| Author |
Message |
Yousuf Khan
Guest
|
 Posted:
Sun Dec 12, 2004 4:18 am Post subject:
Some kind of IBM/AMD manufacturing breakthrough? |
|
|
A press release from IBM & AMD about an announcement about something
they won't say right now. Speculation is that it's a manufacturing
process breakthrough.
Forbes.com: The Week Ahead: Dec. 13-17, 2004
http://www.forbes.com/work/management/2004/12/11/cx_ss_1211peopletowatch.html
Yousuf Khan |
|
| Back to top |
|
 |
David Wang
Guest
|
| Posted:
Sun Dec 12, 2004 7:23 am Post subject:
Re: Some kind of IBM/AMD manufacturing breakthrough? |
|
|
Yousuf Khan <bbbl67@ezrs.com> wrote:
Perhaps it's just the talk about IBM's 45nm process at IEDM
next week.
--
davewang202(at)yahoo(dot)com |
|
| Back to top |
|
|
Yousuf Khan
Guest
|
| Posted:
Sun Dec 12, 2004 9:55 pm Post subject:
Re: Some kind of IBM/AMD manufacturing breakthrough? |
|
|
David Wang wrote:
| Quote: | Yousuf Khan <bbbl67@ezrs.com> wrote:
A press release from IBM & AMD about an announcement about something
they won't say right now. Speculation is that it's a manufacturing
process breakthrough.
Forbes.com: The Week Ahead: Dec. 13-17, 2004
http://www.forbes.com/work/management/2004/12/11/cx_ss_1211peopletowatch.html
Perhaps it's just the talk about IBM's 45nm process at IEDM
next week.
|
There used to be a time when they'd simply put a press release on this
sort of thing and be done with it. Now it's a big press conference style
event.
Yousuf Khan |
|
| Back to top |
|
|
Klaus Fehrle
Guest
|
| Posted:
Sun Dec 12, 2004 10:36 pm Post subject:
Re: Some kind of IBM/AMD manufacturing breakthrough? |
|
|
"Yousuf Khan" <bbbl67@ezrs.com> schrieb im Newsbeitrag
news:lM2dnczuAJGJ5SHcRVn-rA@rogers.com...
| Quote: | David Wang wrote:
Yousuf Khan <bbbl67@ezrs.com> wrote:
A press release from IBM & AMD about an announcement about something they
won't say right now. Speculation is that it's a manufacturing process
breakthrough.
Forbes.com: The Week Ahead: Dec. 13-17, 2004
http://www.forbes.com/work/management/2004/12/11/cx_ss_1211peopletowatch.html
Perhaps it's just the talk about IBM's 45nm process at IEDM next week.
There used to be a time when they'd simply put a press release on this
sort of thing and be done with it. Now it's a big press conference style
event.
Yousuf Khan
|
Forbes says its related to an IEDM presentation on wednesday.
Fwiw, here is the program:
http://www.his.com/~iedm/techprogram/Advance_Program.pdf
KF |
|
| Back to top |
|
|
Klaus Fehrle
Guest
|
| Posted:
Mon Dec 13, 2004 5:06 pm Post subject:
Re: Some kind of IBM/AMD manufacturing breakthrough? |
|
|
"Yousuf Khan" <bbbl67@ezrs.com> schrieb im Newsbeitrag
news:lM2dnczuAJGJ5SHcRVn-rA@rogers.com...
| Quote: | David Wang wrote:
Yousuf Khan <bbbl67@ezrs.com> wrote:
A press release from IBM & AMD about an announcement about something they
won't say right now. Speculation is that it's a manufacturing process
breakthrough.
Forbes.com: The Week Ahead: Dec. 13-17, 2004
http://www.forbes.com/work/management/2004/12/11/cx_ss_1211peopletowatch.html
Perhaps it's just the talk about IBM's 45nm process at IEDM next week.
There used to be a time when they'd simply put a press release on this
sort of thing and be done with it. Now it's a big press conference style
event.
Yousuf Khan
|
Here is the press release:
http://www.amd.com/us-en/Corporate/VirtualPressRoom/0,,51_104_543~91999,00.html
The potential of strains in the dimension announced has been known and
demonstrated in the lab long ago. Apparently, half of it has been opened up
this year, the other half to come next year. Rather a step than a leap, but
well, it could be quite significant when it comes to keep dualcore-chips in
a reasonable thermal envelope without compromising performance on serial
problems too much.
KF |
|
| Back to top |
|
|
Douglas Siebert
Guest
|
| Posted:
Mon Dec 13, 2004 5:40 pm Post subject:
Re: Some kind of IBM/AMD manufacturing breakthrough? |
|
|
"YKhan" <yjkhan@gmail.com> writes:
Well, it'll make transistors switch 24% faster. But that won't increase
clock rate by 24% (which I assume is what you meant by "performance")
since it won't help with wire delays at all.
Can anyone give us a ballpark estimate of how much this would help clock
rate keeping power constant?
--
Douglas Siebert dsiebert@excisethis.khamsin.net
"They that can give up essential liberty to obtain a little temporary
safety deserve neither liberty nor safety" -- Thomas Jefferson |
|
| Back to top |
|
|
YKhan
Guest
|
|
| Back to top |
|
|
Tom Knight
Guest
|
| Posted:
Mon Dec 13, 2004 11:30 pm Post subject:
Re: Some kind of IBM/AMD manufacturing breakthrough? |
|
|
Douglas Siebert <dsiebert@excisethis.khamsin.net> writes:
| Quote: | Well, it'll make transistors switch 24% faster. But that won't increase
clock rate by 24% (which I assume is what you meant by "performance")
since it won't help with wire delays at all.
|
With properly buffered wiring, it will improve wire delay per unit
length as well. Buffered wire delay per unit length is proportional
to the square root of RC*tau where R and C are wire resistance and
capacitance per unit length and tau is the inverter delay. So, this
will improve buffered wire speed by 11.4%. |
|
| Back to top |
|
|
Yousuf Khan
Guest
|
| Posted:
Tue Dec 14, 2004 4:40 am Post subject:
Re: Some kind of IBM/AMD manufacturing breakthrough? |
|
|
Tom Knight wrote:
| Quote: | Douglas Siebert <dsiebert@excisethis.khamsin.net> writes:
Well, it'll make transistors switch 24% faster. But that won't increase
clock rate by 24% (which I assume is what you meant by "performance")
since it won't help with wire delays at all.
With properly buffered wiring, it will improve wire delay per unit
length as well. Buffered wire delay per unit length is proportional
to the square root of RC*tau where R and C are wire resistance and
capacitance per unit length and tau is the inverter delay. So, this
will improve buffered wire speed by 11.4%.
|
That 11.4% figure is interesting, because some reports were saying that
it'll improve performance by 12% rather than the 24% stated in the press
release. I wonder if they were referring to this? Nah, reporters
couldn't be that clever. :-)
Yousuf Khan |
|
| Back to top |
|
|
Niels Jørgen Kruse
Guest
|
| Posted:
Tue Dec 14, 2004 6:05 pm Post subject:
Re: Some kind of IBM/AMD manufacturing breakthrough? |
|
|
YKhan <yjkhan@gmail.com> wrote:
| Quote: | Well, it looks like they found a new straining technique that increases
performance by about 24% on their process:
PC Pro: News: AMD and IBM claim 24 per cent chip speed breakthrough
|
There is a better description at
<http://news.zdnet.com/2100-9584_22-5487544.html?tag=st.prev>.
The current 970FX only has strain on N-channel transistors, so the
P-channel transisors will be running hard to keep up.
If IBM takes the opportunity to optimize for low voltage operation, we
could see a 970FX suitable for powerbooks in Q1 '05 and probably a
faster Xserve G5.
--
Mvh./Regards, Niels Jørgen Kruse, Vanløse, Denmark |
|
| Back to top |
|
|
YKhan
Guest
|
| Posted:
Wed Dec 15, 2004 4:40 am Post subject:
Re: Some kind of IBM/AMD manufacturing breakthrough? |
|
|
Niels Jørgen Kruse wrote:
| Quote: | The current 970FX only has strain on N-channel transistors, so the
P-channel transisors will be running hard to keep up.
|
Cool, so why was it harder to strain the P-channels than the
N-channels? Also what sort of specific functions were the P-channels
were used in, and which ones were the N-channels used in?
Yousuf Khan |
|
| Back to top |
|
|
YKhan
Guest
|
| Posted:
Wed Dec 15, 2004 4:40 am Post subject:
Re: Some kind of IBM/AMD manufacturing breakthrough? |
|
|
Niels Jørgen Kruse wrote:
| Quote: | YKhan <yjkhan@gmail.com> wrote:
Well, it looks like they found a new straining technique that
increases
performance by about 24% on their process:
PC Pro: News: AMD and IBM claim 24 per cent chip speed breakthrough
There is a better description at
http://news.zdnet.com/2100-9584_22-5487544.html?tag=st.prev>.
The current 970FX only has strain on N-channel transistors, so the
P-channel transisors will be running hard to keep up.
If IBM takes the opportunity to optimize for low voltage operation,
we
could see a 970FX suitable for powerbooks in Q1 '05 and probably a
faster Xserve G5.
--
Mvh./Regards, Niels Jørgen Kruse, Vanløse, Denmark |
|
|
| Back to top |
|
|
YKhan
Guest
|
| Posted:
Wed Dec 15, 2004 4:40 am Post subject:
Re: Some kind of IBM/AMD manufacturing breakthrough? |
|
|
Niels Jørgen Kruse wrote:
| Quote: | YKhan <yjkhan@gmail.com> wrote:
Well, it looks like they found a new straining technique that
increases
performance by about 24% on their process:
PC Pro: News: AMD and IBM claim 24 per cent chip speed breakthrough
There is a better description at
http://news.zdnet.com/2100-9584_22-5487544.html?tag=st.prev>.
The current 970FX only has strain on N-channel transistors, so the
P-channel transisors will be running hard to keep up.
If IBM takes the opportunity to optimize for low voltage operation,
we
could see a 970FX suitable for powerbooks in Q1 '05 and probably a
faster Xserve G5.
--
Mvh./Regards, Niels Jørgen Kruse, Vanløse, Denmark |
|
|
| Back to top |
|
|
Niels Jørgen Kruse
Guest
|
| Posted:
Wed Dec 15, 2004 2:42 pm Post subject:
Re: Some kind of IBM/AMD manufacturing breakthrough? |
|
|
YKhan <yjkhan@gmail.com> wrote:
| Quote: | Niels Jørgen Kruse wrote:
The current 970FX only has strain on N-channel transistors, so the
P-channel transisors will be running hard to keep up.
Cool, so why was it harder to strain the P-channels than the
N-channels?
|
It doesn't seem to be any harder, once you know how.
| Quote: | Also what sort of specific functions were the P-channels
were used in, and which ones were the N-channels used in?
Yousuf Khan
|
Start here
<http://www.realworldtech.com/page.cfm?ArticleID=RWT050802020022>.
--
Mvh./Regards, Niels Jørgen Kruse, Vanløse, Denmark |
|
| Back to top |
|
|
Bernd Paysan
Guest
|
| Posted:
Wed Dec 15, 2004 2:56 pm Post subject:
Re: Some kind of IBM/AMD manufacturing breakthrough? |
|
|
YKhan wrote:
| Quote: | Niels Jørgen Kruse wrote:
The current 970FX only has strain on N-channel transistors, so the
P-channel transisors will be running hard to keep up.
Cool, so why was it harder to strain the P-channels than the
N-channels? Also what sort of specific functions were the P-channels
were used in, and which ones were the N-channels used in?
Yousuf Khan
|
The task of N-channels is to drive towards GND, the task of P-channels is to
drive towards VDD. Strains on these channels need to be different: the
N-channel needs a wider lattice to improve mobility, the P-channel a
compressed lattice. So to have strains on both P- and N-channel, you need
two different materials.
On CMOS logic, if I had the choice, I'd rather strain P-channel than
N-channel. P-channel mobility is about a factor 3 less than N-channel
mobility, anyway (the theory tells you that the factor is less than 3, but
the practice tells you that 3 is a good approximation ;-). The gate
capacitance therefore is 1 part N-channel, and 3 parts P-channel. Improving
the 3 part helps a lot, improving the 1 part doesn't give you much.
On domino logic, N-channels do the real work, and the P-channel only
precharges nodes. So for domino logic, improving the N-channel is worth the
effort.
--
Bernd Paysan
"If you want it done right, you have to do it yourself"
http://www.jwdt.com/~paysan/ |
|
| Back to top |
|
|
|
|
|
|
FAQ
Memberlist
Register
Profile
Log in to check your private messages
Log in