amd fightback 2007??

[drunkenmaster]

Thats what I am saying - I cant believe any S939 chips where even fabricated prior to S754 release, just seems HIGHLY unlikekly to say the least (or is he part of AMD's high inner core?? Even then I find it dubious)

es chips find their way out. 939, in terms of cpu releases was extremely, extremely close to 754 launch. they aren't vastly different, you gotta realise that a 939 is 1 pin and one on die mem controller difference, ecc mem thats it. there would have been batches available way before 754's official launch(available to mobo makers, mem, and amd people i mean). september till around, well their were reviews based on es samples around april/june, they would have had them for a few weeks to play with so they got chips at the latest 5 months after the release of 754. there is no way in hell they wouldn't have yet made 939 chips by then, started making them and testing chipsets, motherboards, etc, and get mobo's and everything else ready by the launch date. things just don't happen that quickly.


i'd actually forgotten it was all the way back in 2003 that the 754 came out, (two hazy uni years i guess helped to pass the time), did it launch with a 2.2Ghz part? 2.4 maybe for the 940fx, my early 3400+ would hit what, 2.6Ghz, my first newcastle on 939(3500+) hit 2.75Ghz on air. 0.09 process bought us very very very small increases to clockspeeds which is kinda sucky. i think at the moment we've seen almost 3 years go past with very little stock and air overclock increase in speeds. it seems like more of a design limit than a process limit tbh, i'm very unconvinced a 0.065 process will bring us big increases at all. pretty convinced i'll be on conroe's till at least early next year, haven't had an intel system since before uni 2 years ago.


anyway, just though of a point to add, the main reason es and early review cpu's are often not overclocked as far as possible compared to other components. unlike asus, gigabyte, dfi, msi, g-skill, ocz, xfx and many others who all build products pre-overclocked, for certain higher than spec fsb speeds or boards that have multiply voltage/fsb options for overclocking, amd and intel have always firmly stood in the, "these our stock products, you shouldn't really overclock them" stance.

 

[Angilion]

I recently read rumours of a stopgap measure that's basically a K8 dual core with additional execution units, to bring it closer to performance/clock speed ratio of Conroe.

Conroe would still beat it, partly because the cache handling on dual-core is far better with Conroe than K8 and partly because Conroe should be able to scale up to much higher clock speeds, but it would reduce the gap a lot.

Does this sound like a plausible reworking, or would it entail major design work and take nearly as long as a bona fide new core?

 

[Kamakazie!]

Time and again people forget that K8L is a high end server/workstation part. It won't see the light of day on desktops.

as far as i am aware, a desktop dual core (and quad core) are planned using the K8L architecture. It is certainly not a server/workstation only revision.
not, atleast from what i have gathered from various articles on the web.

 

[drunkenmaster]

I recently read rumours of a stopgap measure that's basically a K8 dual core with additional execution units, to bring it closer to performance/clock speed ratio of Conroe.

Conroe would still beat it, partly because the cache handling on dual-core is far better with Conroe than K8 and partly because Conroe should be able to scale up to much higher clock speeds, but it would reduce the gap a lot.

Does this sound like a plausible reworking, or would it entail major design work and take nearly as long as a bona fide new core?

this is essentially what is happening with the K8L, its not gonna be out till probo q1 2007, but thats not that far off. its beefing up most parts of the core though so its IPC is increasing, noticably, but if its enough is yet to be seen. its quad core part should in theory, be a lot more efficient than the conroe quad core(kentsfield) due to bandwidth available when all 4 cores are working the fsb hard. however, if its just an added two cores that can't be used for anything it won't matter a bit to most of us. though if they can push the ppu type stuff onto 3rd/4th core then you can get along fine without a ppu card.

theres a rumoured(and very very likely) dual socket stop gap measure that will be released/talked about shortly after conroe launches(probo early august for dual socket thing). but the psu, boards, memory(maybe) and two cpu's will be expensive. but will be in the "i've got the biggest wil....." type catagory, which means a lot to joe schmoe public. if amd can, despite cost, say we ARE the fastest, then someone looking to spend much less is going to be under the impression that anything amd have is faster. its a marketing win and something most of us won't be able to afford, or utilise, but when's that ever stopped anyone