• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

anyone interested in 65 nm A64 x2

Permabanned
Joined
31 Oct 2006
Posts
59
AMD debuing 65 nm A64 x2 chips in limited supply. They only have 2x512 KB L2 caches as oppossed to the current 90 nm ones that sport 2x1MB L2 caches. They run at 65 Watt so maybe there is some overclocking head room in them? It does not seem much like progress to me, I guess we really do have to await a new architecture in mid 2007 for that much needed performance increase to match Intel.
 
cache makes little difference in real world applications, there should be a fair bit overclocking in these chips as well, since 90nm were decent clockers, 65nm will naturally generate less heat so, 3.0+Ghz should be very achievable, optamistic for 3.4 or so myself or maybe even higher under watercooling when i set it up :) i need some answers on when were getting them here but so far nobody has said a thing :(
 
councilofdespai said:
AMD debuing 65 nm A64 x2 chips in limited supply. They only have 2x512 KB L2 caches as oppossed to the current 90 nm ones that sport 2x1MB L2 caches. They run at 65 Watt so maybe there is some overclocking head room in them? It does not seem much like progress to me, I guess we really do have to await a new architecture in mid 2007 for that much needed performance increase to match Intel.

Only 2 90nm parts every came with 1mb L2 cache. most came with 512kb and the 1mb parts have been discontinued for a while.
As Gashman said, cache has little effect on A64 performance due to the architecture.
 
Kamakazie! said:
Only 2 90nm parts every came with 1mb L2 cache. most came with 512kb and the 1mb parts have been discontinued for a while.
As Gashman said, cache has little effect on A64 performance due to the architecture.

if anything RAM latency has a much bigger impact than cache, thats why AM2 seems to do so badly against C2D, we need some CAS2.0 DDR2 running at 1000Mhz (DDR) :D and these 65nm parts clocking at 3.6Ghz to compete with intel
 
Kamakazie! said:
Only 2 90nm parts every came with 1mb L2 cache. most came with 512kb and the 1mb parts have been discontinued for a while.

Not entirely true, there's the x2 4400 & x2 4800 and also the AM2 x2 5200 & FX62.
 
the xs4400 and 4800 have been stopped but theres just a lot of stock that you can still get one the other 2 thier high end models so they have 1mb cache
 
Gashman said:
if anything RAM latency has a much bigger impact than cache, thats why AM2 seems to do so badly against C2D, we need some CAS2.0 DDR2 running at 1000Mhz (DDR) and these 65nm parts clocking at 3.6Ghz to compete with intel
The internal memory controller ensures the A64s still have lower memory latency than C2D, however C2D has very good pre fetching algorithms which hides the sluggishness of the FSB some what. Faster RAM also makes little difference with the A54, it is far from bandwidth starved.
K8L will feature better pre fetching and prediction branches as well as a faster HT so that will reduce latency further and extend the lead over the C2D in memory intensive tasks.
I reckon the 65nm parts will struggle to clock that high. Most are topping out at 3gig max at the moment and AMD die shrinks usually don't yield much better clocks, just less power consumption and less heat output.


El Jimben said:
Not entirely true, there's the x2 4400 & x2 4800 and also the AM2 x2 5200 & FX62.
4400 and 4800 have been discontinued for a while and never came to AM2 i don't think.
5200 and FX62 are essentially Opterons and are AM2 only
But yeah, ok, 2 for each socket :)
 
Kamakazie! said:
The internal memory controller ensures the A64s still have lower memory latency than C2D, however C2D has very good pre fetching algorithms which hides the sluggishness of the FSB some what. Faster RAM also makes little difference with the A54, it is far from bandwidth starved.
K8L will feature better pre fetching and prediction branches as well as a faster HT so that will reduce latency further and extend the lead over the C2D in memory intensive tasks.
I reckon the 65nm parts will struggle to clock that high. Most are topping out at 3gig max at the moment and AMD die shrinks usually don't yield much better clocks, just less power consumption and less heat output.



4400 and 4800 have been discontinued for a while and never came to AM2 i don't think.
5200 and FX62 are essentially Opterons and are AM2 only
But yeah, ok, 2 for each socket :)

aren't i right in thinking AMD found a pre-fetch problem with K8, thats gonna be fixed in K8L? and why the hell has the only person i want to answer (gibbo) not posted :( still curious when OCUK are getting these
 
bris01vt5.jpg

got this from extremesystems forum

not bad if true!, on stock volts too :D
 
dante6491 said:
If that was on air, imagine what it would be like under water :eek:

No difference, look at the voltage, he's not limited by heat, just the limit of the chip.

Dont see why he's used a .5 multi though
 
Defcon5 said:
No difference, look at the voltage, he's not limited by heat, just the limit of the chip.

Dont see why he's used a .5 multi though

apparently the Brisbane chips feature a half multi to get around the problem with DDR2 running at proper speeds.
Is that correct?

Meh Photoshop.... that sucks. If they can regularly hit 3.5gig though then it bodes extremely well for K8L.
 
Back
Top Bottom