how do amd cpus keep getting faster

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
27,099
16,014
136
More cache, and possibly core enhancements.
 

DAPUNISHER

Super Moderator CPU Forum Mod and Elite Member
Super Moderator
Aug 22, 2001
31,705
31,607
146
Originally posted by: Markfw900
More cache, and possibly core enhancements.
Dual channel with the new nF4 board don't hurt either :)
 

imported_michaelpatrick33

Platinum Member
Jun 19, 2004
2,364
0
0
Everyone needs to remember as well that AMD always takes their time with a die shrinks and go from the bottom to the top. So their 90nm parts are going from the low end and notebook space and will migrate up to the high end as yields increase to the FX-57 next year.

It is interesting also that most reviewers are getting 2800 and the FX-53 on air cooling and raising the multiplier. I think AMD could almost start stockpiling choice FX-57 on 130nm if they wanted. If they did that mean their 130nm was only 600megahurtz behind the Intel 3.4 130nm fastest part. Pretty impressive engineering
 

iwantanewcomputer

Diamond Member
Apr 4, 2004
5,045
0
0
anybody know how fast the athlon 64 line will get up to on 90 nm. i've seen estimates of only 3.2-3.3 on air, but the 130 nm ones were only esimated to get to 2.2-2.4, and many of the fx 55's will probably be able to do 2.8 or more on air. it would be nice to see amd clock up as high as intel, :)
 

imported_michaelpatrick33

Platinum Member
Jun 19, 2004
2,364
0
0
Originally posted by: iwantanewcomputer
anybody know how fast the athlon 64 line will get up to on 90 nm. i've seen estimates of only 3.2-3.3 on air, but the 130 nm ones were only esimated to get to 2.2-2.4, and many of the fx 55's will probably be able to do 2.8 or more on air. it would be nice to see amd clock up as high as intel, :)


Yeah AMD appears to have the FX-55 overclock on air to 2800 in most of the reviews so that means that AMD could possibly (if they wanted and the roadmaps show a 90nm FX-55) get 2800 on 130nm which is only 600mghrtz behind Intel's 3.4 on 130nm. Yikes and wow to the AMD engineers

AMD only needs it looks like now to get to 2800 to 3000 at most because the dual cores will take over at that point. The 4200 has to be a 2600 part (and has to have 1meg L2cache) to keep the PR ratings (a 4200 operating at 2600 with 512K cache will still be a 4000 according to all the past AMD PR ratings) and I would imagine the 4400 part would be a 2800 part (also with the 1meg L2cache) to keep the PR ratings.

The 4200 will come out next year along with the FX-55 I imagine to keep the fX line ahead of the A64 line. When the 4400 comes out the FX will be dual core. I wish AMD had more Foundry space (come on Dresden) so they could make more chips (I think that is why the high end chips are so expensive [if they lowered prices to much they would sell more chips than they could make and cause a backlash], also making money of course they are a business interested in money) to take more marketshare away from Intel and more importantly build their own brandname recognition. Please do T.V. AMD. Spend some money please. I digress.
 

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
27,099
16,014
136
Originally posted by: DAPUNISHER
Originally posted by: Markfw900
More cache, and possibly core enhancements.
Dual channel with the new nF4 board don't hurt either :)

All 939 boards are dual channel. And with NF4, even if it supports 754, a 754 chip IS single channel, as the controller is on-chip.
 

DAPUNISHER

Super Moderator CPU Forum Mod and Elite Member
Super Moderator
Aug 22, 2001
31,705
31,607
146
Originally posted by: Markfw900
Originally posted by: DAPUNISHER
Originally posted by: Markfw900
More cache, and possibly core enhancements.
Dual channel with the new nF4 board don't hurt either :)

All 939 boards are dual channel. And with NF4, even if it supports 754, a 754 chip IS single channel, as the controller is on-chip.
You have a penchant for stating the obvious, but since you didn't include it in your first post I added it ;) Besides single channel and skt754 doesn't answer his question how it gets faster at the same speed now does it? DC and the new chipset along with the factors you mentioned are how it has gotten faster without clockspeed increases so was that just postcount++ by us both or what :laugh:
 

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
27,099
16,014
136
Sorry, dapunisher, I didn't understand. But yes, I agree with you ON-TOPIC (my fault), that DC and cache and core enhancements are the three ways AMD can make a K8 chip faster at the same clock speed.
 

glugglug

Diamond Member
Jun 9, 2002
5,340
1
81
I seriously think AMD will probably reach 4GHz on 90nm. They are taking their time getting there though, milking the lower speeds for all they're worth.
 

Concillian

Diamond Member
May 26, 2004
3,751
8
81
Originally posted by: darfur
how do amd cpus keep getting faster
While still running at only 2.4ghz???

I believe the REAL question is:

How do Intel P4 processors do so little while running at 3.6 GHz???

In the days of the PIII vs. Athlon, Athlons were slightly faster than PIIIs per clock. Not much though, things were pretty close to AMD MHz = Intel MHz.

Then Intel put out the P4. It was clocked at insanely high speeds like 1700 and 1800 MHz. Intel lost the race to 1000 MHz but was not going to lose the race to 2000 MHz. Unfortunately these cores clocked at such high speeds did not perform anywhere near what a comparable PIII would have.

Traditionally a new "generation" of architecture has yielded a better MHZ to work done ratio.

1 MHz on a 386 > 1 MHz on a 286
1 MHz on a 486 > 1 MHz on a 386
1 MHz on a Pentium > 1 MHz on a 486
1 MHz on a Pentium Pro (pretty much a PII) > 1 MHz on a Pentium
PII and PIII weren't really significant architecture changes.
And
1 MHz on a P4 WAY less than 1 MHz on a PIII

Because of this I feel you have to ask why Intel P4 processors do so little at such a high MHz rather than asking why AMD processors do so much at such a low MHz.

AMD is actually going the more traditional route, where an architecture change = more crunching power per MHz.
Given the history of the microprocessor in a PC, it should not be at all surprising that a 2.4 GHz A64 outpaces a 2.4GHz AthlonXP, which outpaced a similar speed Athlon, which out paced a similar speed K6, which outpaced a similar speed 5x86.
 

cyberknight

Senior member
Sep 3, 2004
378
0
0
Well I'm sure AMD could reach high clock speeds too if they were willing to sacrifice everything to increase the clock speed like Intel has.
 

BW86

Lifer
Jul 20, 2004
13,114
30
91
Originally posted by: glugglug
I seriously think AMD will probably reach 4GHz on 90nm. They are taking their time getting there though, milking the lower speeds for all they're worth.

i dont think amd is going to reach 4ghz(stock), they are going to be using multi cores before they ever reach 4ghz :p
 

glugglug

Diamond Member
Jun 9, 2002
5,340
1
81
Originally posted by: BW86
Originally posted by: glugglug
I seriously think AMD will probably reach 4GHz on 90nm. They are taking their time getting there though, milking the lower speeds for all they're worth.

i dont think amd is going to reach 4ghz(stock), they are going to be using multi cores before they ever reach 4ghz :p

Dual core will be in about 9 months.

4GHz is probably about 18 months away IMHO. So yes, they will have dual core first. And those dual cores will make it up to 4GHz I think.
 

Sahakiel

Golden Member
Oct 19, 2001
1,746
0
86
Originally posted by: glugglug
Originally posted by: BW86
Originally posted by: glugglug
I seriously think AMD will probably reach 4GHz on 90nm. They are taking their time getting there though, milking the lower speeds for all they're worth.

i dont think amd is going to reach 4ghz(stock), they are going to be using multi cores before they ever reach 4ghz :p

Dual core will be in about 9 months.

4GHz is probably about 18 months away IMHO. So yes, they will have dual core first. And those dual cores will make it up to 4GHz I think.

Just one question: Do you think Intel could make a 4GHz part on 90nm?
 

masshass81

Senior member
Sep 4, 2004
627
0
0
Originally posted by: Concillian
Originally posted by: darfur
how do amd cpus keep getting faster
While still running at only 2.4ghz???

I believe the REAL question is:

How do Intel P4 processors do so little while running at 3.6 GHz???

In the days of the PIII vs. Athlon, Athlons were slightly faster than PIIIs per clock. Not much though, things were pretty close to AMD MHz = Intel MHz....

...Because of this I feel you have to ask why Intel P4 processors do so little at such a high MHz rather than asking why AMD processors do so much at such a low MHz.

I think Intel only cares about acheiving high frequencies since the majority of people equate speed with higher Mhz..

For example (I'm sure we all see this), I'll browse the laptops at Best Buy and people are always looking at the 1Ghz-1.5Ghz Mobiles and say 'Wow that is so slow I bet, why don't can't they at least be 2Ghz like the Pentium 4s" or "Yeah, this Athlon system is cheaper than the Pentium 2.6Ghz because its only 2.0Ghz"

IMO, overall, Intel has the advantage because of their higher frequencies - even if they don't perform nearly as well as lower clocked cpus with more efficient architechture.

 

DAPUNISHER

Super Moderator CPU Forum Mod and Elite Member
Super Moderator
Aug 22, 2001
31,705
31,607
146
Originally posted by: masshass81
Originally posted by: Concillian
Originally posted by: darfur
how do amd cpus keep getting faster
While still running at only 2.4ghz???

I believe the REAL question is:

How do Intel P4 processors do so little while running at 3.6 GHz???

In the days of the PIII vs. Athlon, Athlons were slightly faster than PIIIs per clock. Not much though, things were pretty close to AMD MHz = Intel MHz....

...Because of this I feel you have to ask why Intel P4 processors do so little at such a high MHz rather than asking why AMD processors do so much at such a low MHz.

I think Intel only cares about acheiving high frequencies since the majority of people equate speed with higher Mhz..

For example (I'm sure we all see this), I'll browse the laptops at Best Buy and people are always looking at the 1Ghz-1.5Ghz Mobiles and say 'Wow that is so slow I bet, why don't can't they at least be 2Ghz like the Pentium 4s" or "Yeah, this Athlon system is cheaper than the Pentium 2.6Ghz because its only 2.0Ghz"

IMO, overall, Intel has the advantage because of their higher frequencies - even if they don't perform nearly as well as lower clocked cpus with more efficient architechture.
Since they canceled the 4ghz it is evident they will be leaving mhz=speed behind for awhile too :) I don't see any P4 is a can 'o whupass commercials anymore, I see the Blue Men promoting mobile Centrino tech. Damned savy IMO, since that is a higher growth market than the desktop is now, and corporate notebook purchases mean KA-CHING!

I am excited about what the future holds now, the repaved roadmap is far juicier than just waiting for the next speed bump ya know? Dual cores, SLI, and Longhorn could be an awesome departure from what we have seen since the race to 1ghz, and I'm stoked at the prospects :sun:

 

Macro2

Diamond Member
May 20, 2000
4,874
0
0
The REALLY REAL question is how Intel processors keep getting faster without increases in clockspeed........

and the answer is....

New optomised benckmarks! B0pco anyone?
 

imported_RobJ

Member
Jul 27, 2004
90
0
0
Originally posted by: masshass81

I think Intel only cares about acheiving high frequencies since the majority of people equate speed with higher Mhz..

For example (I'm sure we all see this), I'll browse the laptops at Best Buy and people are always looking at the 1Ghz-1.5Ghz Mobiles and say 'Wow that is so slow I bet, why don't can't they at least be 2Ghz like the Pentium 4s" or "Yeah, this Athlon system is cheaper than the Pentium 2.6Ghz because its only 2.0Ghz"

IMO, overall, Intel has the advantage because of their higher frequencies - even if they don't perform nearly as well as lower clocked cpus with more efficient architechture.

True. When i built my computer, people were like, oh cool, how many megahertz, and I'd be like "it's an amd 3500". They'd prolly just assume it was 3500 mHz and I'd be content that I didn't have to explain why 2.4>3.4.

And the speed limit on silicon has partially to do with the limit to how small the pathways can be etched on silicon. The wavelength of light required to do the lithography is getting so small that new sources are being sought. For instance, I've seen researchers attempt to use x-rays.

 

bigal40

Senior member
Sep 7, 2004
849
0
0
Originally posted by: glugglug

Dual core will be in about 9 months.

4GHz is probably about 18 months away IMHO. So yes, they will have dual core first. And those dual cores will make it up to 4GHz I think.

Dual core in 9 months?
Will dual core be socket 939? Is a dual core going to run on, say an, nforce4?
 

miketheidiot

Lifer
Sep 3, 2004
11,060
1
0
Originally posted by: bigal40
Originally posted by: glugglug

Dual core will be in about 9 months.

4GHz is probably about 18 months away IMHO. So yes, they will have dual core first. And those dual cores will make it up to 4GHz I think.

Dual core in 9 months?
Will dual core be socket 939? Is a dual core going to run on, say an, nforce4?

99% yes with a bios update.