AMD Speeds...how?

[klein1jo]

So I was just checking out a parts website. I've always been an Intel man, but after looking I am kinda left astonished. The Intel CPU's max at 3.8 Ghz, but 3.6 and 3.4 are more common with 800Mhz FSB. AMD is hitting 4800+ with averages around 4000+ and FSB's at 2000Mhz. When did this happen and how are they getting bus-speeds like that?

I've heard that Intel hit a brick wall. They were just reducing microns to get faster speeds and now they've got to the point that when they do that, their chips aren't speeding up anymore so they have to redesign their architecture. I also heard that x86 may be going byebye.

Maybe AMD because of the different architecture hasn't hit this snag yet (they are getting 4000+ speeds at around 2.6Ghz), but will when they get their chips running over 3Ghz. But how are they getting busspeeds like that and does that mean that memory is holding them back from some serious speeds (still running at less that 1000Mhz)?

 

[Trifid]

Starting from the top

The AMD 64's use different archecture, so are more efficient, and that number is the average equivilant of 'the main competitor' would need to have to compare to it. The AMD's actual processor speed is not that high. The FX57 (The top gaming proccessor) is 2.8ghz, but due to having shorter pipelines, it can do more calculations per cycle , it can do alot more with that 2.8ghz.

The 2ghz is the hypertransport speed, with a bus speed of 200mhz. AMD doesn't use a FSB as it is not efficient, so instead have a onboard memory controller, and a link to the memory at 2000mhz.

They are both nearing that brick wall, where they can't go any higher (on the 90nm archetecture.) So what they are doing is having more cores on the chip, and let the programmers make use of those cores, to work on the same program together.

With AMD you have to ignore that pure clock speed, as it isn't a representation of what it is capable of. Even comparing a 2ghz socket 939 to a 3ghz socket 754 isn't a accurate representation. So you have to look at benchtests.

When the memory is running slower than the bus speed, that is when the bottlenecks happen (under 200mhz.) AMD will move onto DDR2 at 667mhz for their next socket, the M2. Also AMD make full use of dual channel ram, with PC3200 ram, would nomally transfer data at 3200mb/s, with dual channel it runs at 6400mb/s. Intel systems don't get bandwidth as high as that with dual channel.


Btw. there is a search button.

 

[Ste]

Basically what he said. Good explanation.

I actualy heard that AMD will skip right to DDR3 with their new socket.

 

[klein1jo]

That was awesome. So to clarify in my head, the system bus is the speed of the motherboard. The FSB is the speed that the CPU talks to the memory. AMD got rid of that cause using onboard they can get speeds up to 2000MHz. So the bottleneck here is the RAM.

Intel, if running at 800Mhz FSB, can talk to the memory at that speed. So if you are using 333Mhz you haven't hit the bottleneck until you get 800Mhz ram.

 

[klein1jo]

Oh, so how do video cards work in all this. if the AGP slot connects directly to the system bus, then depite having core speeds up to 1000Mhz, it can only talk at 200Mhz.

I don't know how PCI-E works or where it connects. I'm sure this question will prob make me look like an idiot.

 

[beardo526]

I'm not so good at this stuff, but I know that the different peripheral interfaces (PCI, AGP, PCI-E) all communicate with the chipset of the motherboard, then the chipset talks to the CPU, RAM, etc. I don't know if that's helpful to you or not, but at least it's a bit of information you might not have known already.

 

[Zanjo]

The PCIe x 16 (Graphics card slot) slot is above the PCI slots or PCIe x 1/4 slots. It usually looks like 3 white slots then 2 small blacks slots followed by a white one. The one that is higher (when the MoBo is mounted to the case) is where the PCIe is. Or you can just see which fits?

 

[Static_11]

very well explained... that goes for everyone