Die size & speed limit

aznThunder

Distinguished
Apr 3, 2003
13
0
18,510
I dunno if this should go in the CPU or VPU forum, but here goes:

With the release of the X800, various sites have been saying that the approximate frequency "limit" for a 130nm process is about 600mhz (thus the x800s supposedly can't OC as high as the 6800s). However, this obviously can't be inherent to the 130nm process since 130nm CPUs go over 3ghz. So what am I not getting? Thanks guys.

"I've got a great new idea for a value product-line! We charge the customer the same price, and we give them a slower product!"
"How is that a value?"
"Oh, it's a value for us, of course."
 

TheRod

Distinguished
Aug 2, 2002
2,031
0
19,780
Transistors count, heat and how the 130nm is done can affect max speed. I'm not in expert of micro-chip, maybe someone else will give you more precise clue.

--
Lookin' to fill that <font color=blue>GOD</font color=blue> shape hole!
 

cleeve

Illustrious
It probably has more to do with the amount of transistors in the next-gen GPUs (much more than CPUs already), and the limited size that acceptable cooling devices for them can be.

It's not realistic to put something as big (or heavy) as a Coolermaster CPU cooler or a water cooling system on your everyday video card.

But of course it's theoretically possible to get clockspeeds higher than 600 Mhz with custom cooling jobs; hell, I'm pretty sure it's been done already. Maybe even with 9800XT's.
It's just not realistically do-able with mass-market acceptable parts, is all...

________________
<b>Radeon <font color=red>9500 PRO</b></font color=red> <i>(hardmodded 9500, o/c 340/310)</i>
<b>AthlonXP <font color=red>~2750+</b></font color=red> <i>(2400+ @2.2Ghz)</i>
<b>3dMark03: <font color=red>4,055</b>
 
thus the x800s supposedly can't OC as high as the 6800s
Whomever said/wrote that, has no idea what they are talking about. IF anything the opposite would be true. Due to the fact that it lacks the LOW-K/D process, the GF6800 will never reach the same speeds as the X800 under similar conditions, especially with the increased number of transistors.

Sounds like you read that bogus article Derth linked to.

Don't worry, there'll always be overhead as long as there are tweakers.


- You need a licence to buy a gun, but they'll sell anyone a stamp <i>(or internet account)</i> ! - <font color=green>RED </font color=green> <font color=red> GREEN</font color=red> GA to SK :evil:
 

Vapor

Distinguished
Jun 11, 2001
2,206
0
19,780
715MHz was already reached on the X800XTpe...with a phase-change cooler. Time to let FUGGER loose with LN and dry ice (or whatever else he uses).

I would expect that with a voltmod and a good air cooler that 625ish MHz is reachable, maybe higher if you mix in some luck.

Maxtor disgraces the six letters that make Matrox.
 

Vapor

Distinguished
Jun 11, 2001
2,206
0
19,780
Dude, you gotta work on that!! Get a new card or something (more voltage?)! People are getting up near 800mhz (even though the X800s love bandwidth more) on the core.

Maxtor disgraces the six letters that make Matrox.
 

entium

Distinguished
May 23, 2004
961
0
18,980
low-k dielectric is a component insulation which increase max frequency amounts about 20% of a given micron size and gives a 25% boost of power aswell. So lets say a chip is running at 400 mhz. It performs like its runing at 500 mhz and if the max is 600 mhz it can go to 720 mhz. As most of ya guys know that theoretical limit is not the actual limit. It isn't suggested to use a .13 micron chip with low-k over 725 mhz because of the heat that will be produced due to the resistance. Even with low k without a good cooler it will over heat the chip very quickly if its overclocked more then 725.
 

phial

Splendid
Oct 29, 2002
6,757
0
25,780
So lets say a chip is running at 400 mhz. It performs like its runing at 500 mhz


um. hah that is SO not true

-------
<A HREF="http://www.albinoblacksheep.com/flash/you.html" target="_new">please dont click here! </A>
<A HREF="http://www.subhi.com/keyboard.jpg" target="_new">This is you, interweb junky</A>
 

ytoledano

Distinguished
Jan 16, 2003
974
0
18,980
It's not realistic to put something as big (or heavy) as a Coolermaster CPU cooler or a water cooling system on your everyday video card.
True, but IMO that has to change.
Games today count for the vast majority of resource consuming applications - There are more people gaming than encoding files. And since the GPU is more important than the CPU in games, why is it still condidered just a card? I'd think that by now mother boards would have mounting holes for GPUs like they do for CPUs.

Does PCI-EX change anything that will let you mount a heavier HS on the vid card?

This is the end of my post. No wait, this is...
[-peep-].
 

entium

Distinguished
May 23, 2004
961
0
18,980
Well this is what ATI has been saying the the dielectric insulation process does improve efficiency by quiet a bit, but I've never seen numbers like that in any other type of chips. Many manufactures use it for those two reasons. If ya want just search the web for so articles on it you will see what I said is true.
 

entium

Distinguished
May 23, 2004
961
0
18,980
<A HREF="http://www.a1-electronics.net/General_Interest/2003/IBM_Low-k_Copper_Cu-11.shtml" target="_new">http://www.a1-electronics.net/General_Interest/2003/IBM_Low-k_Copper_Cu-11.shtml</A>

Here ya go IBM says they get a 35% increase in transitor proficiency
 

cleeve

Illustrious
I don't think that's what they mean by efficiency, entium.

If you run a low-k 9600XT at 300/300 and a non-low-k 9600 PRO at 300/300, they're going to have identical performance. And vice-versa (9600 PROs overclocked to 9600XT spec perform identical to XTs)

low-k might give more headway for clocking, but running the same calculations clock-for-clock on low-k VS non-low-k will get identical performance. At least that's what I've seen from the overclocked 9600 PROS out there (which are *not* low-k, if memory serves)...

Perhaps they mean they use less power clock-for-clock or something?

________________
<b>Radeon <font color=red>9700 PRO</b></font color=red> <i>(o/c 329/337)</i>
<b>AthlonXP <font color=red>~2750+</b></font color=red> <i>(2400+ @ 2208 Mhz)</i>
<b>3dMark03: <font color=red>4,876</b>
 

entium

Distinguished
May 23, 2004
961
0
18,980
oh yeah miss read it. I don't know though There are other articles out there that gave the 400 to 500 mhz example. But you are right I never saw any differences when low k chips are clocked at the same speed.

Yeah I've been miss reading it. When they are talking about transistor efficiency they are talking about power usage.

Camparing 400 to 500 mhz transitor effciency is equal with regular to low-k
 
If ya want just search the web for so articles on it you will see what I said is true.
What you said is a misinterpretation of the information provided, and not really 'true'.

First, the link to IBM describe a 35% increase from SOI, not low-k, different things. The mention of a <b>30</b>% increase of Low-K is not due to Xmhz performing better than the same Xmhz, it's due to the reduced cross-talk and efficiency of the chip (less noise, less heat generation and better signalling under more heat), which in turn ALLOWS you to usually increase the chips frequesncies without hiting those barriers at the same point. OR as IBM says <i><font color=blue>low-k dielectric <b>can</b> deliver up to a 30% boost in computing speed and performance.</font color=blue></i>. This is not due to the process in and of itself, but due to the new attainable higher frequencies because of the process. Also power consumption can also be the spin-off benifit.

It does NOT mean; "<i>So lets say a chip is running at 400 mhz. It performs like its runing at 500 mhz"</i>, not at all. A 400mhz/300mhz R9600XT (low-k/d) will perform the same as a stock R9600Pro(non low-k/d). The only difference is that the upper frequency limits are higher on the R9600XT.

At lower frequencies crosstalk is not so much of an issue, it's a geometric progression, which is only an issue as you approach the upper thresholds. The efficiency of low-k at equal lower frequencies will not show an appreciable difference. It may also allow you to pack on more transistors at the same frequencies, in which case the same frequency wil yield more, but that's due to the increased tranies (if that's their function).

Your initial response was correct, but you added that line which is definitely not right.

<A HREF="http://www-306.ibm.com/chips/technology/makechip/breakthroughs/lowk.html" target="_new">Here's IBM's descriptions</A>

and

<A HREF="http://www.spectrum.ieee.org/WEBONLY/publicfeature/feb03/film.html" target="_new">IEEE Blurb</A>


- You need a licence to buy a gun, but they'll sell anyone a stamp <i>(or internet account)</i> ! - <font color=green>RED </font color=green> <font color=red> GREEN</font color=red> GA to SK :evil:
 
Dang, I hate being at work, by the time I had a chance to post the original reply I wrote when I got in, you already covered it HOURS ago. :eek:

Oh, well, so be it.


- You need a licence to buy a gun, but they'll sell anyone a stamp <i>(or internet account)</i> ! - <font color=green>RED </font color=green> <font color=red> GREEN</font color=red> GA to SK :evil: