can anybody tell me what's the difference between a dual dual-core (fake quad core) and a real quad core?
Really, nothing as far as you are concerned. I think even Intel would admit that their current solution is a stop-gap until they integrate everything on one die. But this situation is very similar to the one we were in way back when there was the option on having cache memory on die vs in a separate silicon die but wire bonded on to the same substrate vs cahe on the motherboard.
(Back then there
was a speed difference because you were taking the fastest bus on the chip and leading it across the motherboard, when otherwise it could stay on chip. In this situation, while these are fast interconnects, they are fast interconnects that had to go out anyway, and the inter-cpu traffic isn't all that much in most practical circumstances.)
Back then, there was discussion of which approach was best, but, with the benefit of hindsight it is clear that it was best to regard this a purely an internal issue. What you got and what you paid for it are the things, and the marketing nonsense of 'we've got a better arch, nah, nah, na, nah, na' isn't in the short term as important as who wants your business more, and so who offers a discount to get your business, because that often counterbalances internal production cost issues.
In the longer term, you have to look at the roadmap/plans for the platform, but you were going to do that anyway, right?
So, if you are trying to decide which to buy, you should compare what you get for what amount of money (performance, compatability with stuff you've got already, like motherboards and RAM, and the other stuff like 'future proofing'). You shouldn't 'add extra points' for one approach or the other, the production costs are already baked into the price you get charged, and you don't need to add it in again.
from verndewd
well we cannot derive much from comparison of AMD and Intel. Honestly I dont know if it matters either way. You may think that 4 cores on one die will communicate faster between cores ,But if the architecture is not as efficient it doesnt matter if its native quad or 2 dies.
The trouble is that a benchmark compares
everything in the system and given that we know that Intel and AMD give different results, you know that'll you get different results even if the 'real/false' causes no difference. But, as I argue above, just look at, say, benchies for the whole system and the cost of the whole system and ignore how its done. The biggest bang per buck is the biggest bang per buck. End of story.
Seperate dies can induce latencies with longer pathways unless its designed around that and through put doesnt rely on cross die communication which if not done right can also cause latencies.
While that's also correct, part of that is at the level of how the OS/Application ditributes the workload around and whether it ever has to move stuff around between cores. So, the CPU manufacturer can't really control that.
Here, the AMD arch, with its Hypertransport bus is 'better'. In essence, HT is probably going to work as well as it works (unless the HT clock changes) whether it is working between chips on one slice of silicon, two on one module, or four spread around a motherboard. OTOH, there is little evidence that the Intel approach of using the bus to memory to communicate between cores has a problem
at today's CPU clock speeds and memory speeds. Sometime, if CPU speeds and numbers keep going up and memory bus speeds don't go up at least as fast, it will become a bottleneck, but it hasn't happened yet.
But, you can regard the AMD approach of being more elegant and having better ability to scale, but that really isn't worth much until the Intel approach shows limitations.
from brad81987
umm, i'm only a first year EE so most of that went right over my head...but i think AMD is shooting for something that doesn't have to use the FSB to communicate.
Yes, well that's true, But Intel are using the FSB because that's what Intel always does (until some future arch comes along) and AMD are using the HT bus, because that's what AMD always does. So, the comment is correct as it stands, but it just doesn't come from 'true' or 'false' dual cores, just the difference in approach that the two organisations have in their architectures at this point.