Intel 5x faster per watt.

[Targon]

Ok, I've gone through all 12+ pages of these posts going back and forth, and decided to finally chime in.

Intelia, you can't compare a water cooled or phase change cooled processor to an air cooled chip and say the water cooled is better. Getting a chip to run stable for extended periods of time(not just the 5 minutes it takes to run a benchmark) with normal air cooling, not even with heat pipe cooling is what the processor companies are selling. If a company or person NEEDS fancy cooling methods, it means they arn't running under similar conditions to people who are running with only a heat sink+fan.

Running outside of a case on a desk with a large fan or air conditioning blowing on the setup also isn't a fair test because that's NOT the environment processors are designed to be run at.


Next up, this whole "performance per watt" thing. For a LAPTOP, or where people are trying to lower their power usage due to electricity costs, then the performance per watt issue is valid. BUT, it means NOTHING when you are discussing which processor has the best performance. Either Intel or AMD can release a new version of a chip virtually unchanged, same clock rate but on a newer process(going to .65 micron from .90 micron for example) and generally get an improvement per watt. For a desktop, workstation, or server that has PERFORMANCE as the only metric that matters, performance per watt doesn't matter. So let's ignore all the hype that Intel is trying to spread when it comes to their "new" metric that they are trying to sell.

I agree that Intel MAY have an edge in a MHz to MHz battle between the new chips when they are released and what AMD will have at the time, but if AMD has their top end at 3.2GHz and Intel has their top end for the new processors at 2.2GHz, AMD will be ahead in performance by quite a bit. The difference between Pentium-M(whatever the generation) and the AMD64 processors when it comes to a MHz to MHz comparison isn't anywhere near as big a gap in performance as the Pentium 4 is compared to AMD64 processors.

I am also looking forward to how well the new desktop and server chips from Intel will perform, but I don't hold any illusions about how things are when a chip is released. Unless Intel wants to provide a water cooling setup with each CPU so it can clock high enough to beat the best AMD processors available in the retail/OEM channel, if their top end released processors can't beat the top end AMD processors as sold in a boxed set(which is currently CPU+heat sink+fan), then Intel won't be ahead of AMD in performance.

You also need to understand that you can't compare a $1000 CPU to a $700 CPU, no matter what the clock speed is. Both AMD and Intel sell processors at the $900-$1000 price points. You need to compare based on these issues. Also, a dual/multi-core chip shouldn't be compared to a single core chip of the same price because the design of multi-core chips is MEANT for multi-threaded and multi-tasking environments. A $1000 single-core chip will run single-threaded applications better than a $1000 dual-core chip will, while the multi-core chip will run multi-threaded and multiple applications "better". Apples to apples comparisons are needed. For a review, it makes sense to show the difference in performance in different situations between single and multi-core chips because for those who don't know for certain which is better for THEIR needs they may not know which will be better. Break it down to game performance, multi-tasking performance, workstation application performance, office task performance, and multi-media performance if you want to go into the main categories for a desktop/workstation CPU. Servers have their own benchmarks and because of I/O differences between different chipsets and motherboards, servers need a MUCH greater degree of investigation to see if it's a CPU performance issue in an application, and a I/O issue with the motherboard/chipset.

Note through all of this that I am not bashing Intel, or putting them down to any degree except against their new performance per watt garbage. At the moment, AMD has a lead over Intel. Intel has a lot of new hype, which is expected because of their latest event, but let's keep our eyes on reality. Intel has been behind for a while in MOST benchmarks when it comes to the best performing chip in the desktop, workstation, and server environments. In laptops, Intel has an advantage in battery life, AMD has a performance edge in laptops at the expense of battery life. Intel has a LOT more sales in the laptop and desktop areas than AMD due to marketing and general lack of AMD based laptops in the retail chain stores. AMD has an edge in sales from small businesses(such as my own) and do-it-yourself computers.

Dell being Intel-only and notice advertises more than any other computer company out there is a large reason that AMD hasn't been selling to the general public better than they have been.

My own observations of the whole Intel vs. AMD battle is that AMD has been going for innovation in overall system design as well as CPU design to get the edge. Intel hasn't released anything really exciting except the Pentium-M in quite some time, and even with the latest hype they are sending out, they don't have anything NEW that is going into the retail market right now. They may have plans and things in the works, but if they take another 1-2 years from now before those new technologies are ready for the public, it will give AMD a chance to release even more new approaches to overall system design and performance. I will grant that some of the new things that Intel is moving to the CPU are interesting, but how well they work once released will remain to be seen. Voltage regulation on the CPU may mean that a CPU could die a sudden tragic death if that part were faulty, and there is NOTHING a motherboard manufacturer could do to prevent it.

So, cheer up Intelia, time will tell if you are right or wrong, but you may want to back off a bit for now until Intel is ready to release the new stuff. That water-cooled setup you are using to challenge others with could easily get it's butt kicked by almost anyone with a water-cooled Athlon 64 FX-57. Just because a 3GHz to 3GHz comparison may give it an edge doesn't mean it will win the crown when it comes to water-cooled vs. water-cooled.

 

[Targon]

Originally posted by: Technomancer
Let me point out that Conroe would ONLY be 5x faster than the Northwood (which it was compared against) ONLY at running each at a specific clock cycle, because it seems a lot of you have forgotten that power useage scales nonlinearly.

Therefore, someone claimed for argument's sake, purely theoretical, that if Conroe ran at 3 Ghz for 20 watts then it would be 15 Ghz at 100 watts.

WRONG.

It would be fantastic if Conroe did run at 15 Ghz on 100 watts, but let's see what really happens when we scale:

3 - 20
6 - 50
9 - 110
12 - 180
15 - 280

Nobody's going to run a chip that will burn 280 watts at normal load.

I just gave the previous case as a theoretical example, but it is more in line with what really happens.

As such, it would be more realistic, to compare a 3 Ghz Northwood to a 2 or 2.5 Ghz Conroe that would use less than 1/5th total power of the Northwood to get the 5-1 ratio.

I state this not to bash Intel, but to provide a more correct, if not entirely precise view of looking at what Intel's people are saying.

That said, both AMD and Intel are working on low-power chips, they both have 500 Mhz chips that consume about 5 w/hr at full load and I remember reading that the Chinese developed (or stole the design for) a similar chip. There are newer technologies, or at least technologies that are now more frequently used to reduce power consumption. For what it's worth, Elpia claimed to develop a DDR3 @ 1333 Mhz on 1.5 volts that uses what they call "dual-gate transistors" to reduce current leakage. My bet is that AMD and Intel will be looking at this technology for their processors if they aren't already.

EDIT:

Except for those things I stated about the actual developments with reducing power useage, there are other factors to consider. Intel said 5x the performance per watt, the key word is performance, not Ghz. Also, I don't think they've even given us any means by which to measure their "performance per watt" so until then any theoretical speculation, including my own will be irrelevant.

Intel may also be using multi-core in their formula to decide what increased performance per watt means. They may try to convince people that a four-core chip with each core running at 1GHz provides the performance of a 4GHz chip or something. We all know it doesn't work that way, but Intel may try to use it. A 4 core@1GHz won't require a lot of power, so will meet the demands of the Intel Hype Machine when it comes to performance per watt. But unless you run an application that takes advantage of multi-threading, it won't give you that level of performance in the real world.

When it comes to power usage, the curve is a lot sharper, more of a logarithmic curve.

500MHz at 5 watts
750MHz at 7 watts
1GHz at 11 watts
1.5GHz at 20 watts
2GHz at 30 watts
2.5GHz at 44 watts
3GHz at 60 watts
3.5GHz at 80 watts
4GHz at 120 watts


It's a part of the reason Intel had problems with the power usage of the P4 line, as you ramp up the speeds, you hit a critical point where for every quarter of a GHz you go up by 5 to 15 watts of power needed, and that causes a lot of heat. Also, all the cache that Intel was forced to add to the chip to keep up with AMD raises the power demand.

 

[Intelia]

Nice post Targon. If you read all my post you will see I figured in the DC as compared to single core. Also you Can't say that Conroe will run at 100 watts. Conroe will be 65 watts running at upon release between 2.5-30GHz at 65 watts. so figure in the dual cores as part of what Intel was talking about. The best I can see this running @ as compared to a 3.0 P4C would be a 50% increase in performance and I find those numbers 100% believable . and would be very happy with that performance increase .Now buy those numbers Conroe looks impressive.

Now here,s were I am having some problems. Intel stated also that Merom would have 3x the performance of yonah watt for watt. Both these cpu,s are low watt. . So 3x the performance perwatt increase over a dual core yonah is incrediable number. If this is true the performance of the new Intel Cpu's is going to be incrediable. The 50% better performance than P4C seems like its doable to me . But 3x merom DC compared to a Yonah DC just seems outrageous.As merom and yonah will use about the same watts . Intel seems to be talking 3X the performance of a yonah. 2x seems extreme to me. Well just have to wait and see. It also seems that right after these new 65nm. are released Intel will be releaseing the 45nm right on the tail of the 65nm tech.

 

[TuxDave]

Originally posted by: Targon

Next up, this whole "performance per watt" thing. For a LAPTOP, or where people are trying to lower their power usage due to electricity costs, then the performance per watt issue is valid. BUT, it means NOTHING when you are discussing which processor has the best performance. Either Intel or AMD can release a new version of a chip virtually unchanged, same clock rate but on a newer process(going to .65 micron from .90 micron for example) and generally get an improvement per watt. For a desktop, workstation, or server that has PERFORMANCE as the only metric that matters, performance per watt doesn't matter. So let's ignore all the hype that Intel is trying to spread when it comes to their "new" metric that they are trying to sell.

The rest of the post is ok but I highly disagree with this statement. All high performance design today MUST be power aware whether or not it's for servers, desktops or mobile. I'm sure 99% of the forum will agree (Intel fans since that is what Intel wants and AMD fans since they enjoy making fun of how much power Prescott uses). In the academic world, research has moved away from answering the question of "How fast can we make this" to "What's the fastest version of this that fits below a power envelope".

The power/performance metric is a very old concept in VLSI design and it's not something that Intel invented. Go ask any digital VLSI professor.

Sever farms need to be power aware. The processors in a server rack has a standard for maximum power usage in the processor due to its small form factor. And given thousands of servers in a room, if they can get they can them to be lower power, that's not a bad selling point. They can then buy more servers at the same electricity cost.

 

[linkgoron]

Err I'm no programer or anything, but when they say "integer perforemance per watt" how much will it really affect real world performence?

 

[JEDIYoda]

So has it been established yet whether Intelia is actually Dothan all dressed up in drag??

 

[Shenkoa]

Good question.

Intel claims everything the come up with is original.

Example (The all new patented smoke signals from Intel, the fastest and easiest way to communicate, uses super ultra smoke particles with revolutionary Night Threading so you can see them at night)

 

[Targon]

Originally posted by: TuxDave

Originally posted by: Targon

Next up, this whole "performance per watt" thing. For a LAPTOP, or where people are trying to lower their power usage due to electricity costs, then the performance per watt issue is valid. BUT, it means NOTHING when you are discussing which processor has the best performance. Either Intel or AMD can release a new version of a chip virtually unchanged, same clock rate but on a newer process(going to .65 micron from .90 micron for example) and generally get an improvement per watt. For a desktop, workstation, or server that has PERFORMANCE as the only metric that matters, performance per watt doesn't matter. So let's ignore all the hype that Intel is trying to spread when it comes to their "new" metric that they are trying to sell.

The rest of the post is ok but I highly disagree with this statement. All high performance design today MUST be power aware whether or not it's for servers, desktops or mobile. I'm sure 99% of the forum will agree (Intel fans since that is what Intel wants and AMD fans since they enjoy making fun of how much power Prescott uses). In the academic world, research has moved away from answering the question of "How fast can we make this" to "What's the fastest version of this that fits below a power envelope".

The power/performance metric is a very old concept in VLSI design and it's not something that Intel invented. Go ask any digital VLSI professor.

Sever farms need to be power aware. The processors in a server rack has a standard for maximum power usage in the processor due to its small form factor. And given thousands of servers in a room, if they can get they can them to be lower power, that's not a bad selling point. They can then buy more servers at the same electricity cost.

I said that in environments where power usage due to electicity costs was important in those environments. But to say that across the board that performance per watt should be more important than the performance of the processor fails to understand that for MOST people out there, a higher performing CPU is the reason to upgrade.

Would you go out and spend $500 for a new processor that isn't any faster than your current processor, but uses less power without overclocking comming into play? For every environment, there will be different issues that need to be addressed. If you want to buy the fastest processor available for your budget, you do NOT want to be forced to track down which processor is fastest when the metric that Intel is using to sell processors is performance/watt. If Intel can make the new processors run faster with a lower power draw, then they deserve credit. I just dislike masking the performance of a processor behind a metric that most won't care about.



I take it for granted that the engineers who design processors KNOW that keeping the power usage down is a key when it comes to making future processors at higher speeds.

 

[Targon]

Originally posted by: Intelia
Nice post Targon. If you read all my post you will see I figured in the DC as compared to single core. Also you Can't say that Conroe will run at 100 watts. Conroe will be 65 watts running at upon release between 2.5-30GHz at 65 watts. so figure in the dual cores as part of what Intel was talking about. The best I can see this running @ as compared to a 3.0 P4C would be a 50% increase in performance and I find those numbers 100% believable . and would be very happy with that performance increase .Now buy those numbers Conroe looks impressive.

Now here,s were I am having some problems. Intel stated also that Merom would have 3x the performance of yonah watt for watt. Both these cpu,s are low watt. . So 3x the performance perwatt increase over a dual core yonah is incrediable number. If this is true the performance of the new Intel Cpu's is going to be incrediable. The 50% better performance than P4C seems like its doable to me . But 3x merom DC compared to a Yonah DC just seems outrageous.As merom and yonah will use about the same watts . Intel seems to be talking 3X the performance of a yonah. 2x seems extreme to me. Well just have to wait and see. It also seems that right after these new 65nm. are released Intel will be releaseing the 45nm right on the tail of the 65nm tech.

Watt for watt. So they may be able to cut the power draw of Merom by 66% but the speed might stay the same. Or at the same wattage you get a huge boost in performance. Which will it be? I'm not putting down what engineers are doing, just what the marketing people are doing. It's also why I said I am waiting for these processors to be released. They could also tell the truth and we may see NO improvements in the normal processors if they compare processors running at 500MHz where it's easier to control power demand in a processor.

And finally, Intel is making comparisons to the P4s in most claims of improvements in the performance/watt ratings. We all know that AMD was well ahead of Intel when comparing the AMD64 processors to the Pentium 4s, so it's no wonder that the new stuff will also show a better performance/watt. I do expect Intel to have an advantage in power demand compared to AMD for quite a while once Intel releases the new chips, but how they perform will be the key.

 

[The Linuxator]

Originally posted by: Bona Fide
How about we all just wait and see the production samples before making any hasty judgements?

 

[The Pentium Guy]

Originally posted by: The Linuxator

Originally posted by: Bona Fide
How about we all just wait and see the production samples before making any hasty judgements?

:thumbsup:.

I like the way this thread is going (well, I only read the page 13) - no more heated trollish discussions. This is the way AT should be

 

[Furen]

Die, thread, die...

 

[Markfw]

Originally posted by: The Pentium Guy

Originally posted by: The Linuxator

Originally posted by: Bona Fide
How about we all just wait and see the production samples before making any hasty judgements?

:thumbsup:.

I like the way this thread is going (well, I only read the page 13) - no more heated trollish discussions. This is the way AT should be

Thats because Intelia got banned......

 

[Sentential]

Originally posted by: Intelia
Nice post Targon. If you read all my post you will see I figured in the DC as compared to single core. Also you Can't say that Conroe will run at 100 watts. Conroe will be 65 watts running at upon release between 2.5-30GHz at 65 watts. so figure in the dual cores as part of what Intel was talking about. The best I can see this running @ as compared to a 3.0 P4C would be a 50% increase in performance and I find those numbers 100% believable . and would be very happy with that performance increase .Now buy those numbers Conroe looks impressive.

Now here,s were I am having some problems. Intel stated also that Merom would have 3x the performance of yonah watt for watt. Both these cpu,s are low watt. . So 3x the performance perwatt increase over a dual core yonah is incrediable number. If this is true the performance of the new Intel Cpu's is going to be incrediable. The 50% better performance than P4C seems like its doable to me . But 3x merom DC compared to a Yonah DC just seems outrageous.As merom and yonah will use about the same watts . Intel seems to be talking 3X the performance of a yonah. 2x seems extreme to me. Well just have to wait and see. It also seems that right after these new 65nm. are released Intel will be releaseing the 45nm right on the tail of the 65nm tech.

Intelia please stop while you are ahead. You do not have any facts on this and are only giving people who like Intel a bad name. Some people here do know what they are talking about and 3ghz is not going to happen, yet.

That being said Conroe ought to clock easily past 3ghz but that wont be its stock speed by a long shot. Granted it ought to do near 3.5ghz on the stock cooler (based on estimates from known Presler/Cedar Mill overclocks)

Nor will it by 5X unless you compare it on a core-by-core basis.

Smithfield = 135W / 67W per
Conroe = 65W / 32.5W per
Yonah = 32W / 16 per

The only one that comes close to the 5X value is Yonah, not Conroe

 

[clarkey01]

Originally posted by: Intelia
Nice post Targon. If you read all my post you will see I figured in the DC as compared to single core. Also you Can't say that Conroe will run at 100 watts. Conroe will be 65 watts running at upon release between 2.5-30GHz at 65 watts. so figure in the dual cores as part of what Intel was talking about. The best I can see this running @ as compared to a 3.0 P4C would be a 50% increase in performance and I find those numbers 100% believable . and would be very happy with that performance increase .Now buy those numbers Conroe looks impressive.

Now here,s were I am having some problems. Intel stated also that Merom would have 3x the performance of yonah watt for watt. Both these cpu,s are low watt. . So 3x the performance perwatt increase over a dual core yonah is incrediable number. If this is true the performance of the new Intel Cpu's is going to be incrediable. The 50% better performance than P4C seems like its doable to me . But 3x merom DC compared to a Yonah DC just seems outrageous.As merom and yonah will use about the same watts . Intel seems to be talking 3X the performance of a yonah. 2x seems extreme to me. Well just have to wait and see. It also seems that right after these new 65nm. are released Intel will be releaseing the 45nm right on the tail of the 65nm tech.

FUD FUD FUD..

Jesus...... How the hell do you know this hay? Do you sneak outside Intel fabs with your ear to the walls?.

Im not anti Intel, really.However you seem to state all these amazing "facts" like your husband has a 4 Ghz dothan and 45nm is only just behind 65nm. I am all for Intel facts and news, hell I even posted an Intel thread today but you dont seem to contribute anything worthwhile. You'v upset (maybe a little extreme) but p*ssed some of us off for sure. Even the most die hard Intel fans feel for us.

If this was wingz or another Intel worker posting such things then it may have some credibility.

 

[Pabster]

Don't you know Intel has an uber-secret 45nm Sassaman running 5GHz per core?

Let this one die, folks.

 

[Sentential]

I can tell everyone right here right now that .45nm chips cannot be got. .65nm yes, 45? hell no. He's full of ****. In addition 4ghz isnt possible on dothan. Only one guy has gotten close and that was on LN2.

Intelia the 'facts' coming from your friend are so garbled that if I were him I'd bitch slap you for making me look like a fool. Grow up.

Btw, Merom isnt a 3X (300%) peformance improvement you tool, it was 30%. Once again you prove that you are simply a poser trying to sound cool.

 

[Markfw]

People, read. Intelia got banned, so let this thread die !!!!

 

[Sixtyfour]

Originally posted by: Intelia
Clock for clock yonah is killing AMD 64 X2

ROFLMFAO !!

Click-> Intel next generation CPU is owned by todays AMD Opteron <- Click.