RSX: A G70 shrunk down to 90nm?

[mckmas8808]

I like this part.

Check back on the site for the full interview tomorrow, where we discuss unified shaders, the PlayStation 3, HDR, and the next generation of GeForce cards.

Ok *adds to favorites*

 

[MechanizedDeath]

G70 has 'many' clocks - Nvidia

"It's somewhat hard for us to say 'the core clock in G70 is this single number'", says Kirk. "We didn't want to be accused of exaggerating the clock speed, so we picked a conservative number to talk about the core clock speed. But, yes, that is just one of the multiple clocks."

"People have said that G70 doesn't have any new architecture, but that's not really true. It has new architecture, it's just not always visible.

"The chip was designed from the ground up to use less power. In doing that, we used a lot of tricks that we learned from doing mobile parts. The clock speeds within the chip are dynamic - if you were watching them with an oscilloscope, you'd see the speeds going up and down all the time, as different parts of the chip come under load."

Honestly...sounds like PR fud to me. It doesn't suggest that the clock is any higher than the 430MHz. I thought mobile chips only scaled the clock down, not up. If 430MHz was an average clock speed, that would be neat, but I'm not holding my breath. It could just be preventive damage control against the R520. I'll wait and see. PEACE.

EDIT: If Asus is calling it a "geometric clock delta" than that would indicate a multiplicative increase (essentially a percent increase). But the site notes a ~40MHz increase even for the overclocked part. It workd out to about an 8.8% increase each time. But I don't buy that. It would mean the G70 would jump up to 600MHz if it was running at 550MHz by default.

EDIT2: Maybe I'm misinterpreting that. I read "geometric clock delta" as being related to a geometric series, which increases by a multiplier. I don't think it necessarily related to the vertex shaders or anything. But then that NVidia response specifically notes the fillrate, and since the PS outnumber the VS 3:1, maybe that's why. Maybe the VS run ~8% faster than the PS. Or maybe this is just another snipe hunt. :?

 

[overclocked]

Honestly...sounds like PR fud to me. It doesn't suggest that the clock is any higher than the 430MHz. I thought mobile chips only scaled the clock down, not up. If 430MHz was an average clock speed, that would be neat, but I'm not holding my breath. It could just be preventive damage control against the R520. I'll wait and see. PEACE.

I dont agree with you as it is the community that has brought this up and not nVidia, which i think is strange. And as 430Mhz "is" the reference speed i dont see what your meaning.
There is no downsides as i see it so i dont get your negatism towards the technology.
Btw i agree that G70 isnt´t a whole new chip architecture but rather a refined and polished NV40 one if thats what you meant by "PR fud"..

 

[MechanizedDeath]

Honestly...sounds like PR fud to me. It doesn't suggest that the clock is any higher than the 430MHz. I thought mobile chips only scaled the clock down, not up. If 430MHz was an average clock speed, that would be neat, but I'm not holding my breath. It could just be preventive damage control against the R520. I'll wait and see. PEACE.

I dont agree with you as it is the community that has brought this up and not nVidia, which i think is strange. And as 430Mhz "is" the reference speed i dont see what your meaning.
There is no downsides as i see it so i dont get your negatism towards the technology.
Btw i agree that G70 isnt´t a whole new chip architecture but rather a refined and polished NV40 one if thats what you meant by "PR fud"..

Yeah, I changed my opinion a bit in the EDITs. But I still want to wait for official word on this beyond that brief response from NV. PEACE.

 

[seismologist]

The chip probably has a programmable multiplier in the PLL. They can change the core frequency on the fly depending on the mode. While at the same time shutting down other portions of the chip by clock gating.

 

[PC-Engine]

G70 has 'many' clocks - Nvidia

"It's somewhat hard for us to say 'the core clock in G70 is this single number'", says Kirk. "We didn't want to be accused of exaggerating the clock speed, so we picked a conservative number to talk about the core clock speed. But, yes, that is just one of the multiple clocks."

"People have said that G70 doesn't have any new architecture, but that's not really true. It has new architecture, it's just not always visible.

"The chip was designed from the ground up to use less power. In doing that, we used a lot of tricks that we learned from doing mobile parts. The clock speeds within the chip are dynamic - if you were watching them with an oscilloscope, you'd see the speeds going up and down all the time, as different parts of the chip come under load."

The chip probably has a programmable multiplier in the PLL. They can change the core frequency on the fly depending on the mode. While at the same time shutting down other portions of the chip by clock gating.

What's so special about that with regards to performance? Seems to me like Nvidia needs it to keep their chips cool and quiet. Not really something to write home about.

 

[V3]

What's so special about that with regards to performance?

They can iron out bottleneck, so the performance can improve. But yeah there is nothing really special or new in regard to this tech though. NV probably used them in previous chip too, just not to the extent as they have done with G70.

 

[mckmas8808]

Noticed some other people find out about this too.

I just verified it, during some new test sessions with core clock monitoring we noticed that the G70 although clocked at 430 standard will in 3D applications immediately jump towards 468 MHz. Even overclocked at 480 Mhz the clock will jump towards 522 MHz, again that ~40 MHz difference. What is going on in that G70 GPU huh ??

Link http://www.guru3d.com/newsitem.php?id=2827

Now then. Even if you overclock your GeForce 7800 GTX towards 500 MHz the actual realtime monitoring results would show back 540 MHz, again that 40 MHz difference

The new technology that we found by accident is in fact simply a new innovation, not a weird cheat or something like that.

Link http://www.guru3d.com/newsitem.php?id=2836

 

[PC-Engine]

Noticed some other people find out about this too.

I just verified it, during some new test sessions with core clock monitoring we noticed that the G70 although clocked at 430 standard will in 3D applications immediately jump towards 468 MHz. Even overclocked at 480 Mhz the clock will jump towards 522 MHz, again that ~40 MHz difference. What is going on in that G70 GPU huh ??

Link http://www.guru3d.com/newsitem.php?id=2827

Now then. Even if you overclock your GeForce 7800 GTX towards 500 MHz the actual realtime monitoring results would show back 540 MHz, again that 40 MHz difference

The new technology that we found by accident is in fact simply a new innovation, not a weird cheat or something like that.

Link http://www.guru3d.com/newsitem.php?id=2836

...and I'm sure you can explain why it's innovative right?

 

[seismologist]

G70 has 'many' clocks - Nvidia

"It's somewhat hard for us to say 'the core clock in G70 is this single number'", says Kirk. "We didn't want to be accused of exaggerating the clock speed, so we picked a conservative number to talk about the core clock speed. But, yes, that is just one of the multiple clocks."

"People have said that G70 doesn't have any new architecture, but that's not really true. It has new architecture, it's just not always visible.

"The chip was designed from the ground up to use less power. In doing that, we used a lot of tricks that we learned from doing mobile parts. The clock speeds within the chip are dynamic - if you were watching them with an oscilloscope, you'd see the speeds going up and down all the time, as different parts of the chip come under load."

The chip probably has a programmable multiplier in the PLL. They can change the core frequency on the fly depending on the mode. While at the same time shutting down other portions of the chip by clock gating.

What's so special about that with regards to performance? Seems to me like Nvidia needs it to keep their chips cool and quiet. Not really something to write home about.

Nothing special really. Maybe it's the only way they can get up to the higher clock speed without moving to a faster process.

 

[Carl B]

I really think it has more to do with running 'cool n quiet' than it does increasing the performance. That being said I don't know why people discount it so readily - I feel technology that furthers that front is absolutely essential, and I could see where NVIdia would be guarded/proud of it if they feel they've made some sort of significant advance in this area.

Certainly it's what we'll all take for granted, but what would the alternative without this technology have been? Especially for an environment like a console, better to have it than not.

 

[Shifty Geezer]

I'm skeptical of it's worth in console environment. On a PC if you're not playing 3D gaes then throttling back makes sense. But when are you not pushing the GPU to it's limits on a 3D console?

 

[London Geezer]

I'm skeptical of it's worth in console environment. On a PC if you're not playing 3D gaes then throttling back makes sense. But when are you not pushing the GPU to it's limits on a 3D console?

When watching movies?
Or navigate the menus?

 

[Shifty Geezer]

True. Forgot about that. Especially true if PS3 is to be Linux-box too.

 

[Carl B]

I'm skeptical of it's worth in console environment. On a PC if you're not playing 3D games then throttling back makes sense. But when are you not pushing the GPU to it's limits on a 3D console?

Well I'm just going under the assumption that the throttling is highly dynamic as NVidia seems to be implying, rather than one state or the other. The 40 MHz difference thing would actually imply the later, but David Kirk's comments would imply the former. I guess I'm just waiting for that interview to come out to make more of a judgment on it; I was just trying to convey that I think this tech has more to do with power and heat than it does performance per se.

 

[seismologist]

I'm skeptical of it's worth in console environment. On a PC if you're not playing 3D gaes then throttling back makes sense. But when are you not pushing the GPU to it's limits on a 3D console?

I dont have the block diagram of a G70 but isn't there some other functions on the chip as well like for windows and mpeg 4 acceleration?

 

[Titanio]

So, 430Mhz is the "minimum clock"?

Possible implications for RSX and/or the numbers given out for it at E3? Is 550Mhz its "minimum clock", and are the performance figures given for it taken at that clock?

Perhaps this is something that can be switched off, so to speak, if they can run everything at a desired speed (i.e. 550Mhz) all the time (at 90nm)?

 

[jvd]

Sounds like my r420 . It jumps up to 540/580 at times and down to 520/560 again .

Ati called it over drive when they did this with the 9800xt .

Its just a way to give you a little more performance . Nice feature in a way as if the gpu gets to hot or or certian parts of the chip aren't used they can be clocked back or turned off to prevent damage or waste power .

But its been done before as i mentioned with ati , amd does it cool n quiet and intel did it with the p4s

 

[MechanizedDeath]

Sounds like my r420 . It jumps up to 540/580 at times and down to 520/560 again .

Ati called it over drive when they did this with the 9800xt .

Its just a way to give you a little more performance . Nice feature in a way as if the gpu gets to hot or or certian parts of the chip aren't used they can be clocked back or turned off to prevent damage or waste power .

But its been done before as i mentioned with ati , amd does it cool n quiet and intel did it with the p4s

I think it's Kirk specifically refers to something akin to an asynchronous chip. The articles refer to different clocks for diffrent parts of the chip. Not scaling the entire clock, correct me if I'm wrong. ROPS and PS running at 430MHz, and the VS running at 470MHz? That's different than clock throttling we see now. ATI's Mobility line has had this clock scaling for years. NVidia's probably does the same. My 1.6GHz Centrino also runs from 0.6-1.6GHz.

That said, if it's a consistent 40MHz bump (not shown by the results), then that would be an arithmetic clock delta. Geometric series increase by a multiplier. Arithmetic series increase by addition. I don't know if Asus is using the word "geometric" to refer to a multiplicative increase (percent increase) or to refer to the geometry calculations getting a ramped clock. PEACE.

 

[ralexand]

Can't wait for that ps3 article. Could be what we all been waiting for, well at least most.