Core i7 Reviews

[Idontcare]

Originally posted by: sunnn
also i failed to find power consumption for other tests. maybe i just missed them, can you kindly point me to them?

Here's the link to the data I used from Anand's recent article: http://www.anandtech.com/cpuch...howdoc.aspx?i=3453&p=3

Originally posted by: sunnn
just to caution you on throwing numbers out there. i965 data is hardly a representative of the whole nehalem family. tests were done with turbo on. WITH TURBO OFF , i920 power consumption is hardly (5-10%?) improved compared to q9450.
lastly, i965 are being marketed towards people with multi-gpu's, people who overclock the hell out of their cpu's, performance/watt metric is the least of their concerns.

data taken from nehalem:dark knight page 9:
cpu.......................q9450..................i920..........imrovement...i965-over-q9770
crysis...................0.154fps/w.........0.144fps/w.......-6.9%.......15%
pov-ray 3.7.........13.32pps/w.......17.44pps/w.......31.1%.......53%
cinebench xcpu...65.6cbm/w.........77.71cbm/w......18.6%.......32%
x264 hd...............0.37fps.w...........0.42fpw/w........13.5%.......38%
edit: a bit off topic:
its been generally accepted that in gaming, phenom is 15% and 20% slower compared to kentsfield and yorksfield clock-for-clock. so far, out of the 5 games anand tested here (9950be vs. q9450), the average difference is 19.5%.

If you read my posts above I was thoroughly disheartened by Nehalem's power consumption results as reported in Anand's original article. Then he published the follow-up article (I linked above) wherein he delves into the power-consumption versus yields situation at Intel:

I published an abridged version of these results in the review, basically showing that the Core i7-965 offered much better power consumption, across the board, than the equivalently clocked QX9770 while the Core i7-920 was outshined by the Q9450 which drew less total system power. Both datapoints were valid but there were too many unanswered questions to draw any serious conclusions at that point.

http://www.anandtech.com/cpuch...howdoc.aspx?i=3453&p=3

Note that the idle power on the i7-965 is very low, one thing that must be enabled to achieve this is the QPI power management option in the X58 BIOS which for whatever reason was disabled by default in our original review.

http://www.anandtech.com/cpuch...howdoc.aspx?i=3453&p=3

All three of Intel's Core i7 CPUs that will be available at launch this month are 130W TDP parts. At 3.2GHz that's expected, but at 2.66GHz that's a bit high compared to Intel's other quad-core 2.66GHz processors on the market. The Core 2 Quad Q9450, for example, has a 95W TDP and runs at 2.66GHz.

The lower TDP is made possible by a lower core voltage, which is enabled by the fact that Intel has been building quad-core Penryns for a while and yields are high enough where driving core voltage down is possible. The same will eventually happen to the Core i7, but it's such a new design, such a radical departure from Intel's previous Core based CPUs and so early in the manufacturing process that there simply hasn't been time to get yields high enough to produce < 100W TDP 2.66GHz parts.

http://www.anandtech.com/cpuch...howdoc.aspx?i=3453&p=3

I confirmed that I didn't have a particularly low power Core i7-965 by testing multiple chips, and Intel confirmed that my QX9770 fell within the middle of its distribution for power characteristics of all QX9770s. It looks extremely probably that at the same TDP level, Nehalem has the ability to be much more power efficient than even Penryn - all without so much as a die shrink, remember that both of these CPUs are built on the same 45nm process.

http://www.anandtech.com/cpuch...howdoc.aspx?i=3453&p=3

So that pretty much settles it. 920 power consumption is going to be somewhat of an anomoly for a while as this is effectively the "garbage bin" for all the silicon that is functional at 2.66GHz given enough voltage and allowed a high TDP (130W, same as the 3.2GHz flagship). But if you want to know where the Nehalem family itself is headed as the junk SKU slowly gets optimized then you should really look at the power-consumption performance of the flagship product.

Once I read Anand's follow-up review and analyzed his data I finally see the light on Nehalem's power consumption reduction capabilities. Sure initially the 920 does have a 130W TDP and is clocked below the 965...so performance/watt can't be the same for the 920 as the 965, yes?

In the meantime if we are trying to figure out where the power savings of the PCU and static cmos come in then we need to analyze performance data that are not being skewed by marketing and sales dept desire to sell every hobbled i7 that comes out the fab. (which was my goal here, not trying to make a case for why folks on a budget should buy the 920 despite power-consumption results being skewed to the highside thanks to all that junk silicon being pushed thru the SKU at higher Vcc for yield reasons)

It makes perfect sense to me, but then again I worked many years in such an environment so I freely admit it might not make much sense to an industry outsider.

 

[sunnn]

Originally posted by: Idontcare

Originally posted by: sunnn
also i failed to find power consumption for other tests. maybe i just missed them, can you kindly point me to them?

Here's the link to the data I used from Anand's recent article: http://www.anandtech.com/cpuch...howdoc.aspx?i=3453&p=3

Originally posted by: sunnn
just to caution you on throwing numbers out there. i965 data is hardly a representative of the whole nehalem family. tests were done with turbo on. WITH TURBO OFF , i920 power consumption is hardly (5-10%?) improved compared to q9450.
lastly, i965 are being marketed towards people with multi-gpu's, people who overclock the hell out of their cpu's, performance/watt metric is the least of their concerns.

data taken from nehalem:dark knight page 9:
cpu.......................q9450..................i920..........imrovement...i965-over-q9770
crysis...................0.154fps/w.........0.144fps/w.......-6.9%.......15%
pov-ray 3.7.........13.32pps/w.......17.44pps/w.......31.1%.......53%
cinebench xcpu...65.6cbm/w.........77.71cbm/w......18.6%.......32%
x264 hd...............0.37fps.w...........0.42fpw/w........13.5%.......38%
edit: a bit off topic:
its been generally accepted that in gaming, phenom is 15% and 20% slower compared to kentsfield and yorksfield clock-for-clock. so far, out of the 5 games anand tested here (9950be vs. q9450), the average difference is 19.5%.

If you read my posts above I was thoroughly disheartened by Nehalem's power consumption results as reported in Anand's original article. Then he published the follow-up article (I linked above) wherein he delves into the power-consumption versus yields situation at Intel:

I published an abridged version of these results in the review, basically showing that the Core i7-965 offered much better power consumption, across the board, than the equivalently clocked QX9770 while the Core i7-920 was outshined by the Q9450 which drew less total system power. Both datapoints were valid but there were too many unanswered questions to draw any serious conclusions at that point.

http://www.anandtech.com/cpuch...howdoc.aspx?i=3453&p=3

Note that the idle power on the i7-965 is very low, one thing that must be enabled to achieve this is the QPI power management option in the X58 BIOS which for whatever reason was disabled by default in our original review.

http://www.anandtech.com/cpuch...howdoc.aspx?i=3453&p=3

All three of Intel's Core i7 CPUs that will be available at launch this month are 130W TDP parts. At 3.2GHz that's expected, but at 2.66GHz that's a bit high compared to Intel's other quad-core 2.66GHz processors on the market. The Core 2 Quad Q9450, for example, has a 95W TDP and runs at 2.66GHz.

The lower TDP is made possible by a lower core voltage, which is enabled by the fact that Intel has been building quad-core Penryns for a while and yields are high enough where driving core voltage down is possible. The same will eventually happen to the Core i7, but it's such a new design, such a radical departure from Intel's previous Core based CPUs and so early in the manufacturing process that there simply hasn't been time to get yields high enough to produce < 100W TDP 2.66GHz parts.

http://www.anandtech.com/cpuch...howdoc.aspx?i=3453&p=3

I confirmed that I didn't have a particularly low power Core i7-965 by testing multiple chips, and Intel confirmed that my QX9770 fell within the middle of its distribution for power characteristics of all QX9770s. It looks extremely probably that at the same TDP level, Nehalem has the ability to be much more power efficient than even Penryn - all without so much as a die shrink, remember that both of these CPUs are built on the same 45nm process.

http://www.anandtech.com/cpuch...howdoc.aspx?i=3453&p=3

So that pretty much settles it. 920 power consumption is going to be somewhat of an anomoly for a while as this is effectively the "garbage bin" for all the silicon that is functional at 2.66GHz given enough voltage and allowed a high TDP (130W, same as the 3.2GHz flagship). But if you want to know where the Nehalem family itself is headed as the junk SKU slowly gets optimized then you should really look at the power-consumption performance of the flagship product.

Once I read Anand's follow-up review and analyzed his data I finally see the light on Nehalem's power consumption reduction capabilities. Sure initially the 920 does have a 130W TDP and is clocked below the 965...so performance/watt can't be the same for the 920 as the 965, yes?

In the meantime if we are trying to figure out where the power savings of the PCU and static cmos come in then we need to analyze performance data that are not being skewed by marketing and sales dept desire to sell every hobbled i7 that comes out the fab. (which was my goal here, not trying to make a case for why folks on a budget should buy the 920 despite power-consumption results being skewed to the highside thanks to all that junk silicon being pushed thru the SKU at higher Vcc for yield reasons)

It makes perfect sense to me, but then again I worked many years in such an environment so I freely admit it might not make much sense to an industry outsider.

made more sense. however, i would like to point out, until these numbers on i920 are improved, we will just take them as they are (5 to 10% improvement on power efficiency over q9450).
also, i double checked other numbers on q9770 and i965. here's what i got:
POV-Ray5 PPS/Watt.....................53%
Cinebench (1 thread)...................31%
Cinebench ..................................32%
3dsmax 9 SPECapc CPU..............33%
x264 HD Encode Test..................38%
DivX 6.8.3....................................16%
Windows Media Encoder.............0%......."un-game" ave..29%(turbo on)
Age of Conan..............................31%
Race Driver GRID.........................13%
Crysis..........................................14%
FarCry 2......................................31%
Fallout 3......................................44% .......game ave.....26.6%(turbo on)

i split power efficiency improvement, games vs. synthetic improvements. again, these numbers/improvements were done with turbo on. with TURBO OFF, i965~20% power efficiency improvement and ~5% absolute performance improvement over q9770. now when you consider the cost to migrate to the new platform, it's a hard call. so, 30-40% power efficiency improvement is very very generous. i would be very cautious.

note: you might want to check the numbers in divx, windows media encoder, 3dsmax, race driver, crysis. if i misread, miscalculated anything, feel free as well.
also,i like how you stick with the facts on your responses. very admirable.

 

[IntelUser2000]

I don't know enough about the technical details of any of that to add further value by commenting but it continues to beg the question "what the hell was Intel doing at IDF bothering everyone with these slides and graphs"?

Why do you care?? Its interesting and you never know, the power consumption might have been higher without the static logic. We don't know since there is no Nehalem with full domino logic or Penryn with majority of static logic. They always talk about numbers in tech conferences which are otherwise meaningless to average Joe's. Example like the transistor Ion/Ioff numbers, or even the transistor count/die size.

I wouldn't be surprised to see Intel launch the notebook version as a dual-core without HT enabled. When HT kicks in and works like it was designed power consumption goes up like crazy (more cores are utilized fully). On a desktop that's not a problem but on a laptop that eats battery time fast.

That'd not make sense because Nehalem on a laptop will be made for performance/watt, not purely low power consumption. Its where HT will achieve it as the performance/watt will soar.

EDIT: Nehalem will be better off power wise on the laptop because of the new chipset. The Ibexpeak PCH(Platform Controller Hub) will be much more power efficient. If you look at the CPU level power consumption, Nehalem is at worst, equal to Core 2: http://www.tomshardware.com/re...7-Nehalem,2057-10.html

Most people idle their laptops or close to it(web browsing), the low idle power will benefit power consumption.

 

[Idontcare]

Originally posted by: IntelUser2000

I don't know enough about the technical details of any of that to add further value by commenting but it continues to beg the question "what the hell was Intel doing at IDF bothering everyone with these slides and graphs"?

Why do you care?? Its interesting and you never know, the power consumption might have been higher without the static logic. We don't know since there is no Nehalem with full domino logic or Penryn with majority of static logic. They always talk about numbers in tech conferences which are otherwise meaningless to average Joe's. Example like the transistor Ion/Ioff numbers, or even the transistor count/die size.

I happen to work in the industry, so the technical side of these things happens to tickle my fancy.

I've been the guy on many occasions who had a 100 slide presentation (the iceberg) but VP's approve me publicly showing just one slide out of the heap (the tip of the iceberg) and that one slide ends up being meaningless, pointless, and otherwise worthless when presented in the absence of the other "kept strictly confidential" slides in the presentation package. And I know I am no one special, absolutely everyone in the industry who interfaces with the public domain in some capacity finds themselves similarly gagged at all time.

My suspicion is this is what happened to static CMOS presentation but it is fun to tease the intel guys on the forums as they too can't discuss publicly what they know privately which (to my pleasure) renders them incapable of defending their company's honor against my tiresome and witless comments on the few things that are public

As an enthusiast though, when people want my money its always nice to think I know when I am getting taken advantage of even if knowing it doesn't change my buying habits one iota.

But I happen to enjoy sarcasm even more, even if it comes at the expense of my own humility, hence the repeated denigration of my own attributes in many of my posts (at least I like to think it turns out that way ).

 

[Idontcare]

Originally posted by: sunnn
made more sense. however, i would like to point out, until these numbers on i920 are improved, we will just take them as they are (5 to 10% improvement on power efficiency over q9450).

Yeah I have no issue with that, never did. It all comes down to what you intend to use the numbers for. If you are trying to compare say architectural performances in a broad-sense then it is just sensible to avoid using data from a known "polluted" clockspeed SKU.

But if you are a potential consumer of that same clockspeed SKU then it is worth your time to analyse the data there too so you know what you are getting into when you buy a 920 today with current steppings, etc. I totally agree with that point of view.

Originally posted by: sunnn
note: you might want to check the numbers in divx, windows media encoder, 3dsmax, race driver, crysis. if i misread, miscalculated anything, feel free as well.

Originally posted by: sunnn
also, i double checked other numbers on q9770 and i965. here's what i got:
...
3dsmax 9 SPECapc CPU..............33%

QX9770 scored 13.1 using 220.1 Watts/hour. That comes out to 13.1/220.1 = 0.060 units/watt.

i7-965scored 17.6 using 209.4 Watts/hour. That comes out to 17.6/209.4 = 0.084 units/watt.

100%*(0.084 - 0.060)/0.060 = 41% by my math. (the i7 is 41% more power efficient than the QX9770 on 3dsmax 9 SPECapc CPU)

Not sure how you got 33%, is my maths in error here?

Originally posted by: sunnn
DivX 6.8.3....................................16%
Windows Media Encoder.............0%......."un-game" ave..29%(turbo on)

What I did here was actually calculate how many watts the CPU used for the duration of the test since the metric for these tests was "time to completion" and not an actual benchmark score per say.

For example the QX9770 required 221.7 watts/hour while doing the DivX test, and took 42.4 seconds to complete the test.

This means the QX9770 consumed 2.61 Watts to complete the DivX test.

The i7 965 required 202.1 watts/hour during the test, and required only 32.8seconds to complete the test, which works out to mean the i7 965 consumed 1.84 watts to perform the DivX test.

So 100%*(2.61-1.84)/2.61 = 29% ... meaning the i7 is 29% more energy efficient than the QX9770 when it comes to the DivX test.

By similar methodology the QX9770 used 2.01 Watts for the Media Encoder test while the 965 used 1.34W...a 33% improvement in energy efficiency over the QX9770.

Using these numbers, the "un-game" average increases to 37% by my calcs.

Originally posted by: sunnn
Race Driver GRID.........................13%
Crysis..........................................14%
FarCry 2......................................31%
Fallout 3......................................44% .......game ave.....26.6%(turbo on)

I am not sure why all our numbers are close, but off by a couple percent in all these cases.

Farcry2 for example. QX9770 uses 324.2 watts/hour at 102.6 fps. Thats 0.3165 fps/watt for QX9770.

i7 965 uses 271.9 watts/hour at 115.1 fps. Thats 0.4233 fps/watt for the 965.

100%*(0.4233 - 0.3165)/0.3165 = 33.8% (965 improvement over the QX9770)

Maybe you are dividing by the 965's numbers instead of the QX9770 numbers? That would make your numbers mean "energy efficiency deficit of QX9770 over the 965" which is similar but not the same thing as what I am talking about.

The difference would be the same as saying "the QX9770 uses XX% more power than the 965 for these tests" versus saying "the 965 uses XX% less power than the QX9770 for these tests".

In those two statements the value for XX% is not the same value because you would change the denominator to reflect the fact you are using one system as the reference relative to the other (and your reference system is what you are changing in those two statements)

I get the average "game" efficiency for the 965 over the QX9770 as being 28%.

Originally posted by: sunnn
i split power efficiency improvement, games vs. synthetic improvements. again, these numbers/improvements were done with turbo on. with TURBO OFF, i965~20% power efficiency improvement and ~5% absolute performance improvement over q9770.

I consider Turbo On to be a designed-in feature of the hardware and it does increase power consumption (higher frequency) so its not like it comes for free or does any slight of the hand stuff to the performance/watt data.

As I'm interested in operating my system in a performance/watt maximized mode, the only reason I'd care to know the results of turning off Turbo mode is if they made the results better...as then I'd turn off turbo mode on my system for my apps.

Originally posted by: sunnn
now when you consider the cost to migrate to the new platform, it's a hard call. so, 30-40% power efficiency improvement is very very generous. i would be very cautious.

I think we both realize we are discussing the nitpicking of the details at this point. The cost to migrate is always present no matter the decade or the computer system involved, Bloomfield adds nothing new to this aspect of the end-user's decision point. Clearly the cost analysis must begin with the user's current system.

Migrating from a P4 or K7 prolly isn't much of an argument.

Migrating from a QX9770 to an i7 965 can not be supported by power reduction numbers alone.

As for 30-40 versus any other number...the fact that I used a 10% range to describe my "impressions" from Anand's short-list of benchmarks ought to provide the reader with ample information as to how "hard" my impression of the numbers are.

10% of 30% is 1/3 the range. I would hope no one is making purchasing decisions off of "idontcare's estimation of nehalem's power efficiency is 30-40% over penryn" comments.

To me there is nothing to be cautious about, were the maths to work out that the hard numbers from 100's of benchmarks resulted in the 965 being a mere 20% more efficient over the QX9770 it really wouldn't change my overall impression of the Nehalem's improvements in power consumption at this point.

Were I a consultant to the IT dept at Fortune 500 companies then you can bet I would be more interested in narrowing that range down from "30-40%" to a real number representative of the client's apps and operating conditions (which would then make such "real" numbers useless and meaningless for any other company or person on the planet).

Originally posted by: sunnn
also,i like how you stick with the facts on your responses. very admirable.

The US presidential elections are over so I've got nothing else to talk about but the facts these days