#30, cool. When's Part 2? :)
:
DrDillyBar
(#31)
«
| Edit Reply |
|
moritzgedig |
speculative execution and branch-prediction have such an impact!
not only on the execution but also on the powerdraw. as AMD is unable to do it as good as intel. They should find a way around it. hyperthreading and more not so speculative cores should be looked into. now that the singlecore performance has reached it's limit (until there is some major tech change) the code will have to be parallel anyways, thus one can change to another thread, instead of using resources on chance. likely the number of cores per CPU will double every 3 years, that would be 8 cores in every PC in 2014. Until then, programmers will have to have found, a way of easyly making multithreaded code. It can't be that hard, humanity has achived harder things befor.  |
#8, Good advice!
:
moshpit
(#17)
«
| Edit Reply |
|
Vaughn |
it shouldn't feel like a dinosaur, im still running a Opteron 170 tho with 4GB of ram just upgraded from 2GB. Vista runs great on my rig and I haven't really found anything that is forcing me to upgrade yet.
|
| Edit Reply |
|
Homerr |
I use an E8400 system at work and have an E6850 at home. I had to work on an older s939 4200+ X2 system yesterday and was pleasantly surprised that it didn't feel like a dinosaur.
|
| Edit Reply |
|
StuG |
I know alot of people have always said the X2 architecture as less efficent than a C2D, but i'm glad to see someone just break it down into simliar terms. I still <3 my X2, put that thing through hell and it still is running great after many years of my stupid abuse.
|
| Edit Reply |
|
A_Pickle |
Branch prediction tests showed "vastly" greater accuracy on the Intel front, with "about 50% fewer mispredicted branches per instruction retired."
This should be the unsurprising part. Ah, NetBurst, you brought one great thing to the table for us. :) |
| Edit Reply |
|
ybf |
CPU + game = inadequate use of resources.
I keep my nVidia 8800 GT warm running the CUDA version of the Folding At Home client, and use the other 95% of my Yorkfield cycles on a stack of BOINC grid projects. Interestingly, I can do this 24/7, and have found only one application that has any problem running at "normal" interaction rates while all this is going on in the background, and its only problem is initial startup (it wants me to pause BOINC or it takes 5 minutes instead of 5 seconds; probably a stupid choice of default task priority buried in some component of the .NET library the slow app is based on). Microsoft may have screwed up a lot of fiddly things in Vista, but the scheduler was not one of them. I'd love to spend 8 months inside Redmond pushing the knobs until it's got a true real-time mode... Oh, and it runs Allied Assault: Airborne at full screen-res and full fidelity with zero visible inadequacies, but for that I kill the charityware. |
| Edit Reply |
|
Hattig |
Good article. Will be interesting to see how Phenom compares to the older K8 core design as used however.
|
| Edit Reply |
|
tfp |
It's nice that people still write in depth articles. I always enjoy reading Real World Techs write-ups.
|
| Edit Reply |
|
DancinJack |
Rather than delve into the obscure architectural details of the two offerings, RWY used apps called VTune and CodeAnalyst to poke around under the hood and get hard numbers for things like instructions per clock, branch predictor accuracy, and how the chips handle their L1 and L2 caches.
Should that be RWT? or am I mistaken? EDIT: you fixed it Cy. :) |
2006 called. It wants its review back.