engine, it will become more powerful. You'll get a higher top speed, faster acceleration, albeit with diminishing returns. However, what if you add lubrication? Suddenly you can use a smaller engine and still get better performance. What about slick tyres? Now you have much better acceleration, breaking, handling.

A more efficient design yields better performance, less heat, and requires less power.

Incidentally multi-core CPUs won't last very long, at least not in the current design. Adding more cores is crude, and again, results in diminishing returns. Current multicore designs are basically multiple CPUs glued together, which maybe more efficient communications (Athlon64/Core Duo), and/or sharing some cache (Core Duo). Future will be adding lots and lots of units, similiar to traditional x86 design. However, the method of execution will be different, more like SMP so that the units can be utilised properly. Again - a much, much more efficient design.

One day, one day a new IC tech will come along. That day will be a great day.

manoz · 2 Oct 2006 at 18:43

Am I the only one who noticed the fact that you can see the table through the fan enclosure through the card and in the picture with the card turned around it's completely solid :confused:

pinkacidbootson · 2 Oct 2006 at 18:56

manoz said:
Am I the only one who noticed the fact that you can see the table through the fan enclosure through the card and in the picture with the card turned around it's completely solid

If you look at the last picture, you can see that the surface under the fan is metal, which just happens to look almost the same as the surface of the table in the first picture.

fornowagain · 2 Oct 2006 at 18:59

manoz said:
Am I the only one who noticed the fact that you can see the table through the fan enclosure through the card and in the picture with the card turned around it's completely solid

It looks like an alloy back plate to me. If you look behind the fan, there is a gap with the green pcb showing.

manoz · 2 Oct 2006 at 21:49

fornowagain said:
It looks like an alloy back plate to me. If you look behind the fan, there is a gap with the green pcb showing.

No I mean look at this picture

then look at this picture

If you look at the fan area you can see that the top pic implies that the card is see through yet the card is solid from the back.(unless of course the backplate is made from the same material from the table which I highly doubt :confused:

)

Sumanji · 2 Oct 2006 at 21:57

That is the base of the heatsink, it just appears the same colour as the table dude...

manoz · 2 Oct 2006 at 22:04

Ok I realise now I'm a bit paranoid.

. Stupid coincidences............

VeNT · 2 Oct 2006 at 22:26

its like that quad core Voodoo card, MOOSIVE!
doubt the relese cards will look anything like that in real life!

no_1_dave · 2 Oct 2006 at 23:04

I think its fake.
Nvidia would never make a card that long & ugly as a main stream product. Also, that much power? No way, wouldnt even expect that from ATI...

fornowagain · 2 Oct 2006 at 23:05

manoz said:
No I mean look at this picture

If you look at the fan area you can see that the top pic implies that the card is see through yet the card is solid from the back.(unless of course the backplate is made from the same material from the table which I highly doubt )

I know what you were talking about. I'm sat here looking at it, it's a shade of silver that looks close to the beige of the table is all.