• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

90nm Athlon 64 Socket 939 Winchesters are here!

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.
OC Detective, I just have a hard time believing two 800mhz processors based on the same architecture and with no other benefits, would beat a 2.2ghz-2.8ghz chip in anything. It doesn't work in dual processor systems, I just don't see it working in dual core either.

If it isn't 1.6ghz a chip, they must have pulled off an amazing architecture change.
 
the problem i see with dual cores is that they will be sharing memory bandwidth, and we all know how that kills performance........ but then dual channel doesnt help a single a64 that much so maybe it wont kill performance?

also i doubt a 1.6ghz dual core will be 2x800mhz, Amd and Intel know this not to be the case. surely?
 
man_utd said:
OC Detective, I just have a hard time believing two 800mhz processors based on the same architecture and with no other benefits, would beat a 2.2ghz-2.8ghz chip in anything. It doesn't work in dual processor systems, I just don't see it working in dual core either.

If it isn't 1.6ghz a chip, they must have pulled off an amazing architecture change.
Not me that is saying it - the links provide a quote from AMD's manager of Opteron architecture.
Here is another link saying the same thing. Actually in this one they are talking about 1.4Ghz total for the dual core. Remember these are primarily designed for servers with multiple cpus and the comparison would be done in this arena.
http://www.pcworld.com/news/article/0,aid,118078,00.asp
Quote
"The size of the performance increase going to the new processor depends on the speed of the two chips. Comparing a computer today with one that uses two chip sockets with 90-nanometer processors, performance increases between 30 percent and 55 percent on various benchmarks, McGrath said. The dual-core models run 600MHz to 1GHz slower than the single-core models to prevent overheating problems, he said."
Seems fairly clear when they say model they are talking about the whole thing rather than each processing engine.
 
Last edited:
How will they boost the IPC 50% to make up for crapy clocks of dual core? If thats the case, why not make a 2GHz single core perform like 3GHz
 
OK once again the comparison is likely a dual core cpu v 2 single core cpus (in a server).
Once again the quote
"Comparing a computer today with one that uses two chip sockets with 90-nanometer processors, performance increases between 30 percent and 55 percent on various benchmarks"
The "computer today" being the dual core being demonstrated! Everyone seems to forget we are talking servers primarily.

Further quote:
"A two-processor dual-core system clocked 5 clock speeds down from AMD's fastest part will generate performance of approximately 125 to 140 percent performance of the dual-processor unicore system, McGrath said. A dual-core, dual-processor system clocked at only three grades slower than the fastest AMD chip will perform at between 130 to 160 percent of the performance of the base system, McGrath's slide indicated. McGrath's performance figures used synthetic benchmarks, such as SPECint_rate 2000 and SPECifp_rate 2000. "

So we are not comparing a dual core against a single core cpu (i.e dual core v single cpu desktop) but dual core dual processor v dual processor unicore!
 
For the past few years I've been periodically asking how future processors will effect video rendering, with very few people ever posting opinions.

Keeping camcorder recordings on magnetic VHS tapes is simply not a good idea given that inexpensive DVD media and DVD burners are here. I see video rendering as the primary reason for upgrading from Athlon XP,

so... who here can post

how much of an improvement do you think a 2.75 GHz Socket 939 Athlon 64s would give you over a, say 2.4 GHz Athlon XP, if other parts of the system are similar?

How about future dual core Toledo over Athlon XP?
 
:) I was saying primary reason to upgrade (if there ever is one) since I don't do too much rendering, it's just that I can't justifying upgrading for any other reason since games/apps run just fine for my needs without upgrading... I'm trying to (so far unsuccessfully) find out just how much of an improvement will future CPUs give me in this particular area.

Overall cost to performance to overclockability looked to be greatly in AMD's favor again with the release of relatively inexpensive Winchesters,

or is it?...:

:eek: Oct 11, 2004 AMD's 90nm Athlon 64 Have Equal Performances to 130nm Chips :eek:

Inq's called X-Bit "those crazy Russians..." once or twice.
:) :)
 
OC Detective said:
The 2GHz for BOTH (ie total) is roughly correct for those at launch. Indeed it may well be optimistic some are saying they will launch at 1.6GHz. (clockspeeds expected to be 600Mhz to 1Ghz lower than at present in TOTAL).
The shortfall is then made up by performance improvements over single core of between 30 to 55%.
http://www.tomshardware.com/hardnews/20041005_205914.html
http://www.extremetech.com/article2/0,1558,1666805,00.asp
http://www.x86-secret.com/?option=newsd&nid=767 (for the french speakers!)
The notion of 1.4 to 1.6GHz EACH cpu is highly unlikely given the near doubling of transistors and only a 95W TDP!

I read that it will be 25%-40% faster that is why I expected such speeds. I read the articles you linked, interesting that one writes these will be 939 the other 940 ones.
When they compared performance they said 2 dual cpu systems, which would mean it is server performance not desktop. So we should compare to Opteron speeds not 939.
Even so it would be hard to belive that 4*1 Ghz cpu with 2 memcontroller can be faster than 2*2,4Ghz cpu with 2 memcontroller by 25%-40%. Doubling the cache would not make up for the 800Mhz in clockspeed not mentioning they said it will be faster. A64's do not benefit much from the cache on desktop, I am not sure about server performance, but the systems scale up well with higher clockspeed while cache might not give a boost at all.

About the TDP I did not read about 90nm A64's underclocked but PM was working without heatsink at 1Ghz with only 0.8V even 2 such cores would not make much heat around 1.4-1.6Ghz. I have no idea how much voltage will AMD need for these chips but IMO it can be the need four double memcontroller what keeps them raising from clockspeed in the future and not heat.
 
Considering 3500+ A64 90nm has TDP of only 67W then there won't be a problem to make 2 GHz Dual core with TDP 95W. Dual core is two 2GHz cores minus 1 mem controller (about 10W of power) minus 3 HT links - you do not need to duplicate those.

Intel won't launch desktop dual core on Pentium M architecture. Only mobile dual cores will be Pentium M, desktops continue to use NetBurst... yes, those 151W / 119Amps Prescotts will be doubled - I guess they will have to lower frequency to maybe 2.6 - 3.0 GHz what will seriously hurt performance and be no competitive to 2 GHz AMD dual core (... and not even to 2 GHz dual core Yonah Pentium M).
 
Petr said:
Intel won't launch desktop dual core on Pentium M architecture. Only mobile dual cores will be Pentium M, desktops continue to use NetBurst... yes, those 151W / 119Amps Prescotts will be doubled - I guess they will have to lower frequency to maybe 2.6 - 3.0 GHz what will seriously hurt performance and be no competitive to 2 GHz AMD dual core (... and not even to 2 GHz dual core Yonah Pentium M).

I know Intel won't launch such desktop unfortunately :confused:, what I tried to show that lower speed requires less voltage and TDP is not an issue with such CPUs.
 
enduro said:
Why's that? I wasn't really looking at the FX anyway. I was looking at the overclocked results for the 3200+

Look at it.
These results are illogical. It doesn't make sense that the 1MB- cached FX is slower that the 256KB Sempron
 
Back