Can someone briefly explain to me....

[Sulaco]

Hey guys,

I'm looking to upgrade in the near future and have been looking at CPUs. I've had my eye on the current top line X2s, namely the 6400 at over 3.0ghz.

But whenever I see threads about it, pitted against slower clocked C2Ds, everyone dismisses the X2 and recommends the C2d. What is it about the C2Ds that make them superior?

Can someone briefly explain to a noob why the slower clocked C2Ds are better? Thanks!

 

[bryanW1995]

it does everything faster. games are faster. benchmarks are faster. burning cd's, dvd's, etc is faster. it's just a much better cpu. Most accounts give c2d a 25% or so clock advantage over x2. what is so confusing about this?

by the way, c2d also runs cooler. x2 is OLD technology, if you're die-hard amd wait for phenom to come out in november.

 

[myocardia]

Well, the difference performance-wise compared to their clockspeeds is just like it was with the Athlon 64's vs. the Pentium 4's. While the Pentium 4 was about 50% faster in clockspeed, it was considerably slower in processing power. And now the same applies to the Athlon 64's, compared to the C2D's; the C2D's process information at a faster rate (which is all processors do), even when they are clocked slower than the Athlon X2's.

 

[AmberClad]

Clock speed isn't the end all, be all. There's a little something called IPC -- Instructions Per Cycle. A chip with a slower pure clock speed might still perform better than a higher clocked chip because it essentially gets more done per clock cycle.

Someone else correct me on these numbers if I'm wrong:
Intel Pentium 4 = 2 IPC
Intel Core architecture = 4 IPC
AMD K8 = 3 IPC

This works both ways. In the days of Pentium 4, you'd see higher clocked P4s get outperformed by lower clocked AMD CPUs.

 

[Arkaign]

Originally posted by: bryanW1995
it does everything faster. games are faster. benchmarks are faster. burning cd's, dvd's, etc is faster. it's just a much better cpu. Most accounts give c2d a 25% or so clock advantage over x2. what is so confusing about this?

by the way, c2d also runs cooler. x2 is OLD technology, if you're die-hard amd wait for phenom to come out in november.

:thumbsup:

 

[Cogman]

Originally posted by: Sulaco
Hey guys,

I'm looking to upgrade in the near future and have been looking at CPUs. I've had my eye on the current top line X2s, namely the 6400 at over 3.0ghz.

But whenever I see threads about it, pitted against slower clocked C2Ds, everyone dismisses the X2 and recommends the C2d. What is it about the C2Ds that make them superior?

Can someone briefly explain to a noob why the slower clocked C2Ds are better? Thanks!

Wow, not to be mean, but you are a perfect example of how Intel kind of shot itself in the foot (as others have explained).

To go to analogy mode, You can do things fast, and you can do things efficiently. Imagine a pile of dirt, the P4 at 3.2 ghz is like one person shoveling the dirt (with a smaller shovel then the C2D I might add..) from one location to another, now this person is able to shovel pretty dang fast. Now along comes C2D, first off, he is two people, they have bigger muscles and larger shovels, They still move pretty fast but not quite as fast as the P4. Who moves more dirt? The C2D of course.

To compare with the X2, it is kind of the same thing, except it is just one on one again. C2D has more muscle and a bigger shovel, hence it moves the most dirt. (yes this analogy doesn't even begin to explain what is really going on and really has quite a few inaccuracies but it is good enough).

The guide,

Speed = clock speed.
Worker Muscle = IPC
Shovel = Cache and Instruction set.

 

[Cogman]

Originally posted by: Arkaign

Originally posted by: bryanW1995
it does everything faster. games are faster. benchmarks are faster. burning cd's, dvd's, etc is faster. it's just a much better cpu. Most accounts give c2d a 25% or so clock advantage over x2. what is so confusing about this?

by the way, c2d also runs cooler. x2 is OLD technology, if you're die-hard amd wait for phenom to come out in november.

:thumbsup:

I agree with everything except for the burning of cd's and dvd's part. That depends much more on the cd/dvd burner then the cpu. Now DVD encoding, that goes much faster.

BTW, if you are looking at upper end AMD, you might as well go for a Q6600 intel wise, 4 is better then 2 usually

 

[Markfw]

I have both. Example. I have an X2 4400@2.5 ghz. Takes over 2 days to do a F@H SMP unit. I have a C2D@3.2 that does it in one day. I have a C2Q (quad) @3.2 that does it in less than 12 hours.... You do the math as to what is faster.The X2 would need to be over 5 ghz to compete with the C2D.

 

[Sulaco]

First off, for some of the people here, I am not set on AMD at all, and I don't believe anything in my original post would lead anyone to that conclusion. In fact, I stated I was aware that C2D was superior, I merely wondered why; what it was that gave it the edge.

I've been out of the processor loop since before the C2D launched, and I haven't followed it religiously, as you might tell from my post count combined with my join date, and most articles I was able to dig up didn't compare the two directly, and most forum discussions didn't detail the reasons. I simply sought those reasons.
I know I opened myself to being condescended to by calling myself "noob", but that frankly was out humility and a genuine interest in the matter, ego aside.

Thanks for all the non-patronizing answers from everyone though. Much appreciated!

 

[Viditor]

The answer is in the name...
C2D has one of the most powerful cores ever produced.
While AMD still has a superior platform and interconnect, Intel's Core2 and their other modifications are more than a match for this...

Some terms you can Google are:
Memory Disambiguation which is a part of Intel's Smart Memory Access is probably the most profound advancement (IMHO).
Article Here

Advanced Smart Cache is basically Intel's shared L2...this was a very significant breakthrough as well. While a shared cache has been done before, I can't think of any example where it was this efficient.

To counter AMD's latency advantage from the On Die Memory Controller, C2D uses a very large L2. Intel is able to do this cost effectively (cache is a very expensive part of the chip because of it's size) because they have a huge process advantage in cache density...in other words, they can manufacture the same amount of cache in a much smaller area.

There are other things, but I beleive those are the biggest reasons...does that answer your question?

Edit: BTW, you will also see that C2D retires 4 instructions/clock as a reason, but this is a Red Herring...
C2D retires 4 uOps per clock while AMD decodes 3 full x86 instructions/clock (a uOp can be a part of a full instruction). Also, I can't think of any app that will allow this in any event...usually it's closer to 2 decodes/clock

 

[mouthwash]

well..........

its because INTEL excellent marketing strategy makes everyone believes that core2 is the better choice! dont fell for that!

lol, just j/k


core2 is just a much better architecture. It was derived from pentium m, the mobile processor in laptop, so it runs cooler, and consumes less power than previous cpus

thanks to the Isreal team.





correct me if i am wrong

 

[CTho9305]

Originally posted by: AmberClad
Clock speed isn't the end all, be all. There's a little something called IPC -- Instructions Per Cycle. A chip with a slower pure clock speed might still perform better than a higher clocked chip because it essentially gets more done per clock cycle.

Someone else correct me on these numbers if I'm wrong:
Intel Pentium 4 = 2 IPC
Intel Core architecture = 4 IPC
AMD K8 = 3 IPC

This works both ways. In the days of Pentium 4, you'd see higher clocked P4s get outperformed by lower clocked AMD CPUs.

Be careful that you don't buy into oversimplifications. While that is true using one way of interpreting peak IPC, taken alone it's incredibly misleading. For example, on this benchmark, K8 sustains 3 x86 instructions per cycle, but the "wider" core 2 can only do about 2.3 x86 instructions per cycle. It's a simple program too. Core 2 probably has a different mix of functional units available (for example, K8 has 3 ALUs and 3 AGUs), which just don't map well to the instruction mix of this benchmark. Alternately, the instructions used in that benchmark might cost multiple instruction "slots" in Core 2's microarchitecture. That bit is particularly tricky, since the 4/5 uop wide parts of Core 2 are counted using simpler uops than the 3 uop width of K8 is measured and some classes of instructions that fit in one K8 "uop" take multiple Core 2 "uops". I wrote a long post about this in another thread and copied it here if you're interested in more details.

Just to be clear, I'm not disputing that Core 2 performs very well in many real-world applications - I'm arguing that looking at those supposed IPC / pipeline width numbers is a good way to come to a bad conclusion. In this case, it works out for many real-world benchmarks... but not everything. There's no guarantee that the next "3 wide" machine won't demolish Core 2, or that a future "6-wide" machine wouldn't be destroyed by a "narrower" machine. There's gotta be a reason that large businesses find Barcelona interesting - how wide is Barcelona, compared to the current Xeons or future penryn-derived ones?

 

[Viditor]

Originally posted by: CTho9305

Originally posted by: AmberClad
Clock speed isn't the end all, be all. There's a little something called IPC -- Instructions Per Cycle. A chip with a slower pure clock speed might still perform better than a higher clocked chip because it essentially gets more done per clock cycle.

Someone else correct me on these numbers if I'm wrong:
Intel Pentium 4 = 2 IPC
Intel Core architecture = 4 IPC
AMD K8 = 3 IPC

This works both ways. In the days of Pentium 4, you'd see higher clocked P4s get outperformed by lower clocked AMD CPUs.

Be careful that you don't buy into oversimplifications. While that is true using one way of interpreting peak IPC, taken alone it's incredibly misleading. For example, on this benchmark, K8 sustains 3 x86 instructions per cycle, but the "wider" core 2 can only do about 2.3 x86 instructions per cycle. It's a simple program too. Core 2 probably has a different mix of functional units available (for example, K8 has 3 ALUs and 3 AGUs), which just don't map well to the instruction mix of this benchmark. Alternately, the instructions used in that benchmark might cost multiple instruction "slots" in Core 2's microarchitecture. That bit is particularly tricky, since the 4/5 uop wide parts of Core 2 are counted using simpler uops than the 3 uop width of K8 is measured and some classes of instructions that fit in one K8 "uop" take multiple Core 2 "uops". I wrote a long post about this in another thread and copied it here if you're interested in more details.

Just to be clear, I'm not disputing that Core 2 performs very well in many real-world applications - I'm arguing that looking at those supposed IPC / pipeline width numbers is a good way to come to a bad conclusion. In this case, it works out for many real-world benchmarks... but not everything. There's no guarantee that the next "3 wide" machine won't demolish Core 2, or that a future "6-wide" machine wouldn't be destroyed by a "narrower" machine. There's gotta be a reason that large businesses find Barcelona interesting - how wide is Barcelona, compared to the current Xeons or future penryn-derived ones?

Yeah...what he said.

Nice post CTho!