Isn't conroe cheating a bit with its performance?

[Philippine Mango]

"Secondly, two technologies that Intel pulled from its rabbit's hat are intelligent prefetchers and Advanced Smart Cache. "

Article

I know conroe has been out for a while and this seems like a bit late for me to be discussing this but I've recently come to the realization of what I'm about to say...

Isn't having prefetching and having smart cache a bit unfair? Sure it's good to take advantage of repetitive tasks and speed things up a bit but doesn't that make apparent performance a bit unfair or unreal? It just seems like it's kindof cheating kind of like zip compression where it can just skip a bunch of bits because it knows it has to do the same thing again. This type of thing is also similar with SSE and other technologies like that, right?

Meh just thought I'd say that... Also btw, I also heard that conroe can be slower or about the same speed as AMD's processors, so what cases does that occur?

Also, is conroe much faster than the P4s when it comes to folding at home? I ask this because I assume but have no idea but I'd believe folding at home doesn't have any repetitive tasks, or it could be the complete opposite and it's entirely repetitive... dunno

 

[jpeyton]

Not cheating.

 

[Cheex]

No way. It is just simply THAT GOOD!!

:thumbsup: :beer:

 

[Dadofamunky]

Ummmm.... noooooo......

 

[hardwareking]

not cheating.Just intel's unique way of improving performance.
And core 2 chips are much faster than P4's in folding @home.

 

[Gatt]

Originally posted by: Philippine Mango
"Secondly, two technologies that Intel pulled from its rabbit's hat are intelligent prefetchers and Advanced Smart Cache. "

Article

I know conroe has been out for a while and this seems like a bit late for me to be discussing this but I've recently come to the realization of what I'm about to say...

Isn't having prefetching and having smart cache a bit unfair? Sure it's good to take advantage of repetitive tasks and speed things up a bit but doesn't that make apparent performance a bit unfair or unreal? It just seems like it's kindof cheating kind of like zip compression where it can just skip a bunch of bits because it knows it has to do the same thing again. This type of thing is also similar with SSE and other technologies like that, right?

No, it isn't unfair, it's the way computers work.

The slowest action a computer can take is to load/store something from memory into the registers on the processor where the work takes place. It's several times slower than performing any other kind of instruction. A large part of this is because the processor has to go off it's Ghz level core to it's Mhz level memory(With the bandwidth between the memory being the slowest part, and actually finding it in memory taking more time).

Thus we have cache. One of the best ways to speed up a processor at this point is to reduce the penalty for memory access, and aside from adding more cache, the best way is to use that cache more efficiently. Hence the prefetchers and smart cache.

It's not cheating any more than AMD moving the memory controller on die, which is another approach to reducing the penalty for memory access.

The goal of both is to keep the processor(s) fed, a cycle lost waiting for memory access is that much less performance, a cycles lost to a mispredict is also that much less performance.

Increasing speed isn't the only way to increase performance. Increasing efficiency also helps, and as time progresses, these things will be absolutely neccessary. When we're looking at 4-8 cores running at 4ghz, keeping these things fed on a bandwidth limited FSB is going to be the next big issue. Better prefetchers, cache, and on-die memory controllers will be absolutely necessary to remove bottlenecking.

Consider it for a moment, what's going to happen when 4-8 processors all hit that FSB at the same time with load/stores? It's going to be some lost clock cycles as someone waits for their data, much easier to make it stay in the cache, especially once it's shared.

 

[Bateluer]

I can't speak for F@H, but my E6600 burns through S@H WUs.

 

[sdsdv10]

Originally posted by: Philippine Mango
Isn't having prefetching and having smart cache a bit unfair?

Isn't this quote, kind of like saying; "Isn't it unfair of AMD to have an on die memory controller?"

Chip designers/companies make choices and then they try to optimize those choices.
Nothing unfair or "cheating" about it. Just good business.

 

[Ichigo]

So bruteforcing is the only "non-cheating" way? End-user performance is all that counts.

 

[Duvie]

I think this whole thread is more about "how stupid is AMD not to have implemented it themselves?"......seems like a bunch of whining.....prefetching and prediction has been used by both parties for sometime....this has become an integral part of speeding up computing....

<<Also, is conroe much faster than the P4s when it comes to folding at home? I ask this because I assume but have no idea but I'd believe folding at home doesn't have any repetitive tasks, or it could be the complete opposite and it's entirely repetitive... dunno>>

I am not 100% sure about the P4....I can tell you it KICKED THE HECK out of my A64's, and hence why I got rid of them ALL...

 

[Hulk]

Not cheating, just being clever.

If the object is to get to the finish line at all costs you jump the fence instead of running around it.

That's just smart.

 

[StrangerGuy]

If anything, AMD got complacent and stuck with K8 for too long, and cannot produce a performance reply for Core 2 until for at least a year after Conroe debut.

 

[Corporate Thug]

by cheating do you mean being efficient?

 

[mithrandir2001]

Originally posted by: Philippine Mango
Sure it's good to take advantage of repetitive tasks and speed things up a bit but doesn't that make apparent performance a bit unfair or unreal?

By your logic, then, it would be better for the CPU to take longer to complete tasks because at least it is doing it "the hard and honest way".

Geez, you practically sound like a Puritan.

 

[garkon8]

Originally posted by: mithrandir2001

Originally posted by: Philippine Mango
Sure it's good to take advantage of repetitive tasks and speed things up a bit but doesn't that make apparent performance a bit unfair or unreal?

By your logic, then, it would be better for the CPU to take longer to complete tasks because at least it is doing it "the hard and honest way".

Geez, you practically sound like a Puritan.

Funny. Unfortunately, Nice guys finish last. The on-die memory controller was the killer at one time, but human ingenuity always finds a better, quicker way to the top.

 

[Conky]

Yes, it is faster so Intel must be cheating. :roll:

What a retarded thread. :thumbsdown:

 

[coldpower27]

Originally posted by: Philippine Mango
"Secondly, two technologies that Intel pulled from its rabbit's hat are intelligent prefetchers and Advanced Smart Cache. "

Article

I know conroe has been out for a while and this seems like a bit late for me to be discussing this but I've recently come to the realization of what I'm about to say...

Isn't having prefetching and having smart cache a bit unfair? Sure it's good to take advantage of repetitive tasks and speed things up a bit but doesn't that make apparent performance a bit unfair or unreal? It just seems like it's kindof cheating kind of like zip compression where it can just skip a bunch of bits because it knows it has to do the same thing again. This type of thing is also similar with SSE and other technologies like that, right?

Meh just thought I'd say that... Also btw, I also heard that conroe can be slower or about the same speed as AMD's processors, so what cases does that occur?

Also, is conroe much faster than the P4s when it comes to folding at home? I ask this because I assume but have no idea but I'd believe folding at home doesn't have any repetitive tasks, or it could be the complete opposite and it's entirely repetitive... dunno

It's a legitimate way to increase performance.

What matters is (to the consumer) are perf/watt, wattage, overall performance, and price. Some factors to a higher degree then others. At any rate, what matters is that it spits out performance, and if it were two monkeys banging on a drum, and gave extreme performance, with all the above characteristics, I would buy it in a heartbeat. How it get's there doesn't matter just so long as it does.

 

[judasmachine]

Originally posted by: Bateluer
I can't speak for F@H, but my E6600 burns through S@H WUs.

I can it burns through F@H WUs too.

 

[jiffylube1024]

By doing something faster how exactly is it cheating?

That's like saying unperfect Anisotropic filtering is cheating, or AA that only does edges, not the whole scene, etc.

Performance "tricks" are necessary for performance to keep improving. Why not call SSE or SIMD a cheat when the same instructions can be done without SIMD in a mere 100 CPU cycles while you're at it.