damn right! I'm staying behind Conroe / read too much benchmarks Did any1 thought that this "new" amd tech is used by Conroe right now?
hmm, so much speculation. until conroe is out, and i see for myself whether it owns amd, isnt that good, or whether amd have a 'secret weapon', im trying not to get too excited. valid point about how 'reverse hyperthreading' will handle multiple single threaded apps or a multi and a single etc, should be interesting. i doubt it will be released under 'reverse hyperthreading'. doubtless, hundreds of people have already been dedicated to finding a confusing acronym for it that contains at least one 'X'. roll on the end of july ...
I'd rather reserve judgement on what's going to be better until I see the benchmarks from both sides when it's released. I suspect that Conroe won't be as fast as the PI benchmarks indicate, but it's certainly going to be very competitive. We'll just have to wait and see...
This is interesting at the very least. This reverse HT was pointed out awhile back, and could well work as a way of transforming the processor from a dual processor to a single wide order processor - not by emulation, but by literally altering the way information goes through it. The die on the AM2 CPUs is 20% larger than that of Socket 939 - that cant all come from making the memory controller DDR2 based IMO.
O Rly? i didnt know about this? Maybe they sliped this in without anyone notecing or asking questions.
Oh God.... what to believe? I'm just gonna sit this one out, and wait until we have a definitiev answer. I'm not upgrading again until all this is resolved once and for all. I agree with Josh: "Reverse Hyperthreading" is the most pants name for a technology ever.
Looks interesting, lets hope its succesfull. The worrying thing for AMD is that surely intell could copy this quite soon? I think i'll reserve judgement untill we see some independant benchmarks from both contenders
Like a few others I'm reserving my judgement untill a few solid benchmarks are out. Would be halarious if this actually made them better/as good as intel's core architecture. Intel: Hey look we completely redesigned out architecture to make these amazingly fast CPUs. Oh suck on that AMD! *AMD flicks switch* AMD: You were saying? Intel: *Few months later* AMD: Heres our new architecture Intel: I would like AMD to win but if intel hips are actually better then I'll go with them.
Interesting idea. I hope it does prove to work (and work well), as I've just purchased an AM2 4000+ X2 whilst I could.
So that implies that what AMD are gonna do with a driver, Intel will do in hardware yum. Also: http://www.intel.com/technology/magazine/research/speculative-threading-1205.htm ch424
That diagram does look very interesting. It implies that the Conroe can do the same as reverse HT, especially the "dynamic switching operation" note. On when it helps, off when you need as many threads as possible. Core2 seems full of suprises doesnt it! What else is in there?
But if I'm not mistaken, crossfire and SLI use two GPUs that are designed to stand work alone, I can use an SLI enabeled GPU in my system even though I don't have SLI. With the GPUs you have to have a system for the link to be made between the two, and I think they are not designed to work togother as well as two identical cores on one piece of silicon. L J
You're basically saying that SLI is just two chips plus interfacing hardware to link them and split up the data. Surely AMD are doing the same thing? Two identical cores, just with extra hardware to split up the data and link the two cores, hence 20% increased die size. Although I see what you're getting at, if it were truly in hardware, it wouldn't need drivers. The fact that it need drivers implies that it works like SLI/xfire, not a proper hardware solution. ch424
It hs been noted that dual core cpu's dont help gameing very much because games arent writtin for dual core. Although when you have dual graphics cards you get a fairly good boost in FPS even if the game isnt writtin for two gpus. Im assuming this patch will be abot the same as adding that second card, effectively useing both cores.
At the very end of it all, we can all just wait for launch days and see the benchmark results first hand. All we can say for sure is that both sides are fighting hard - and that AMD will be trying to really counter with K8L. Of course, the ball game starts all over again with Nehalem...
The reality is thred level parallelism and thread sync issues are not so clear cut. If it was, we would already have optimising compilers that could generate decent multithreaded code for us. That is in essance what this reverse HT tech sounds like, only it's done at run time instead of build time. I'm rather skeptical. Otherwise, it is simply more instruction level parallelism, and there is hords of info on the limits of that. Considering big blue only just started sinking money into compiler theory to generate multithreaded code, I don't know how far AMD could have gotten. I guess I'll just have to wait and see.
