Sign in with
Sign up | Sign in
Your question

"FX-64 To Be Released The 8th Of August"

Tags:
Last response: in CPUs
Share
June 21, 2006 6:28:09 PM

Quote:
HEXUS can report that the AMD Athlon FX-64 will be available to system integrators in the UK at a price point of $999 per CPU from the 8th August 2006.


http://www.hexus.net/content/item.php?item=5990

More about : released 8th august

June 21, 2006 6:45:25 PM

Quote:
HEXUS can report that the AMD Athlon FX-64 will be available to system integrators in the UK at a price point of $999 per CPU from the 8th August 2006.


http://www.hexus.net/content/item.php?item=5990
cool, i will not buy it
June 21, 2006 6:48:06 PM

8) :lol:  :lol:  :lol: 

Quote:
Intel is about to unleash a complete range of dual-core CPUs, in volume, ready for launch day. Until the transition to 65nm (Valentine's Day 2007?) and a core re-jig, AMD's best defense is speed bumps.


Interesting, another indicator that Conroe will be available in volume at launch day!!!
Related resources
June 21, 2006 7:03:46 PM

Quote:
8) :lol:  :lol:  :lol: 

Intel is about to unleash a complete range of dual-core CPUs, in volume, ready for launch day. Until the transition to 65nm (Valentine's Day 2007?) and a core re-jig, AMD's best defense is speed bumps.


Interesting, another indicator that Conroe will be available in volume at launch day!!!
I bet 9nm just posted the thread without reading it. The article says bad and scary things about his love.
June 21, 2006 7:17:03 PM

Quote:
HEXUS can report that the AMD Athlon FX-64 will be available to system integrators in the UK at a price point of $999 per CPU from the 8th August 2006.


http://www.hexus.net/content/item.php?item=5990
cool, i will not buy it

don't think I will either...its time for a new Intel Machine I think......however I will prolly buy AMD again....just dont know when
June 21, 2006 7:20:19 PM

Holy crap imagine how much power it'll use.
June 21, 2006 7:27:15 PM

Quote:
Holy crap imagine how much power it'll use.

ENIAC was spending 174kW, I bet the FX-64 will spend less. :lol: 
June 21, 2006 7:32:00 PM

Quote:
HEXUS can report that the AMD Athlon FX-64 will be available to system integrators in the UK at a price point of $999 per CPU from the 8th August 2006.


http://www.hexus.net/content/item.php?item=5990

Thats cool so I guess the FX-62 to be in the low $800 to high $700. I think AMD is replacing its X2's with the FX's and its single core 64's with the X2's. Price wise its sounding good.

The FX-64 should outperform the E6600 which is somewhat out of the price range but will have to wait an see. AMD needs a CPU to combat the E6700 and X6800 extreame. I would guess atleast a FX-68 will be needed to even get pass the E6700 which isn't good news.
June 21, 2006 7:41:12 PM

I would be surprised if AMD prices the FX-64 to as low as $800 at launch, regardless of Conroe...
AMD, imo, just can not sell them that low and make money to... especially so considering their volumes...
June 21, 2006 7:44:35 PM

Quote:
HEXUS can report that the AMD Athlon FX-64 will be available to system integrators in the UK at a price point of $999 per CPU from the 8th August 2006.


http://www.hexus.net/content/item.php?item=5990

Thats cool so I guess the FX-62 to be in the low $800 to high $700. I think AMD is replacing its X2's with the FX's and its single core 64's with the X2's. Price wise its sounding good.

The FX-64 should outperform the E6600 which is somewhat out of the price range but will have to wait an see. AMD needs a CPU to combat the E6700 and X6800 extreame. I would guess atleast a FX-68 will be needed to even get pass the E6700 which isn't good news.

What are you smoking, AMD cannot replace the X2 series with FX chips, FX denotes the best, having 4 versions of "the best" doesnt make sense, especially when they will still be beaten by "regular" Intel Cpus.

plus the FX-64 is probably going to be a 3Ghz chip, which might finally allow it to compete with a E6600, and it will be lower in yield than the FX-62 given the increased clock.

AMD is not going to be producing more than 1 dual core FX chip for the AM2 platform, what you might see is a 512K cache 2.8ghz 5200+ or something.
June 21, 2006 8:06:57 PM

Do you see this guys? in the past 3 month we got FX-60, FX-62 and now FX-64.
They are seeping up their production, espacilly when we think that the move from FX-51 to FX-57 took almost 2 years!
They are speeding up, higher speeds might be their only diffence from Conro.
If the FX-64 is 3.0, how high can it OC?
It might not be such a blood bath for AMD after all, but it is a must that they lower their prices.
OH man, I want a 939 FX-60 so bad and I might be able to buy it after all :) 
a b à CPUs
June 21, 2006 8:09:49 PM

A bit funny reading about the "cpu battle" Funny how a Intel is so much better, but cannot be bought. It's like nonne actually reads TOMS HARDWARE, just blabs.
By the time the Intel is out in force with their new cpu, AMD probably will be using the 65nm. Just the change to 65nm will kill the advantage of the new Intel cpu. I'll wait for the cpus to actually start shipping (both 65nm AMD/Intel cpus).
Ever read, "Don't count your chicks before they hatch?"
June 21, 2006 8:13:24 PM

Quote:
Do you see this guys? in the past 3 month we got FX-60, FX-62 and now FX-64.
They are seeping up their production, espacilly when we think that the move from FX-51 to FX-57 took almost 2 years!
They are speeding up, higher speeds might be their only diffence from Conro.
If the FX-64 is 3.0, how high can it OC?
It might not be such a blood bath for AMD after all, but it is a must that they lower their prices.
OH man, I want a 939 FX-60 so bad and I might be able to buy it after all :) 


Answer, not very much at all, similar to the FX-60 and FX-57...
Overclocking is not the FX series strongpoint... for you are already buying AMD's top chip overclocked... just done at the factory...
June 21, 2006 8:13:24 PM

Fast, but it'll use a lot of energy and probably won't overclock to well.
June 21, 2006 8:39:18 PM

Do you ever read what you type?

You are hoping to afford an FX-60?

How about as soon as the conroe E6600 ships buy that for MUCH cheaper "since for you and me the money does count" and performs as well or better at stock speeds and then can overclock like mad!!

Why buy an FX-60 for the sake of saying I own an FX-60? Is this a badge for an AMD fan or something? Just curious?

Why not buy what is affordable and has the best perfomance? Be it AMD or Intel?

I just will never understand the "Brand Loyalty" thing?
June 21, 2006 8:40:36 PM

Wow, someone that actually is thinking here, BRAVO, more of your kind are really needed here.
So so tired of these idiots.
June 21, 2006 8:46:00 PM

Master, Teeee Heeeee

What about his comments points to a significant increase in performance? A die shrink may result in better power handling but the Uarch is the same isn't it?

Did I miss something? Does a die shrink to 65nm magically mean they can be that much faster?

Please enlighten me OH Master Teeee Heeee!
June 21, 2006 8:54:11 PM

Quote:
Do you ever read what you type?

You are hoping to afford an FX-60?

How about as soon as the conroe E6600 ships buy that for MUCH cheaper "since for you and me the money does count" and performs as well or better at stock speeds and then can overclock like mad!!

Why buy an FX-60 for the sake of saying I own an FX-60? Is this a badge for an AMD fan or something? Just curious?

Why not buy what is affordable and has the best perfomance? Be it AMD or Intel?

I just will never understand the "Brand Loyalty" thing?


It is not, I just want the fastest 939 because I don't want to change my RAM's and Mobo
I was using P4C 3.2 when it was kiking Athalon XP, then I swithced to Athalon 64 4000+
It is as simple as what is best for me out there.
a c 99 à CPUs
June 21, 2006 9:06:14 PM

Here's what I think AMD will have at Conroe's real widespread launch late this year:

Athlon 64 Line:
3500+ (2.2 GHz, 2x512K) $80-100
3800+ (2.4 GHz, 2x512K) $110-130

X2 Line:
3800+ (2.0 GHz, 2x512K) $150
4200+ (2.2 GHz, 2x512K) $200
4600+ (2.4 GHz, 2x512K) $250-300
5000+ (2.6 GHz, 2x512K) $325-400
5400+ (2.8GHz, 2x512K) $450-550
5800+ (3.0 GHz, 2x512K) $600-750

FX Line:
FX-64 (3.0 GHz, 2x1MB) $850-900
FX-66 (3.2 GHz, 2x1MB) $999



The AMD chips will be a roughly as expensive clock-for-clock as the Core 2s, but I bet these will be the prices and offerings that are out there and will sell at an okay rate.
June 21, 2006 9:41:53 PM

Quote:
Holy crap imagine how much power it'll use.

Somehow that's precisely what came to mind as soon as I saw the topic. I'll pass on the FX64 and grab something more reasonable like the E6600.

I've been a big fan of AMD for a while now, but it's pretty much set that I'll soon be the proud owner of a Cornrow rig, and that fact gets more firmly cemented every time I read a new article about it. As an engineer, my purchase decisions are based on a combination of performance and overall quality of design, and from what I have seen so far there's just no denying that intel's engineers really stepped it up and created a quality product. Even more appealing is the fact that ATi seems to have created a top-notch chipset to go along with it. I've been given every reason to go all out with this computer build 8)
June 21, 2006 9:47:39 PM

For the price differential between the 60 and the 6600 you should be able to afford a MB and Memory! And still have enough pieces to start another PC.
June 21, 2006 10:13:24 PM

Quote:
I would be surprised if AMD prices the FX-64 to as low as $800 at launch, regardless of Conroe...
AMD, imo, just can not sell them that low and make money to... especially so considering their volumes...

believe I stated the FX-62 because the FX-64 will sale for $999 or so the artical states.
June 21, 2006 10:29:18 PM

Quote:
HEXUS can report that the AMD Athlon FX-64 will be available to system integrators in the UK at a price point of $999 per CPU from the 8th August 2006.


http://www.hexus.net/content/item.php?item=5990

Thats cool so I guess the FX-62 to be in the low $800 to high $700. I think AMD is replacing its X2's with the FX's and its single core 64's with the X2's. Price wise its sounding good.

The FX-64 should outperform the E6600 which is somewhat out of the price range but will have to wait an see. AMD needs a CPU to combat the E6700 and X6800 extreame. I would guess atleast a FX-68 will be needed to even get pass the E6700 which isn't good news.

What are you smoking, AMD cannot replace the X2 series with FX chips, FX denotes the best, having 4 versions of "the best" doesnt make sense, especially when they will still be beaten by "regular" Intel Cpus.

plus the FX-64 is probably going to be a 3Ghz chip, which might finally allow it to compete with a E6600, and it will be lower in yield than the FX-62 given the increased clock.

AMD is not going to be producing more than 1 dual core FX chip for the AM2 platform, what you might see is a 512K cache 2.8ghz 5200+ or something.
I believe at 1 point the single core 64's were "the best" and they got replaced. All the FX denotes to me is higher amounts of cache and "the best" FX chip is in question. The FX-57 kicks the crap out the FX-60 or FX-62 in games.
I really dont see them making 1 chip only for the 4X4 mobo because that will kill the 4X4 from the start. Answer me this why is the sempron over taking the single core 64's for? Change is a comming. What is the differance between the FX-62 and X2 5200? The cache and when the CPU hits the Ghz limit AMD will increase the Cache for more performance. I see AMD making for FX CPU's because the X2 are as close to the Ghz limit as the FX's.
June 21, 2006 10:33:21 PM

Are you serious, I hope you do know that switching from process technologies gives no tangible performance boost. AMD wont be winning any time soon no matter what you think. The FX 62 is maxed out, the overclockability of the chip is simply lame. The FX64 isnt going to be a marvel. Have you forgotten about the 3.2 ghz XE conroe ?, the lead will be even bigger, and intel has an overclocker on their hands. Theres alot more overhead. I personally say intel should max out the conroe, to its absolute limit, and pound amd while they are down, thats competition not releasing products which they could on launch but dont. They will have a great lead, make that lead bigger, and keep hitting hard.
a c 99 à CPUs
June 21, 2006 10:37:47 PM

A switch in process technology given no real performance benefit AT THE SAME CLOCK SPEEDS. However, it reduces the die size of the chip, making it cheaper to produce, and it also allows for clock speeds to be raised, and that does increase performance. Also, with a good die shrink, one can fit more cache on the die and keep sizes roughly the same as before or even a little smaller.
June 21, 2006 10:52:13 PM

Quote:
For the price differential between the 60 and the 6600 you should be able to afford a MB and Memory! And still have enough pieces to start another PC.

Your correct but having to give your current mobo and RAM makes the prices between the 2 get tight. Intel more than likly will get the customers without AMD 64 systems and thous with older 64 systems which the mobos will not work with the X2 or FX CPU's. The E6600 sounds good to me because all I have is a celoron 1.8 with a Nvidia 5500 GPU.
June 21, 2006 10:56:09 PM

Quote:
A switch in process technology given no real performance benefit AT THE SAME CLOCK SPEEDS. However, it reduces the die size of the chip, making it cheaper to produce, and it also allows for clock speeds to be raised, and that does increase performance. Also, with a good die shrink, one can fit more cache on the die and keep sizes roughly the same as before or even a little smaller.

I'm guessing AMD maybe got about 600 to 800Mhz extra over the 90nm process. I dont see AMD topping the X6800 by increasing to the higher Ghz but may could with a larger cache.
June 21, 2006 11:04:31 PM

Only minor boosts in clock speeds will be acheivable if at all, that comes when the manufacturing is more matured, and amd hasnt even got 300 nm wafers online making cash off of these. So im not really to confident in their manufacturing process, even if their processors are being produced in the smartest country (germany). Amd also announced that they werent going to produce very many 2x1 mb cached processors. They are focusing more on 2x512 kb.
June 21, 2006 11:25:11 PM

Quote:
For the price differential between the 60 and the 6600 you should be able to afford a MB and Memory! And still have enough pieces to start another PC.

Your correct but having to give your current mobo and RAM makes the prices between the 2 get tight. Intel more than likly will get the customers without AMD 64 systems and thous with older 64 systems which the mobos will not work with the X2 or FX CPU's. The E6600 sounds good to me because all I have is a celoron 1.8 with a Nvidia 5500 GPU.

I think it'll intel is fair game for anyone with a 754 cpu or slower on the AMD side, and ppl with Pentium D8xx cpus and slower on the intel side. those are the ppl yearning for more performance
June 21, 2006 11:51:57 PM

Quote:
HEXUS can report that the AMD Athlon FX-64 will be available to system integrators in the UK at a price point of $999 per CPU from the 8th August 2006.


http://www.hexus.net/content/item.php?item=5990


I guess we'll find out if CTI will be worth it. SInce they are getting 35W procs with AM2 they may get the FX64 down to 89W. I'm sure they realize that they have to limit the increase in power over the FX62. Not for bragging rights but to keep total system power down.

That may be the selling point. Either way I'm confident that 939/940 was not a fluke. In this case I guess I'm just a fan of the underdog.
June 21, 2006 11:52:55 PM

Quote:
A switch in process technology given no real performance benefit AT THE SAME CLOCK SPEEDS. However, it reduces the die size of the chip, making it cheaper to produce, and it also allows for clock speeds to be raised, and that does increase performance. Also, with a good die shrink, one can fit more cache on the die and keep sizes roughly the same as before or even a little smaller.

I'm guessing AMD maybe got about 600 to 800Mhz extra over the 90nm process. I dont see AMD topping the X6800 by increasing to the higher Ghz but may could with a larger cache.

K7-250 nm 500-700mhz=200mhz
K75-180 nm 550-1000mhz=450mhz
Tbird C-180 nm 900-1400mhz=500mhz
Palimino-180nm 1333-1733mhz=400mhz
Tbred B-130nm 1400-2133mhz=733mhz
Newcastle-130 nm 1800-2400mhz=600mhz
San Diego-90 nm 2600-2800mhz=200mhz

You know if I didn't know any better it almost looks like the K8 doesn't scale well with the process being used. I give AMD 400mhz at the most for moving the K8 to 65nm pretty generous considering.
June 22, 2006 1:00:10 AM

"fan of the underdog"? Underdog=AMD. Fan of AMD. AMD fanboy?
June 22, 2006 1:02:26 AM

Intel wont scale as well ether. its because 130Nm to 90Nm was 40 NM difference, 90 to 65 is 25, etc. The absolute size is going down even if its always scales down 1/2.
June 22, 2006 1:10:58 AM

Quote:
A switch in process technology given no real performance benefit AT THE SAME CLOCK SPEEDS. However, it reduces the die size of the chip, making it cheaper to produce, and it also allows for clock speeds to be raised, and that does increase performance. Also, with a good die shrink, one can fit more cache on the die and keep sizes roughly the same as before or even a little smaller.

I'm guessing AMD maybe got about 600 to 800Mhz extra over the 90nm process. I dont see AMD topping the X6800 by increasing to the higher Ghz but may could with a larger cache.

K7-250 nm 500-700mhz=200mhz
K75-180 nm 550-1000mhz=450mhz
Tbird C-180 nm 900-1400mhz=500mhz
Palimino-180nm 1333-1733mhz=400mhz
Tbred B-130nm 1400-2133mhz=733mhz
Newcastle-130 nm 1800-2400mhz=600mhz
San Diego-90 nm 2600-2800mhz=200mhz

You know if I didn't know any better it almost looks like the K8 doesn't scale well with the process being used. I give AMD 400mhz at the most for moving the K8 to 65nm pretty generous considering.
I would have to see cache sizes and design changes with those else you can really compare them.
June 22, 2006 1:12:22 AM

Quote:
For the price differential between the 60 and the 6600 you should be able to afford a MB and Memory! And still have enough pieces to start another PC.

Your correct but having to give your current mobo and RAM makes the prices between the 2 get tight. Intel more than likly will get the customers without AMD 64 systems and thous with older 64 systems which the mobos will not work with the X2 or FX CPU's. The E6600 sounds good to me because all I have is a celoron 1.8 with a Nvidia 5500 GPU.

I think it'll intel is fair game for anyone with a 754 cpu or slower on the AMD side, and ppl with Pentium D8xx cpus and slower on the intel side. those are the ppl yearning for more performance
Yes the 754 falls in the no upgrade path to a X2 or FX as well. Thanks for pointing that out.
June 22, 2006 1:18:12 AM

Quote:
A switch in process technology given no real performance benefit AT THE SAME CLOCK SPEEDS. However, it reduces the die size of the chip, making it cheaper to produce, and it also allows for clock speeds to be raised, and that does increase performance. Also, with a good die shrink, one can fit more cache on the die and keep sizes roughly the same as before or even a little smaller.

I'm guessing AMD maybe got about 600 to 800Mhz extra over the 90nm process. I dont see AMD topping the X6800 by increasing to the higher Ghz but may could with a larger cache.

K7-250 nm 500-700mhz=200mhz
K75-180 nm 550-1000mhz=450mhz
Tbird C-180 nm 900-1400mhz=500mhz
Palimino-180nm 1333-1733mhz=400mhz
Tbred B-130nm 1400-2133mhz=733mhz
Newcastle-130 nm 1800-2400mhz=600mhz
San Diego-90 nm 2600-2800mhz=200mhz

You know if I didn't know any better it almost looks like the K8 doesn't scale well with the process being used. I give AMD 400mhz at the most for moving the K8 to 65nm pretty generous considering.


What you have to remember is that AMD always went for IPC and NOT clockspeed. When designing a process for efficiency, you aren't designing for headroom as much as immediate performance. You develop your envelope and optimize from there.

That might sound like a line but who cares. It's the truth. I'm sure that the 65nm will add functional improvements and not improvements for OC. Anand said AMD had a trick left up it's sleeve and it maybe this Inverse HT. Truthfully, I don't think any of it is "speculative," but.....

Does anyone know why the AM2 is so much bigger than the 939? Maybe because the IMC is the thread control and arbiter they talked about in the PDF for the patent. The technology described only requires under 64 extra bits and pathways in the IMC. AMD dual defines one IMC per die. My estimates of 20-30% are actually pessimistic because if any buffers are added to the IMC, they could contain a scratchpad for result verification and even provide support for simultaneous operation between single thread and multi thread apps.


SInce the diagrams in the patent material call for three additional "paths" to coordinate the start of threads and retiring of data and instructions.

Does anyone know why the initial IMC Revs were so bad? The patent was approved in 2003 and CnQ debuted with the Rev E 939(?). This microcode could actually determine both when to lower the clockspeed and when to raise it. If the clock throttles up for the second core it maybe possible to increase the clockspeed of each core while presenting a unified "core" to single thread apps and dual core to multithread apps. Multithread is a relative term in that "main UIs" can start as many threads as are necessary to accomplish SYNC and ASYNC execution.


The PDF talks about abstract technologies and clear paths to the process that CnQ, DDR2 IMC and the "larger" die size of REv F could actually perform. The Windows patch may actually just change time slice duration to a smaller value. It's hard to say whether they will need to lengthen the default 2-5 ms or shorten it. It varies between natice and .Net code.

I'm not sure abotu Linux but I'm sure that it coud replicate the slight change in "return timing" that the purported mechanism would allow - if the kernal refreshes it's own "cache faster and can use values that are returned faster fron the HW, the kernel will run faster. Apps don't talk to the CPU, they talk to the kernel, whether it's user mode, kernel mode ring0, ring3, root, etc.
June 22, 2006 1:33:23 AM

but Intel has that same "trick" of Inverse HT under way to.
June 22, 2006 1:36:23 AM

AMD fanboys pretend thats not true.
June 22, 2006 1:48:40 AM

Quote:
Intel wont scale as well ether. its because 130Nm to 90Nm was 40 NM difference, 90 to 65 is 25, etc. The absolute size is going down even if its always scales down 1/2.


K7-250 nm 500-700mhz=200mhz
K75-180 nm 550-1000mhz=450mhz
Tbird C-180 nm 900-1400mhz=500mhz
Palimino-180nm 1333-1733mhz=400mhz
Tbred B-130nm 1400-2133mhz=733mhz
Newcastle-130 nm 1800-2400mhz=600mhz
San Diego-90 nm 2600-2800mhz=200mhz
Manchester-90nm 2200-2400mhz=200mhz

(P6)Mondocino-250nm 300-533mhz=233mhz
(P6)Coppermine-180nm 533-1133mhz=600mhz
(P4)Willamette-180nm 1300-2000mhz=700mhz
(P6)Tulatin-130nm 1000-1400mhz=400mhz
(P4)Northwood-130nm 1600-3400mhz=1800mhz
(P6)Banias-130nm 1300-1700mhz=400mhz
(P4)Prescott-90nm 2800-3800mhz=1000mhz
(P6)Dothan-90nm 1600-2266mhz=666mhz
(P4)Smithfield-90nm 2800-3200mhz=400mhz
(Core)Yonah-65nm 1666-2166mhz=500mhz
(Core)Conroe-65nm 1833*-3333+=1500mhz

Hmm well have to see now won't we.

Moo.
June 22, 2006 1:49:41 AM

Quote:
A switch in process technology given no real performance benefit AT THE SAME CLOCK SPEEDS. However, it reduces the die size of the chip, making it cheaper to produce, and it also allows for clock speeds to be raised, and that does increase performance. Also, with a good die shrink, one can fit more cache on the die and keep sizes roughly the same as before or even a little smaller.

I'm guessing AMD maybe got about 600 to 800Mhz extra over the 90nm process. I dont see AMD topping the X6800 by increasing to the higher Ghz but may could with a larger cache.

K7-250 nm 500-700mhz=200mhz
K75-180 nm 550-1000mhz=450mhz
Tbird C-180 nm 900-1400mhz=500mhz
Palimino-180nm 1333-1733mhz=400mhz
Tbred B-130nm 1400-2133mhz=733mhz
Newcastle-130 nm 1800-2400mhz=600mhz
San Diego-90 nm 2600-2800mhz=200mhz

You know if I didn't know any better it almost looks like the K8 doesn't scale well with the process being used. I give AMD 400mhz at the most for moving the K8 to 65nm pretty generous considering.
I would have to see cache sizes and design changes with those else you can really compare them.

It's their uArch min to max clock speed nothing further to explain.
June 22, 2006 1:50:21 AM

Quote:
A switch in process technology given no real performance benefit AT THE SAME CLOCK SPEEDS. However, it reduces the die size of the chip, making it cheaper to produce, and it also allows for clock speeds to be raised, and that does increase performance. Also, with a good die shrink, one can fit more cache on the die and keep sizes roughly the same as before or even a little smaller.

I'm guessing AMD maybe got about 600 to 800Mhz extra over the 90nm process. I dont see AMD topping the X6800 by increasing to the higher Ghz but may could with a larger cache.

K7-250 nm 500-700mhz=200mhz
K75-180 nm 550-1000mhz=450mhz
Tbird C-180 nm 900-1400mhz=500mhz
Palimino-180nm 1333-1733mhz=400mhz
Tbred B-130nm 1400-2133mhz=733mhz
Newcastle-130 nm 1800-2400mhz=600mhz
San Diego-90 nm 2600-2800mhz=200mhz

You know if I didn't know any better it almost looks like the K8 doesn't scale well with the process being used. I give AMD 400mhz at the most for moving the K8 to 65nm pretty generous considering.


What you have to remember is that AMD always went for IPC and NOT clockspeed. When designing a process for efficiency, you aren't designing for headroom as much as immediate performance. You develop your envelope and optimize from there.

That might sound like a line but who cares. It's the truth. I'm sure that the 65nm will add functional improvements and not improvements for OC. Anand said AMD had a trick left up it's sleeve and it maybe this Inverse HT. Truthfully, I don't think any of it is "speculative," but.....

Does anyone know why the AM2 is so much bigger than the 939? Maybe because the IMC is the thread control and arbiter they talked about in the PDF for the patent. The technology described only requires under 64 extra bits and pathways in the IMC. AMD dual defines one IMC per die. My estimates of 20-30% are actually pessimistic because if any buffers are added to the IMC, they could contain a scratchpad for result verification and even provide support for simultaneous operation between single thread and multi thread apps.


SInce the diagrams in the patent material call for three additional "paths" to coordinate the start of threads and retiring of data and instructions.

Does anyone know why the initial IMC Revs were so bad? The patent was approved in 2003 and CnQ debuted with the Rev E 939(?). This microcode could actually determine both when to lower the clockspeed and when to raise it. If the clock throttles up for the second core it maybe possible to increase the clockspeed of each core while presenting a unified "core" to single thread apps and dual core to multithread apps. Multithread is a relative term in that "main UIs" can start as many threads as are necessary to accomplish SYNC and ASYNC execution.


The PDF talks about abstract technologies and clear paths to the process that CnQ, DDR2 IMC and the "larger" die size of REv F could actually perform. The Windows patch may actually just change time slice duration to a smaller value. It's hard to say whether they will need to lengthen the default 2-5 ms or shorten it. It varies between natice and .Net code.

I'm not sure abotu Linux but I'm sure that it coud replicate the slight change in "return timing" that the purported mechanism would allow - if the kernal refreshes it's own "cache faster and can use values that are returned faster fron the HW, the kernel will run faster. Apps don't talk to the CPU, they talk to the kernel, whether it's user mode, kernel mode ring0, ring3, root, etc.

Moo.
June 22, 2006 2:46:24 AM

Quote:
but Intel has that same "trick" of Inverse HT under way to.



I have no doubt that theItanium engineers aren't trickling down "processes" and that Intel is working on a similar tech but if the PDF lives up to it's tech, AMD may hand Core 2 its a$$. K8 is a formidable architecture and these improvements based on kernel response may make a really big difference in overall system perf.

If they can turn this on in JULY.........
June 22, 2006 4:16:46 AM

Quote:
but Intel has that same "trick" of Inverse HT under way to.



I have no doubt that theItanium engineers aren't trickling down "processes" and that Intel is working on a similar tech but if the PDF lives up to it's tech, AMD may hand Core 2 its a$$. K8 is a formidable architecture and these improvements based on kernel response may make a really big difference in overall system perf.

If they can turn this on in JULY.........

I think you need to admit the K8 is beat uArch wise, its time to move on. Intel did their homework on this one. AMD needs a new uArch, or well the outcome is something that would hurt you ego further I'm afraid.
June 22, 2006 4:43:21 AM

i dunno if thats fair, id take the maximum increase in clockspeed compared to the last die shrink, aka 190nm 2.0Ghz, 130nm 2.5ghz etc. o and i know that my example isnt accurate but its just an example.
June 22, 2006 4:46:34 AM

ur assuming its "in" the AM2 to be tunred on, also that Conroe doesnt have this, also assuming intel doesnt have something the same or better. I think we should all lay off the subject til we have something much more substancial to feed off of.
June 22, 2006 5:40:15 AM

Quote:
but Intel has that same "trick" of Inverse HT under way to.



I have no doubt that theItanium engineers aren't trickling down "processes" and that Intel is working on a similar tech but if the PDF lives up to it's tech, AMD may hand Core 2 its a$$. K8 is a formidable architecture and these improvements based on kernel response may make a really big difference in overall system perf.

If they can turn this on in JULY.........

I think you need to admit the K8 is beat uArch wise, its time to move on. Intel did their homework on this one. AMD needs a new uArch, or well the outcome is something that would hurt you ego further I'm afraid.
Word.
June 22, 2006 5:50:59 AM

so they're releasing a preshot? my my the tides have turned.

As for the 65 nm transition:
Considering they can both hit the same clockspeeds
- i still believe intel has more experience in that arena 8) -
intel's architecture is more efficient clock for clock.

So unless AMD redoes their architecture, like this reverse hyperthreading i keep hearing about,
you're going to see the same thing... 3.5 ghz athlon gets beaten by 3.33 conroe :lol: 
June 22, 2006 6:12:02 AM

Quote:
Are you serious, I hope you do know that switching from process technologies gives no tangible performance boost. AMD wont be winning any time soon no matter what you think. The FX 62 is maxed out, the overclockability of the chip is simply lame. The FX64 isnt going to be a marvel. Have you forgotten about the 3.2 ghz XE conroe ?, the lead will be even bigger, and intel has an overclocker on their hands. Theres alot more overhead. I personally say intel should max out the conroe, to its absolute limit, and pound amd while they are down, thats competition not releasing products which they could on launch but dont. They will have a great lead, make that lead bigger, and keep hitting hard.


I do not think they will simply knock AMD into the ground and keep kicking them. If they were to accidentaly put AMD out of business they would be doing nothing short of screwing their own company over, for with no competition, the market grows stale. Without AMD, today Intel would be nothing as we know it. After all, that's why Intel created AMD.
June 22, 2006 11:26:28 AM

Cool, but...

WHO THE HECK CAREZ?
!