Conroe & AM2...disappointed by both...

[imported_inspire]

Originally posted by: harpoon84

Originally posted by: inspire
Fine. I read your 'more than match' as a 'beat by a significant difference'. I'm still trying to figure out the math you used and how it applies to my arguement, because if you take both of them together, the obvious conclusion is that the E6600 scales just as well, if not better than the X6800 in a hypothetical OC. Which may be true.

You'll have to excuse me for geeting a little defensive when the entire first page of the forums is about how much Intel rocks right now - I think most people have gotten the picture, and the who haven't, won't.

I see where you are coming from, but the thing is, it's harder to use the X6800 for comparison because you can't put a definitive 'it is 20% faster' tag on it - in some tests it is only 10% faster, while in some tests it would be 30% faster, and unless someone has spreadsheeted all the benchmark numbers and graded the X6800 v FX-62 victories in % form, it is just much easier to use a CPU that is *comparable* to the FX-62, in this case an E6600, and then make the relevant calculations from there.

Agreed. So, then it comes to be:

The FX-62 runs at 2.8 GHz Stock, so an OC to 4.0 GHz is a 42.9% Overclock.

On the other hand, the E6600 runs at 2.4 GHz Stock, so an OC to 3.5 GHz is a 45.8% Overclock.

This is significant, considering the E6600 has an advantage in the majority of the current benchmarks.

I would add, that such a spreadsheet wouldn't take much more than 10 minutes in Excel, seeing as how Anandtech did all the same benches with the FX-62 and the X6800. Maybe later today I'll get to that.

 

[harpoon84]

Originally posted by: inspire

Originally posted by: harpoon84

Originally posted by: inspire
Fine. I read your 'more than match' as a 'beat by a significant difference'. I'm still trying to figure out the math you used and how it applies to my arguement, because if you take both of them together, the obvious conclusion is that the E6600 scales just as well, if not better than the X6800 in a hypothetical OC. Which may be true.

You'll have to excuse me for geeting a little defensive when the entire first page of the forums is about how much Intel rocks right now - I think most people have gotten the picture, and the who haven't, won't.

I see where you are coming from, but the thing is, it's harder to use the X6800 for comparison because you can't put a definitive 'it is 20% faster' tag on it - in some tests it is only 10% faster, while in some tests it would be 30% faster, and unless someone has spreadsheeted all the benchmark numbers and graded the X6800 v FX-62 victories in % form, it is just much easier to use a CPU that is *comparable* to the FX-62, in this case an E6600, and then make the relevant calculations from there.

Agreed. So, then it comes to be:

The FX-62 runs at 2.8 GHz Stock, so an OC to 4.0 GHz is a 42.9% Overclock.

On the other hand, the E6600 runs at 2.4 GHz Stock, so an OC to 3.5 GHz is a 45.8% Overclock.

This is significant, considering the E6600 has an advantage in the majority of the current benchmarks.

I would add, that such a spreadsheet wouldn't take much more than 10 minutes in Excel, seeing as how Anandtech did all the same benches with the FX-62 and the X6800. Maybe later today I'll get to that.

Exactly. Granted it is a crude estimation, for all intents Conroe may scale better or worse against A64 as clockspeeds increase, we don't know for sure, I was just making the assumption that they scaled identically.

Anyhow, this is all hypothetical because I would sell my left nut if AMD came out with a 4GHz A64 tomorrow.

Kind of like people wishing Intel would release a 5GHz P4 to counter the A64, it just ain't gonna happen!

 

[Hard Ball]

Originally posted by: fatty4ksu
conroe is best cpu ever

Vintage Fatty, served up.

 

[imported_inspire]

Got it done.

They make you wait like 30 seconds before you can download it, but I took a lot more than 30 seconds making it, so I hope nobody minds waiting.

Keep in mind that the cumulative scores are all equally weighted - not that I think they necessarily should be, I just didn't feel like putting much thought into it. And, plus - this way my figure of a 20% average advantage for the X6800 over the FX-62 is fairly substantiated.

I just thought that originally, it was a pretty gutsy thing to say without showing how you arrived at such a conclusion. In the end you were right, but the margins, by these numbers, were close enough to take a second look at.

 

[JPH1121]

I'll be the first to admit that I certainly can't make a chip let alone improve, from scratch, upon what is already out there. Hell, I doubt I could improve upon what's out there WITH their blueprints in hand.

But despite that fact, that was an irrelevant and asinine comment and quote.

With three years time, a few thousand engineers SHOULD be able to make an improvement architectually of greater than a generous 30%.

Anybody have some numbers for the performance improvements that have been made by the significantly smaller companies ATI and nVidia for comparison?

 

[coldpower27]

Originally posted by: JPH1121
I'll be the first to admit that I certainly can't make a chip let alone improve, from scratch, upon what is already out there. Hell, I doubt I could improve upon what's out there WITH their blueprints in hand.

But despite that fact, that was an irrelevant and asinine comment and quote.

With three years time, a few thousand engineers SHOULD be able to make an improvement architectually of greater than a generous 30%.

Anybody have some numbers for the performance improvements that have been made by the significantly smaller companies ATI and nVidia for comparison?

That's entirely different as it's alot easier to increase performance when processing is inherently parralell as with graphics. not to mention generational cycles happen way quicker.

 

[harpoon84]

Originally posted by: JPH1121
I'll be the first to admit that I certainly can't make a chip let alone improve, from scratch, upon what is already out there. Hell, I doubt I could improve upon what's out there WITH their blueprints in hand.

But despite that fact, that was an irrelevant and asinine comment and quote.

With three years time, a few thousand engineers SHOULD be able to make an improvement architectually of greater than a generous 30%.

Anybody have some numbers for the performance improvements that have been made by the significantly smaller companies ATI and nVidia for comparison?

True, and both AMD and Intel are guilty in all this.

Intel ran into a thermal wall in Prescott, and had all kinds of problems scaling it up, eventually giving up on a 4GHz model altogether. From 03 to 06 the clockspeed on the fastest P4 increase a 'whopping' 600MHz, from 3.2GHz to 3.8GHz. Yes then we had dual core blah blah but that was because Intel could go no further clockspeed wise.

Seeing as Intel couldn't effectively catch them with the Netburst architecture, AMD rested on their laurels and stagnated the A64 clockspeed increases too. From 03 to 06, AMD has increased the clockspeed on A64s from 2GHz to 2.8GHz. That really is quite ridiculous, seeing as AMD didn't have any of the thermal problems Intel had, they had a smooth transistion from 130nm to 90nm. There just wasn't the competition from Intel for AMD to get on their bike and produce something better.

And for that, AMD is paying the price now because a mid range Conroe is outperforming their flagship CPU.

 

[WhoKnowsWho]

Look at how slow multi threaded apps are coming along. I wonder how many improvements we could get just from better software.

 

[dmens]

Originally posted by: JPH1121
I'll be the first to admit that I certainly can't make a chip let alone improve, from scratch, upon what is already out there. Hell, I doubt I could improve upon what's out there WITH their blueprints in hand.

But despite that fact, that was an irrelevant and asinine comment and quote.

With three years time, a few thousand engineers SHOULD be able to make an improvement architectually of greater than a generous 30%.

Anybody have some numbers for the performance improvements that have been made by the significantly smaller companies ATI and nVidia for comparison?

So despite your admitted ignorance, you insist both companies are slacking off in regards to new architectures? Please. The fact you're comparing the absurdly parallel world of 3D processing with general computing performance is bogus to begin with.

But hey, I'll give you some credit. If the issues of power/heat, die area, yields/costs, design schedule, resources to deal with complexity, tool/simulation limitations, risk management, validation, silicon/test debug, and more could be ignored, then yeah, we'd be seeing much more than 30% performance increase on the same clock. Too bad reality is a bitch.

 

[ayabe]

The problem is that we really aren't taking advantage of the dual core architecture even now with quad cores right around the corner. That's why I think a lot of people are disappointed with the "advances" of the last couple years. Multicore is the new roadmap for CPU improvements but unfortunately the software isn't keeping pace.

 

[phantom404]

All i see here is Intel finally admitting that AMD was on the right track in focusing on the Architect of the cpu instead of clock speeds so they decided to switch and copy.....errr come up with their improved cpu architect instead of focusing on cpu speeds alone.

 

[JPH1121]

So despite your admitted ignorance, you insist both companies are slacking off in regards to new architectures? Please. The fact you're comparing the absurdly parallel world of 3D processing with general computing performance is bogus to begin with.

But hey, I'll give you some credit. If the issues of power/heat, die area, yields/costs, design schedule, resources to deal with complexity, tool/simulation limitations, risk management, validation, silicon/test debug, and more could be ignored, then yeah, we'd be seeing much more than 30% performance increase on the same clock. Too bad reality is a bitch.[/quote]


so...you're trying to tell me that those issues don't exist for nVidia and ATI? I was unaware that their chips were made from...er...non...real...stuff? Last I checked, most of the companies for sale on Wall Street suffer from the same or similar problems.

Furthermore, even maintaining the same number of pipelines there are shocking differences in performance. Compare a x1800xt to a x850xt. At times there is a 25% performance improvement @ 1280x1024. I personally can't tell you how long it took for them to jump from the 850 to the 1800...but it sure as hell appears that they did something right.

I don't know, as I said, maybe I'm ignorant. But REALLY people, an architecture can obviously make a difference and if a company a third the size of AMD can do it, why the hell can't AMD? (I can't divide well enough to throw Intel's name in there )

 

[FallenHero]

Originally posted by: phantom404
All i see here is Intel finally admitting that AMD was on the right track in focusing on the Architect of the cpu instead of clock speeds so they decided to switch and copy.....errr come up with their improved cpu architect instead of focusing on cpu speeds alone.

To be fair, clock speed was keeping up for awhile. Of course now, Intel is beating AMD at their own game...

 

[zsdersw]

Why is it that Conroe fails to impress versus the K8? If the K8 was the "kick ass" design everyone thinks it is.. and Conroe beats it.. what's not impressive about that? Oh, that's right.. I forgot. When AMD does something better, it's the greatest thing since sliced bread.. but when Intel does something better, it's what's "expected". Sorry, but either they're both impressive when they leapfrog each other.. or they're both just doing what's "expected".

 

[mhahnheuser]

Originally posted by: WhoKnowsWho
Look at how slow multi threaded apps are coming along. I wonder how many improvements we could get just from better software.

Yep, and multi threads a long way off too, i reckon. Thank the good lord that celly's and sempy's have finally moved to 64 bit so we might be lucky and start to get some 64 bit software shortly starting with Ms Vista, that will start to make our investments seem a bit more worthwhile. Just as well A64 and Conroe are giving us some performance boosts or things would be really crappy. (Probably only because Nvidia and ATI have kept demanding faster cpu's to go with their lightning fast gpu's)

 

[coldpower27]

Originally posted by: JPH1121
So despite your admitted ignorance, you insist both companies are slacking off in regards to new architectures? Please. The fact you're comparing the absurdly parallel world of 3D processing with general computing performance is bogus to begin with.

But hey, I'll give you some credit. If the issues of power/heat, die area, yields/costs, design schedule, resources to deal with complexity, tool/simulation limitations, risk management, validation, silicon/test debug, and more could be ignored, then yeah, we'd be seeing much more than 30% performance increase on the same clock. Too bad reality is a bitch.

so...you're trying to tell me that those issues don't exist for nVidia and ATI? I was unaware that their chips were made from...er...non...real...stuff? Last I checked, most of the companies for sale on Wall Street suffer from the same or similar problems.

Furthermore, even maintaining the same number of pipelines there are shocking differences in performance. Compare a x1800xt to a x850xt. At times there is a 25% performance improvement @ 1280x1024. I personally can't tell you how long it took for them to jump from the 850 to the 1800...but it sure as hell appears that they did something right.

I don't know, as I said, maybe I'm ignorant. But REALLY people, an architecture can obviously make a difference and if a company a third the size of AMD can do it, why the hell can't AMD? (I can't divide well enough to throw Intel's name in there )[/quote]

There is about a 17% improvement in clockspeed between the X850 XT and X1800 XT, while on the other hand memory bnadwidth is improved by a bit over 33%.

So on the IPC front your looking at maybe an 8-10% improvement. In terms of making the Shader pipelines more powerful.

As well R480 technology hasn't seen IPC improvements since the original Radeon 9700 Pro from back in September 2002.

Graphics vendors have more options to increase performance.

IPC per pipeline, usually little is done here as this is difficult to do.

clockspeeds, moderately used.

increase the number of total functional units, the primary source of performance improvements.

For CPU vendors they have these options.

increasing the number of functional units, rare and on the whole useless unless software takes advantage of it.

IPC increases, this is geerally the moderately used arena

clockspeed, generally the most used to increase performance

See how the CPU vendors places and different emphasis on what to increase over what GPU vendors increase. It's not the same concept, increase the stuff for which is optimal for GPU's doesn't work that well with CPU's

 

[dmens]

Originally posted by: JPH1121
so...you're trying to tell me that those issues don't exist for nVidia and ATI? I was unaware that their chips were made from...er...non...real...stuff? Last I checked, most of the companies for sale on Wall Street suffer from the same or similar problems.

Furthermore, even maintaining the same number of pipelines there are shocking differences in performance. Compare a x1800xt to a x850xt. At times there is a 25% performance improvement @ 1280x1024. I personally can't tell you how long it took for them to jump from the 850 to the 1800...but it sure as hell appears that they did something right.

I don't know, as I said, maybe I'm ignorant. But REALLY people, an architecture can obviously make a difference and if a company a third the size of AMD can do it, why the hell can't AMD? (I can't divide well enough to throw Intel's name in there )

not to dismiss the challenges of GPU design, but a CPU design is far more complex on basically all fronts. a GPU is much more easily pipelined, runs at a lower frequency, and the job is so parallel, tossing more resources results in a linear gain in performance (or close). none of that applies in the CPU world.

 

[theteamaqua]

i think thats why next year they have quad cores kentsfield and K8L

i truly hope that K8L can be better than kentsfield otherwise AMD is doomed and out of business and i dont want to use Intel for the rest of my life. umm monopolize anyone??