Synthetickiller

Distinguished
Aug 25, 2006
340
0
18,780
This has probably been discussed, so if it has, just link me so you guys don't waste your time.

Over the past couple years, dual core has become really popular. Would having 2 separate cores with 2 FSB be faster (as long as you're comparing the same type of cores and same amount of possible usage cache and ram)?

It seems more logical to me to have dedicated ram for each core vs a shared L2 cache as well as a FSB dedicated to each core. I doubt there would be progamming issues (unlike the sega saturn. If anyone knows anything about the 2 RISC CPUs it uses, u know what I am talking about) associated with it in terms of software development.

Does this seem logical or am I totally off base? :lol:
 

battousai831

Distinguished
May 30, 2006
251
0
18,780
think of it this way tho, unless you need that much ram, ie over 4gb then its not worth it. this would be because if 1 cpu needs something thats in the other cpu's ram it has to travel over the bus and same for l2 cache and so on (or htt link). so yes there are some cases that it is better and there are some when it is worse, it all depends on what you are doing.
 

ElMoIsEviL

Distinguished
This has probably been discussed, so if it has, just link me so you guys don't waste your time.

Over the past couple years, dual core has become really popular. Would having 2 separate cores with 2 FSB be faster (as long as you're comparing the same type of cores and same amount of possible usage cache and ram)?

It seems more logical to me to have dedicated ram for each core vs a shared L2 cache as well as a FSB dedicated to each core. I doubt there would be progamming issues (unlike the sega saturn. If anyone knows anything about the 2 RISC CPUs it uses, u know what I am talking about) associated with it in terms of software development.

Does this seem logical or am I totally off base? :lol:

Interestingly you have applied the AMD approach using Intel terminology. AMD sinlge core Opterons are linked up via a serial hyper transport link which is faster than the FSB approach by intel, so Opterons scaled better in multisocket platforms.

Dual core CPUs starting coming on the scene, AMD put two on the same die and Intel put two die in the same package (i.e. socket). AMD's solution was sexier, but the relative performance boost dual vs single was roughly the same---AMD's X2 (dual core) were simply better than any P4 dual core just by fact of architecture.

Technically speaking --- two core on the same die are going to 'cooperate' better than two separate CPUs in different sockets, simply due to signal propogation and associated latency of having components spread out spatially by large distances.

Jack

Exactly.. and there are tests out there I believe THG has one that proves this. I have to argue this all the time.. especially at 2CPU.com. Lot's of thick headed peeps out there.
 

Synthetickiller

Distinguished
Aug 25, 2006
340
0
18,780
This has probably been discussed, so if it has, just link me so you guys don't waste your time.

Over the past couple years, dual core has become really popular. Would having 2 separate cores with 2 FSB be faster (as long as you're comparing the same type of cores and same amount of possible usage cache and ram)?

It seems more logical to me to have dedicated ram for each core vs a shared L2 cache as well as a FSB dedicated to each core. I doubt there would be progamming issues (unlike the sega saturn. If anyone knows anything about the 2 RISC CPUs it uses, u know what I am talking about) associated with it in terms of software development.

Does this seem logical or am I totally off base? :lol:

Interestingly you have applied the AMD approach using Intel terminology. AMD sinlge core Opterons are linked up via a serial hyper transport link which is faster than the FSB approach by intel, so Opterons scaled better in multisocket platforms.

Dual core CPUs starting coming on the scene, AMD put two on the same die and Intel put two die in the same package (i.e. socket). AMD's solution was sexier, but the relative performance boost dual vs single was roughly the same---AMD's X2 (dual core) were simply better than any P4 dual core just by fact of architecture.

Technically speaking --- two core on the same die are going to 'cooperate' better than two separate CPUs in different sockets, simply due to signal propogation and associated latency of having components spread out spatially by large distances.

Jack


The hyper transport link has been something I've been doing a little reading into ( at least in terms of MOBO comparisons) since I just purchased a replacement mobo for my 2 opty troy 250s @ 2.4 ghz. I've never had a problem with speeds with dual cpu (not dual core) CPUs, but I can definitally see why a dual-core or quad-core (if they can make applications use the tech) would be supperior in many applications.

It'd be curious to see a dual socket Conroe setup (although I seriously doubt this will happen) VS a duo core QUAD set up.



I do find it funny though how people brag about CPU speeds and power when they can't even be realized. I just ordered an x1900xt and although its not the xtx, I doubt I'll see a difference (not planning to overclock or open up pixel pipes till more people have done it.). Hardware is always ahead of software, its just too bad people don't realize this.

Don't get me wrong, a lot of you guys on here use these new conroes for more than just gaming. I don't understand people my age (college students) running systems that cost $3,000+ to play half life 2 or quake 4.

I remember getting a new althon 550 SLOT A ( WAY BACK WHEN), and having it fly. At least back then, I could see a difference in power. I think for average computer use, hardware has jumped forward 8 or 9 steps ahead of software ( at least until the whore known as vista ultimate rears its ugly head and sucks our souls dry :lol: ).

Thanks for clearing up some of that technical stuff. I haven't really been messing with computer hardware for 3 years b/c of college. Now getting back into the thick of things, I see just how far things have come. Makes me glad i'm waiting till the next AMD vs INTEL war gets overwith so I can purchase a CPU based on raw data, no hype.
 

TRENDING THREADS