AMD 4x4 @ 3GHz VS. Intel QX6700 @ 2.66GHz

[anandtechrocks]

I didn't expect the AMD 4x4 to win this battle (and they didn't) but the performance was closer than I thought.

Besides the horrendous power usage of the 4x4 system, it doesn't look too bad. But that doesn't take into account that the QX6700 is, by default, clocked lower, probably overclocks better, is easier to cool, and in this test the 4x4 has 4GB RAM and the QX6700 only has 2GB.

3GHz 4x4 vs. QX6700

Here are some pictures of the setup:

CPU
Mobo
Mobo and CPUs

Here is the power usage comparison:

Power Consumption in Watts

 

[Cookie Monster]

On one hand you got one chip that doesnt run much hot give you the power of 4 cores.

But on the other hand, you have one very expensive motherboard (with tight layout), 2 CPUs which means more heat and power usage (add in extra fan noise), and negligable performance increase compared to the former, i have no idea where AMD is going with this 4x4 technology.

 

[coldpower27]

Originally posted by: Cookie Monster
On one hand you got one chip that doesnt run much hot give you the power of 4 cores.

But on the other hand, you have one very expensive motherboard (with tight layout), 2 CPUs which means more heat and power usage (add in extra fan noise), and negligable performance increase compared to the former, i have no idea where AMD is going with this 4x4 technology.

Right now AMD needs a 4 core platform to attempt to compete with Intel, remember having a solution at least puts you in the running regardless of how bad the solution may be at the current time.

The neat thing about the 4x4 platform is that it will be upgradable to K8L/K10 Quad Core derivatives, so AMD will have a 8 Core platform on the desktop, when they released their K8L/K10 technology by Mid 2007.

It's isn't a great solution now, but at least AMD is in the running. Remember Intel also released their Pentium EE 840 even though their technology has significant disadvantages compared to the offerings of the competition at the time. A company is not just going to sit by and give up when a competitor has a better product out there. You need to put a product out there ASAP, to show that your willing to compete in the segment.

How effective the competition is however is another matter entirely.

 

[tylerw13]

wow is the graph on the power consumption correct????? dang freakin amd eats power like no tomorrow

 

[Avalon]

Jeesh, look at that power consumption. I hope AMD's K8L and 65nm brings on some significant power savings if they hope to compete on the power envelope front when they release quad core.

 

[myocardia]

Originally posted by: Cookie Monster
On one hand you got one chip that doesnt run much hot give you the power of 4 cores.

Where are you getting your information from? A QX6700 runs >60°C under load, with the world's best heatpipe heatsink. With that same heatsink, people with overclocked 90nm 3.0 Ghz dual-core Opterons run about 55°C load.

 

[The-Noid]

I never hit 60. I don't know where you are coming from. I barely hit 50 on Big Typhoon when I am folding for hours.

 

[Duvie]

First off...

One they didn't really test the gpus in the way they were supposedly designed...What I mean is it is a poor translation but I dont think they really ran it with 4 cards in SLI and at optimal config..so you can throw out most of the game test...This is as much of a marketing of a platform then a cpu...For the reviewers to even the playing field was retarded. We knew how it was going to wash per cpu. AMD has gimick they are tryiong to sell to go after the gamers, not compete with quad cores in all arenas

Secondly...

Why would anyone expect it to do better mhz per mhz? I mean the comparison seems to be very similar to what we already see with current AMD versus C2D...i am not surprised....I knew it would basically be 3ghz of AMD to about 2.6ghz of INtel

Third...

It doesn't matter cpu speed...haven't you intel fanboys realized that yet? More important thing is cost per cpu comparison and overall cost of the platform...Whether one cpu is more efficient makes no difference. At least that is what all you cry babies said when AMD use to wax cpus clocked 1ghz higher....

Fourth and Final

Both of them are so extreme and beyond most users other then a paper pissig match most will not use either one of these. 4 cores is better served inthe business sector and professional markets...

Bottomline most of you can throw your 2 cents in, but most of you will never buy one anyways.

Until AMD comes out with new K8L core derivative (true quad core single chip) we wont know anything more then we already know now. This was not unexpected.


EDIT: I will let you know how well a QX6700 does in temps in ocing in about a week......


Let me alos state it is obvious from my testing that AMD would have been better served by uping the L2 cache. In some of thos items 4mb over 2mb is worth an extra 150-200mhz in performance. That could have been one way for AMD to up its head to head comparison to Intel. we all knew AMD can't clock it high enough to beat Intels current offerings. 65nm aside I dont think they will get above 3.4ghz if that...intels clearly look like with dual cores they could go 3.2ghz now (default vcore) and quad cores perhaps 3ghz with a higher default vcore

 

[nyker96]

I can't say 4x4 is a great solution compare to QX6700, it losses in all performance, power and price. But it does have the option of going higher with 2x quad cores when available. So I guess if you don't care for running 2xquad later. It would be pointless to get this.

 

[mamisano]

I would also like to know about how the OS was configured... was it set up to enable NUMA support?

 

[coldpower27]

Originally posted by: Duvie
First off...

One they didn't really test the gpus in the way they were supposedly designed...What I mean is it is a poor translation but I dont think they really ran it with 4 cards in SLI and at optimal config..so you can throw out most of the game test...This is as much of a marketing of a platform then a cpu...

Secondly...

Why would anyone expect it to do better mhz per mhz? I mean the comparison seems to be very similar to what we already see with current AMD versus C2D...i am not surprised....I knew it would basically be 3ghz of AMD to about 2.6ghz of INtel

Third...

It doesn't matter cpu speed...haven't you intel fanboys realized that yet? More important thing is cost per cpu comparison and overall cost of the platform...Whether one cpu is more efficient makes no difference. At least that is what all you cry babies said when AMD use to wax cpus clocked 1ghz higher....

Fourth and Final

Both of them are so extreme and beyond most users other then a paper pissig match most will not use either one of these. 4 cores is better served inthe business sector and professional markets...

Bottomline most of you can throw your 2 cents in, but most of you will never buy one anyways.

Until AMD comes out with new K8L core derivative (true quad core single chip) we wont know anything more then we already know now. This was not unexpected.


EDIT: I will let you know how well a QX6700 does in temps in ocing in about a week......

First off..
Well the game test shows what the system can do with a 7900 GTX SLI setup, it's just 2 cards rather then 4. Performance should be inline with cards of that caliber on Dual Core CPU's.

Second off...
Some people expected the HyperTransport links to provide some sort of an advantage over the traditional older FSB design that Intel uses for the Core 2 Extreme QX6700. So they expected the gap to close on MHZ to MHZ basis. I seriously doubt anybody will expect the 4x4 platform to close the clock for clock advantage that Core micro-architecture has entirely.

Third off...

What matters is the performance of the solution, and the energy it consumes as well as the price of acquiring the platform. To some people, branding also actually plays a role, though to whether or not you feel this is justified is another issue entirely. Depending on the individual one or more of these factors might have more weight then another.

Clock per clock efficiency comes into play if the architectures have similar clockspeed, then the winner will be the core which can get a greater amount of work done per clock cycle.

It also comes into play if the architectures have wildly differing clockspeed to normalize the amount of performance each processor gets. So overall efficiency is an important factor to consider.

Depends on where you draw the line of "so extreme" the high end solutions are always the ones that get the most praise this is obvious, whether or not you buy one. They are designed to be the most glamourous and to bring good rep to a company.

This is a chicken and the egg scenario, ever heard the phrase of if you build it they will come?, the introduction of Quad Core Systems now gives developers time to create programs that do indeed take advantage of the processing power 4 cores have to offer. The sooner it is introduced to the high end sector for desktops, the sooner it can start filtering down into the mainstream segments. I think it's also pretty obvious people are aware of how much income they have to spend on computer hardware.

 

[BigDH01]

Such a mistake for AMD... a 2220 Santa Rosa costs about 800 dollars on its own. Now they will sell 2 for that price just to say they are in desktop quad core game. If I was a shareholder, I would not be happy. They are already selling out Santa Rosas, why give them away? And for what? To come close to matching the performance of the Intel quad core? I think this is a foolish move on AMD's part. Just wait until you have native quad core and then compete, market penetration will be low until then anyway.

 

[Duvie]

Originally posted by: coldpower27

Originally posted by: Duvie
First off...

One they didn't really test the gpus in the way they were supposedly designed...What I mean is it is a poor translation but I dont think they really ran it with 4 cards in SLI and at optimal config..so you can throw out most of the game test...This is as much of a marketing of a platform then a cpu...

Secondly...

Why would anyone expect it to do better mhz per mhz? I mean the comparison seems to be very similar to what we already see with current AMD versus C2D...i am not surprised....I knew it would basically be 3ghz of AMD to about 2.6ghz of INtel

Third...

It doesn't matter cpu speed...haven't you intel fanboys realized that yet? More important thing is cost per cpu comparison and overall cost of the platform...Whether one cpu is more efficient makes no difference. At least that is what all you cry babies said when AMD use to wax cpus clocked 1ghz higher....

Fourth and Final

Both of them are so extreme and beyond most users other then a paper pissig match most will not use either one of these. 4 cores is better served inthe business sector and professional markets...

Bottomline most of you can throw your 2 cents in, but most of you will never buy one anyways.

Until AMD comes out with new K8L core derivative (true quad core single chip) we wont know anything more then we already know now. This was not unexpected.


EDIT: I will let you know how well a QX6700 does in temps in ocing in about a week......

First off..
Well the game test shows what the system can do with a 7900 GTX SLI setup, it's just 2 cards rather then 4. Performance should be inline with cards of that caliber on Dual Core CPU's.

Second off...
Some people expected the HyperTransport links to provide some sort of an advantage over the traditional older FSB design that Intel uses for the Core 2 Extreme QX6700. So they expected the gap to close on MHZ to MHZ basis. I seriously doubt anybody will expect the 4x4 platform to close the clock for clock advantage that Core micro-architecture has entirely.

Third off...

What matters is the performance of the solution, and the energy it consumes as well as the price of acquiring the platform. To some people, branding also actually plays a role, though to whether or not you feel this is justified is another issue entirely. Depending on the individual one or more of these factors might have more weight then another.

Clock per clock efficiency comes into play if the architectures have similar clockspeed, then the winner will be the core which can get a greater amount of work done per clock cycle.

It also comes into play if the architectures have wildly differing clockspeed to normalize the amount of performance each processor gets. So overall efficiency is an important factor to consider.

Depends on where you draw the line of "so extreme" the high end solutions are always the ones that get the most praise this is obvious, whether or not you buy one. They are designed to be the most glamourous and to bring good rep to a company.

This is a chicken and the egg scenario, ever heard the phrase of if you build it they will come?, the introduction of Quad Core Systems now gives developers time to create programs that do indeed take advantage of the processing power 4 cores have to offer. The sooner it is introduced to the high end sector for desktops, the sooner it can start filtering down into the mainstream segments. I think it's also pretty obvious people are aware of how much income they have to spend on computer hardware.

All very true...On the last part though I cant fully agree because I am a bit concerned at how much we have seen in development for dual core systems that have now been out for 14-16 months....I am not sure we would see benefits to that within 2-3 years IMO....


I htink this was meant to be an uber gaming system for extreme users that dont care about cost. To be honest SLI systems dont really pencil out for bang for the buck.

Also I am stil concerned that the system even using the 7900's was not set up right. Hence the bad translation but the article says something sort of like that...albeit very broken english. I think this is a pltform comparison and the one AMD designed for was not being ran. Also this really should have been no different then a dual dual core opteron setup. I think some of the hyp[ertransport isues with the SLI chip may be early on (not ready for primetime) drivers that we are accustomed from Nvidia products...hence why they have so many new driver updates...Like a weekly event...Performance may be enhanced when the system is configured correctly. When It is I see it as being the gaming champ...For cost, power consumption, and cpu efficiency I see it has being 2nd fiddle and not worth it....

But to each his own!!!

 

[myocardia]

Originally posted by: Yoxxy
I never hit 60. I don't know where you are coming from. I barely hit 50 on Big Typhoon when I am folding for hours.

71C load temps with stock fan @ 100%: link And I can't remember her name, but that girl here on anandtech with the QX6700 ES said she was getting >60C when she had hers on air, with a Scythe Infinity. I questioned that, and she gave me a link, to another forum, with other people complaining about how toasty their QX6700's were, with good heatsinks.

 

[Duvie]

Originally posted by: myocardia
71C load temps with stock fan @ 100%: link And I can't remember her name, but that girl here on anandtech with the QX6700 ES said she was getting >60C when she had hers on air, with a Scythe Infinity. I questioned that, and she gave me a link, to another forum, with other people complaining about how toasty their QX6700's were, with good heatsinks.

However I dont think that is a good example. It s obvious a similar speed Quad core would not have the same TDP especially since it is 2 C2D's slapped together. So using the same C2D stock cooler is a bad example.

So in similar setup as he has there in the review, my E6600 at 3.4ghz w/ 1.47v only idles at 29c and loads at 50c with orthos. What does that say about the performance my Sycthe Ninjas should give me with a QX6700? I dont plan on giving it that much vcore first off.

 

[myocardia]

Originally posted by: Duvie
However I dont think that is a good example. It s obvious a similar speed Quad core would not have the same TDP especially since it is 2 C2D's slapped together. So using the same C2D stock cooler is a bad example.

So in similar setup as he has there in the review, my E6600 at 3.4ghz w/ 1.47v only idles at 29c and loads at 50c with orthos. What does that say about the performance my Sycthe Ninjas should give me with a QX6700? I dont plan on giving it that much vcore first off.

But, the concept is identical, Duvie. You're still cooling 2 complete E6700's, with one heatsink. And you must have missed the part where they said that those temps were at stock speed, with stock voltages. Wait, do you have a special quad core Scythe Ninja?

 

[The-Noid]

Could be poor voltage regulation. Also the stock cooler for the Quad just came out with the retail the ES's were not sent the new cooler. It is much bigger than the C2D cooler and also has a fan that spins much faster.

 

[Duvie]

Originally posted by: myocardia

Originally posted by: Duvie
However I dont think that is a good example. It s obvious a similar speed Quad core would not have the same TDP especially since it is 2 C2D's slapped together. So using the same C2D stock cooler is a bad example.

So in similar setup as he has there in the review, my E6600 at 3.4ghz w/ 1.47v only idles at 29c and loads at 50c with orthos. What does that say about the performance my Sycthe Ninjas should give me with a QX6700? I dont plan on giving it that much vcore first off.

But, the concept is identical, Duvie. You're still cooling 2 complete E6700's, with one heatsink. And you must have missed the part where they said that those temps were at stock speed, with stock voltages. Wait, do you have a special quad core Scythe Ninja?

Actually what I was saying in the first part is that using a stock E6700 HSF is not fair because it is designed for the TDP of the dual core...Therefore anyone with any commonsense would not use that heatsink. To use a stock E6700 and then come to conclusion you wont be able to cool the QX6700 is a stretch. Obviously it should be hotter. that goes without saying.


On the scythe end of the comment i was just referring to the fact it seems to cool my much overclocked E6600 (similar 4mb) quite well so if we look at the delta of the 2 on the stock cooler under load at minimum we should seem similar delta. In that case I should be able OC it well and not hit 60+ temps.

We will see ofcourse but my E6400 at similar OC speed of 3.4ghz with similar vcore to the E6300 I had with stock cooler is almost 18c cooler under orthos....These Ninjas are no slouch and the swing from idle to load is much better.

Other then stress testing these cpu at maximum will run 24/7 folding whic usually runs 10c cooler then dual orthos. The rest of the time it will be rendering 3d stuff (which does not effectively use all 4 cores now)

 

[myocardia]

Originally posted by: Duvie
Actually what I was saying in the first part is that using a stock E6700 HSF is not fair because it is designed for the TDP of the dual core...Therefore anyone with any commonsense would not use that heatsink. To use a stock E6700 and then come to conclusion you wont be able to cool the QX6700 is a stretch. Obviously it should be hotter. that goes without saying.


On the scythe end of the comment i was just referring to the fact it seems to cool my much overclocked E6600 (similar 4mb) quite well so if we look at the delta of the 2 on the stock cooler under load at minimum we should seem similar delta. In that case I should be able OC it well and not hit 60+ temps.

We will see ofcourse but my E6400 at similar OC speed of 3.4ghz with similar vcore to the E6300 I had with stock cooler is almost 18c cooler under orthos....These Ninjas are no slouch and the swing from idle to load is much better.

Other then stress testing these cpu at maximum will run 24/7 folding whic usually runs 10c cooler then dual orthos. The rest of the time it will be rendering 3d stuff (which does not effectively use all 4 cores now)

That wasn't my point, Duvie. My point was that your E6600 runs almost 50C with only 4MB of cache, and only two cores. You're now going to be cooling 8MB of L2 cache, along with 4 cores. Don't expect similar temps, unless you don't overclock at all, in other words. And yeah, I have my dual-core Scythe Ninja installed on my X2 finally. I know how well they cool.

 

[Duvie]

Originally posted by: myocardia

Originally posted by: Duvie
Actually what I was saying in the first part is that using a stock E6700 HSF is not fair because it is designed for the TDP of the dual core...Therefore anyone with any commonsense would not use that heatsink. To use a stock E6700 and then come to conclusion you wont be able to cool the QX6700 is a stretch. Obviously it should be hotter. that goes without saying.


On the scythe end of the comment i was just referring to the fact it seems to cool my much overclocked E6600 (similar 4mb) quite well so if we look at the delta of the 2 on the stock cooler under load at minimum we should seem similar delta. In that case I should be able OC it well and not hit 60+ temps.

We will see ofcourse but my E6400 at similar OC speed of 3.4ghz with similar vcore to the E6300 I had with stock cooler is almost 18c cooler under orthos....These Ninjas are no slouch and the swing from idle to load is much better.

Other then stress testing these cpu at maximum will run 24/7 folding whic usually runs 10c cooler then dual orthos. The rest of the time it will be rendering 3d stuff (which does not effectively use all 4 cores now)

That wasn't my point, Duvie. My point was that your E6600 runs almost 50C with only 4MB of cache, and only two cores. You're now going to be cooling 8MB of L2 cache, along with 4 cores. Don't expect similar temps, unless you don't overclock at all, in other words. And yeah, I have my dual-core Scythe Ninja installed on my X2 finally. I know how well they cool.

I know....I am not expecting to hit 3.4ghz or let alone want to apply that much vcore...So right off the bat I will be quite different. Like I mentioned if you look at anything useful out of that link is what the delta difference is using 4 cores....even after second look the dual core cpu is the 2.93ghz model....They had 7c differnce on a stock cooler that was already quite overmatched....

Talking to yoxxy and some others they have been able to oc up intot he range I will be with no water cooling and temps more reasonable then was listed in that link....

 

[sandorski]

4x4 is very niche and was never meant to be put into widespread use, at least not at this time. Which is most likely why AMD favours Asus as the sole Mobo manufacturer, just not expected to sell in numbers large enough to support competition. That said, who knows what the future might bring? If 4x4 shows some advantages for gaming, perhaps AMD will use the concept in the future and open it up to a broader CPU line(regular Consumer CPUs). Perhaps AMD is just testing some future platform Tech or Demoing Tech for others(future Consoles?)?

I don't know, it might all be about making the best Gaming Machine(yet to be tested) on the Planet just for the recognition. Kinda like how the Auto Industry makes a Super Fast car for bbreaking Records or a Concept Vehicle for an Auto Show.

 

[myocardia]

Originally posted by: Duvie
I know....I am not expecting to hit 3.4ghz or let alone want to apply that much vcore...So right off the bat I will be quite different. Like I mentioned if you look at anything useful out of that link is what the delta difference is using 4 cores....even after second look the dual core cpu is the 2.93ghz model....They had 7c differnce on a stock cooler that was already quite overmatched....

Talking to yoxxy and some others they have been able to oc up intot he range I will be with no water cooling and temps more reasonable then was listed in that link....

Ahh, I had missed where you mentioned delta in that previous post. Well, actually, since you didn't say "temp delta", it just kind of didn't ring any bells. You're right about that, there ought to be a much larger temp delta than 7°C between an X6800 and two E6700's, using the same heatsink, if they were that hard to cool.

BTW, what video card are you using for rendering now that you can't use your modded 6800GT? The reason I ask is because you might consider buying an X1900/1950XT for your E6600. You'd put out way WU's than an overclocked QX6700 would, unless it also had an X1900/1950XT.

 

[Duvie]

I pm'd you Myocardia

 

[Cookie Monster]

Originally posted by: myocardia

Originally posted by: Cookie Monster
On one hand you got one chip that doesnt run much hot give you the power of 4 cores.

Where are you getting your information from? A QX6700 runs >60°C under load, with the world's best heatpipe heatsink. With that same heatsink, people with overclocked 90nm 3.0 Ghz dual-core Opterons run about 55°C load.

Link

Here is one of MANY examples of the temps for kensfield.

Regardless, within this test setup environment, the Core 2 Extreme QX6700 idles at a mild 31C. Under full 100% load with a 30 minute burn-in time, the new quad-core CPU only topped out at only 46-47C. Talk about a huge contrast to Intel's Pentium D thermal characteristics from days gone by. A quad-core Pentium D 965 would have been a toaster of a chip but the Core 2 Extreme QX6700 is a relatively cool operator. Next up - power consumption.

Frankly many other reviews back this claim. The tempurature may differ (figures all vary but the conclusion is always the same), but they all agree its much cooler than last generation dual core chips.

They mentioned that the stock cooler is very poor. But, any normal enthusiasts would go off and by a good CPU cooler if theyre shelling out on a very expensive CPU anyway.

 

[aka1nas]

Originally posted by: sandorski
4x4 is very niche and was never meant to be put into widespread use, at least not at this time. Which is most likely why AMD favours Asus as the sole Mobo manufacturer, just not expected to sell in numbers large enough to support competition. That said, who knows what the future might bring? If 4x4 shows some advantages for gaming, perhaps AMD will use the concept in the future and open it up to a broader CPU line(regular Consumer CPUs). Perhaps AMD is just testing some future platform Tech or Demoing Tech for others(future Consoles?)?

I don't know, it might all be about making the best Gaming Machine(yet to be tested) on the Planet just for the recognition. Kinda like how the Auto Industry makes a Super Fast car for bbreaking Records or a Concept Vehicle for an Auto Show.

IMO, what killed the 4x4's only real advantage is that the slot layout doesn't allow you to fit 4 dual slot cards. Not much of a way around that with ATX of course. This platform will be more interesting next year with dual quad-core K8Ls and more power-efficient DirectX 10 cards with single-slot coolers.