G70 Core Clock Variance

P

pharma

Legend
Hilbert Hagedoorn and Unwinder have revealed the G 70 core clock is quite variable, jumping to about 40 MHz over the initial clock.

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.
Click to expand...

We currently think that the Pixel and Vertex pipes or triangle setup might be clocked differently to explain that differential. Or even Shader Clock 430 MHz/Rop Clock:430MHz/Geometry Clock:470MHz.
Click to expand...

http://www.guru3d.com/newsitem.php?id=2827
 
Interesting. Such technology definitively does have a lot of potential in the longterm. And this might also explain the lower-than-expected idle power usage. And this could also bring the question of what the "primary" clock of the RSX is, also documents seem to indicate 550Mhz all over the pipeline (which seems unlikely to me).
 
Well, it would only make sense. Having lower clockspeeds during idle usage would definitely help with power consumption. I'm not exactly sure why nVidia's reporting the lower clockspeed, though. It may be that, as suggested, the higher clockspeed is not maintained by the entire core.
 
Very interesting topic, as Chalnot wrote there must be some parts "only" that clocks, cause surely this would be something you would think nVidia would salute in their PR atleast.

One a related note the thing about the ASUS card that has 470MHz thats surely missleading AD and they will get alot of shit i guess.
Atleast from what i have seen in the shops here in sweden is that the manufacters differents there cards with clocks now instead of bundle/price
only.
 
Be interesting to know what is happening with the BFG OC; is it doing this 40 mhz thing or not.

And those who have *thought* they did like clock testing of NV40 vs G70 to test relative IPC. . that'll have to be revisited now.
 
Yeah, it does make our lives a bit more challenging when attempting to figure out what the hell the architecture is doing, but it's a good move for gamers in general.

Still, the real argument regarding per-clock performance largely revolves around speculation about what other product releases are going to be capable of. I'm not sure this actually does much to change that speculation.
 
What is that tool that people are using to monitor clocks? I've seen several screenshots but never a link.
 
digitalwanderer said:
How do they monitor it in game? Alt-tab out and check rivatuner real quick?
Click to expand...
Computers have memory, you know? And they can use that to keep a history of samples. ;)
 
so the 6800 ultra core is at 400 and the GTX is running at 470 instead of 430, that puts the core at 17% faster than the 6800 ultra. Most test show the GTX to be about 25% faster than the utra. It would seem to me, that most of the performance increase is comming from the increase of the core speed and not the 24 pipes.
 
Sinistar said:
so the 6800 ultra core is at 400 and the GTX is running at 470 instead of 430, that puts the core at 17% faster than the 6800 ultra. Most test show the GTX to be about 25% faster than the utra. It would seem to me, that most of the performance increase is comming from the increase of the core speed and not the 24 pipes.
Click to expand...

Well we don't know what's running at 470 so it's too early to judge. And where does the 50% increase in pipecount + extra MADD and MUL fit into your analysis ? I'm thinking the core speed is the least contributing factor to pixel shader performance (it should have a bigger impact on the vertex shaders)
 
trinibwoy said:
Sinistar said:
so the 6800 ultra core is at 400 and the GTX is running at 470 instead of 430, that puts the core at 17% faster than the 6800 ultra. Most test show the GTX to be about 25% faster than the utra. It would seem to me, that most of the performance increase is comming from the increase of the core speed and not the 24 pipes.
Click to expand...

Well we don't know what's running at 470 so it's too early to judge. And where does the 50% increase in pipecount + extra MADD and MUL fit into your analysis ? I'm thinking the core speed is the least contributing factor to pixel shader performance (it should have a bigger impact on the vertex shaders)
Click to expand...

What if the 16 rops are a bottleneck, would that make the core speed more important.

I really don't know what I am talking about. thats why I am asking.
 
geo said:
Be interesting to know what is happening with the BFG OC; is it doing this 40 mhz thing or not.

And those who have *thought* they did like clock testing of NV40 vs G70 to test relative IPC. . that'll have to be revisited now.
Click to expand...

I have the BFG 7800GTX, and yes it does this. It jumps to 501Mhz core speed whenever I enter a game.
________
Toyota tf109 history
 
Last edited by a moderator:
trinibwoy said:
. . .it should have a bigger impact on the vertex shaders. . .
Click to expand...

Yeah, I've been thinking about that lately, in the great clocks vs more units debate.
 
Sinistar said:
What if the 16 rops are a bottleneck, would that make the core speed more important.
Click to expand...

If the rops are bottlenecking it's probably due to bandwidth limitations so faster memory would be the key there. I'm sure it'll be sorted out soon enough and we'll have all the answers.
 
Thanks Pharma, I just frontpaged this one.

Did y'all see nVidia's response to Hiiiiilbert's inquiry?

Update 1:

Answer from NVIDIA:

Hey Hilbert,

As our chips become more advanced, we are implementing more complex clocking inside the chip. 430MHz is the primary clock speed of the chip and can be verified by fill rate tests.

We will work with Rivatuner to read the correct registers in order to report the right clock.

Hope this makes sense.

Now if you disect that answer (as vague as it honestly is) we can make note of the fact that NVIDIA uses "Primary" clock speed. Obviously, there's a secondary clockspeed running also. Very likely there are surely different clocks for different pipes.

If 430 remains the primary clock then why can Asus (check that here) advertise it as 470 ?

That would be somewhat misleading to you, the consumer as you think it's an uber Ultra version or something like that. They sincerely use this to advertise:

Engine Clock 470 MHz**
**NV clock(430MHz)+Geometric Delta clock(40MHz)

As it seems nobody can explain the 40 MHz differential properly, yet is is being used as marketing gimmick, consumers automatically assume it's 470 MHz where that's not 100% true.
Click to expand...
 
You must log in or register to reply here.
Back
Top