Bloomfield comes in 130W TDP flavor?

[Rammsteiner]

It was rather interesting to notice that all Nehalem chips have the same 130W TDP. Either this is a typo in Intel's official documents that we've seen or it means that 3.2GHz which is 600MHz faster than 2.6GHz Nehalem will have the same TDP.

We've seen the same numbers before, which might indicate that these are the real numbers. We know that even 2.93GHz Nehalem can easily hit over 4GHz with air-cooling which indicates that there is still enough room for some overclocking in this chip.

Nehalem is probably the best chip that Intel ever engineered even if its a heavily improved Core 2 architecture, but it looks that it works well. November should be the time when you should be able to buy one.
Source (http://www.fudzilla.com/index.php?option=com_content&task=view&id=9371&Itemid=1)

Weird that 2.6Ghz even has 130W TDP, and that on 45nm high/k and metal gates?:shocked: Or doesnt Bloomfield have high/k and metal gates?:shrug:

I think 2.6Ghz as usually wont be close to 130W TDP and the higher clocked one might be near/exceed it. Will have to see, if it's true in the first place of course.

[Shintai]

Alot got 130W TDP but not even using half that. Or 65W TDP and uses 25-30W.

130W TDP is also to allow automatic OC up to 266Mhz I guess.

And this is also a 130W TDP chip....

http://images.anandtech.com/graphs/penrynoc_12040751228/16135.png

But its the same principle of why a 3.00Ghz dualcore and why a 1.6Ghz dualcore is both 65W...

And no, it will never go over 130W. never ever and then...just never. Unless its OCed.

Also we already have tested the powerconsumption. And thats on prototype boards and so on. And the X58 chipset being HOT.

http://images.anandtech.com/graphs/nehalempreview_060508030043/17022.png

FUD just saw a 130W TDP and mixed it together with his epic journalist talents.

[PetNorth]

Source (http://www.fudzilla.com/index.php?option=com_content&task=view&id=9371&Itemid=1)

Weird that 2.6Ghz even has 130W TDP, and that on 45nm high/k and metal gates?:shocked: Or doesnt Bloomfield have high/k and metal gates?:shrug:

I think 2.6Ghz as usually wont be close to 130W TDP and the higher clocked one might be near/exceed it. Will have to see, if it's true in the first place of course.

yeah, it seems Nehalem is power hungry. OBR has tested it, and thats what he stated here http://www.xtremesystems.org/forums/showpost.php?p=3263689&postcount=1196

[Mechromancer]

Source (http://www.fudzilla.com/index.php?option=com_content&task=view&id=9371&Itemid=1)

Weird that 2.6Ghz even has 130W TDP, and that on 45nm high/k and metal gates?:shocked: Or doesnt Bloomfield have high/k and metal gates?:shrug:

I think 2.6Ghz as usually wont be close to 130W TDP and the higher clocked one might be near/exceed it. Will have to see, if it's true in the first place of course.

WTF is strange about 2.6Ghz and 130w TDP? Nehalem has an integrated memory controller and all 4 cores on a single die. You Intel boys are just getting to the point where AMD's Phenom has been for a while now. Memory controllers use more power and yall need to understand that. I hope now you can appreciate what AMD has done with a 65nm 2.6Ghz 9950 at only 125w. I can't wait to see what they'll do with 45nm and high-k metal gates later on in 2009.

TDP won't mean much to yall after you see how Nehalem performs so stop making a big deal out of nothing.

EDIT: Rammsteiner, you're definitely not an Intel boy so my statements don't really apply to you.

[bowman]

130W rating for Bloomfield has been known for a very long time. Fudzilla at it again reporting old info as 'news'.

[Extelleron]

Intel's TDP is such a vast overstatement of the actual power consumption, even under the worst conditions, that it is almost meaningless. Intel lumps processors with the same TDP regardless of power consumption; i.e. the E8400 and E6850 have the same TDP despite the E8400 consuming much less, and consuming maybe 30W under full load.

I'd bet that a 130W 2.66GHz Bloomfield consumes, at most, 70-80W under full load.

[ryboto]

Intel's TDP is such a vast overstatement of the actual power consumption, even under the worst conditions, that it is almost meaningless. Intel lumps processors with the same TDP regardless of power consumption; i.e. the E8400 and E6850 have the same TDP despite the E8400 consuming much less, and consuming maybe 30W under full load.

I'd bet that a 130W 2.66GHz Bloomfield consumes, at most, 70-80W under full load.

the 130W 9770 draws over 110W as tested by LostCircuits, so for the high end, its not as much an overstatement. Besides Nehalem isnt Penryn, you cant expect it to be exctly the same in TDP labeling.

[Rammsteiner]

Alot got 130W TDP but not even using half that. Or 65W TDP and uses 25-30W.

130W TDP is also to allow automatic OC up to 266Mhz I guess.

And this is also a 130W TDP chip....

http://images.anandtech.com/graphs/penrynoc_12040751228/16135.png

But its the same principle of why a 3.00Ghz dualcore and why a 1.6Ghz dualcore is both 65W...

And no, it will never go over 130W. never ever and then...just never. Unless its OCed.

Also we already have tested the powerconsumption. And thats on prototype boards and so on. And the X58 chipset being HOT.

http://images.anandtech.com/graphs/nehalempreview_060508030043/17022.png

FUD just saw a 130W TDP and mixed it together with his epic journalist talents.
Heh, Shintai. First off your post seems to be like the counter PetNorth's post.

Second, you ignored my links for the 2nd time in some other thread regarding Intel's TDP (where Extelleron might want to look in as well apparently). I dont know really, but somehow I cant believe your statement that Bloomfield's TDP would be 'much lower', nor that 'you believe 130W TDP includes the 266Mhz boost'. Also AMD 7-series run hot while consuming almost nothing, so a hot running chipset does not automatically mean it uses a lot of power;).

I know Bloomfield is basicly what Phenom is now, a lot more integrated stuff and this would result in a lot more hungry chip. Although, as Mechro noted, this shows that Phenom is by far not all that bad and I think with Bloomfield already being at 45nm and high/k and metal gates (now is that right or not?:shrug:), it's quite weird for such a production process? I think with later revisions and/or with the move to 32nm it will become better since it's Intel's first, well, 'Phenom-like' CPU to say it that way. We've seen that happen with the Core 2 line already.

Then again, Bloomfield is also a CPU aimed at servers and dont get me wrong, but the last thing huge server/database owners are looking for is to OC their components, be it an easy 'one click' thing or not. Following that, it would be nothing less than stupid to rate it 130W TDP for any chip because it 'might include the 266Mhz boost'. This is from what I understand out of Shintai's post btw, because the part below seems a lot more reliable and truth;

I might be completely retarded for this so correct me if Im wrong, it seems Bloomfield's reference clock is 133Mhz right? So 266Mhz would basicly double it's clock speed, although be it for only a few cores and not all to increase single threaded performance. Increased clockspeed for a few cores sounds reasonable, but a 100% reference boost? I think the multiplier will be lowered a lot then? Cause 3.2Ghz quad running one core at 6.4Ghz sound just wrong:p:

Does anyone have some more info about that Turbo boost though? Cause if that 266Mhz is right, then it cant be at the same MP. So what clock increase might we see from that? I tried to google, but well, I cant really find a lot besides the things I posted already. Also Ive had some booze so maybe my google skills drop a lot from it:p:

Thx in advance:up:

[Donnie27]

And reasonable they are; for a 20 - 50% increase in performance, total system power consumption only went up by 10%.

So much for the FUD LOL!

[metro.cl]

Heh, Shintai. First off your post seems to be like the counter PetNorth's post.

Second, you ignored my links for the 2nd time in some other thread regarding Intel's TDP (where Extelleron might want to look in as well apparently). I dont know really, but somehow I cant believe your statement that Bloomfield's TDP would be 'much lower', nor that 'you believe 130W TDP includes the 266Mhz boost'. Also AMD 7-series run hot while consuming almost nothing, so a hot running chipset does not automatically mean it uses a lot of power;).

I know Bloomfield is basicly what Phenom is now, a lot more integrated stuff and this would result in a lot more hungry chip. Although, as Mechro noted, this shows that Phenom is by far not all that bad and I think with Bloomfield already being at 45nm and high/k and metal gates (now is that right or not?:shrug:), it's quite weird for such a production process? I think with later revisions and/or with the move to 32nm it will become better since it's Intel's first, well, 'Phenom-like' CPU to say it that way. We've seen that happen with the Core 2 line already.

Then again, Bloomfield is also a CPU aimed at servers and dont get me wrong, but the last thing huge server/database owners are looking for is to OC their components, be it an easy 'one click' thing or not. Following that, it would be nothing less than stupid to rate it 130W TDP for any chip because it 'might include the 266Mhz boost'. This is from what I understand out of Shintai's post btw, because the part below seems a lot more reliable and truth;

I might be completely retarded for this so correct me if Im wrong, it seems Bloomfield's reference clock is 133Mhz right? So 266Mhz would basicly double it's clock speed, although be it for only a few cores and not all to increase single threaded performance. Increased clockspeed for a few cores sounds reasonable, but a 100% reference boost? I think the multiplier will be lowered a lot then? Cause 3.2Ghz quad running one core at 6.4Ghz sound just wrong:p:

Does anyone have some more info about that Turbo boost though? Cause if that 266Mhz is right, then it cant be at the same MP. So what clock increase might we see from that? I tried to google, but well, I cant really find a lot besides the things I posted already. Also Ive had some booze so maybe my google skills drop a lot from it:p:

Thx in advance:up:


Nope turbo mode will allow nehalem to boost 266MHz in its final speed not the refeernce clock. Example the 3.2GHz part will be able to go up to 3.46GHz if needed

[Rammsteiner]

So much for the FUD LOL!
Somehow I did not even find 20% IPC improvement in all benchmarks. In quite a few benchmarks Bloomfield did awesome, but not all and also not awesome.

And so much for Anandtech, PetNorth's post shows other stuff and so does the 130W TDP statement. Only 10% power usage, against what? QX9775 x2?

To go a tad OT and to be used as a little side note only to 'underline' Mechro's post, K10 did hardly add power consumption to K8's fastest CPU and yet it gained IPC wise improvements upto IIRC 40%.

Nope turbo mode will allow nehalem to boost 266MHz in its final speed not the refeernce clock. Example the 3.2GHz part will be able to go up to 3.46GHz if needed
266Mhz in total!? Then Shintai might have a real valid point for even noting this since what's 266Mhz all about. Almost any CPU can gain that on stock Volts, although Ive to say it's nice Intel included this as an automatic feature. Then again it only takes 3 seconds in BIOS messing around with P-states and you've the very same results:confused: Although as I said, server/database owners might not be interested in messing around in the BIOS though;)

Anyway, thx for explaining. Somehow I failed hard at finding that:p:.

[deathman20]

Isn't Intel's TDP more or less worst case senerio if the CPU is at said speed, with the maximum rated voltage that they state?

TDP is always overated as how I understood it. They say it is that but really ends up being quiet a bit less. Especially in like the C2D Case with a E2k series and a E8k series having the same TDP even though clearly the later would use be using more power due to its speed, and cache size. What I've been starting to use and more less its guessing and using charts that people saying for load. 85% of the TDP fro the highest speed chip in that series is what it probably uses typically. So slower speeds will use less, and you can guesstimate using that wattage/speed/voltage equation that floats around.

[metro.cl]

Somehow I did not even find 20% IPC improvement in all benchmarks. In quite a few benchmarks Bloomfield did awesome, but not all and also not awesome.

And so much for Anandtech, PetNorth's post shows other stuff and so does the 130W TDP statement. Only 10% power usage, against what? QX9775 x2?

To go a tad OT and to be used as a little side note only to 'underline' Mechro's post, K10 did hardly add power consumption to K8's fastest CPU and yet it gained IPC wise improvements upto IIRC 40%.


266Mhz in total!? Then Shintai might have a real valid point for even noting this since what's 266Mhz all about. Almost any CPU can gain that on stock Volts, although Ive to say it's nice Intel included this as an automatic feature. Then again it only takes 3 seconds in BIOS messing around with P-states and you've the very same results:confused: Although as I said, server/database owners might not be interested in messing around in the BIOS though;)

Anyway, thx for explaining. Somehow I failed hard at finding that:p:.

Also the 266MHz boost depends on idle cores, you dont get 4x266MHz it only "overclocks" 1 core 266MHz if the software is single threaded and then you get a perf boost. So TDP doenst go up

[Donnie27]

Somehow I did not even find 20% IPC improvement in all benchmarks. In quite a few benchmarks Bloomfield did awesome, but not all and also not awesome.

And so much for Anandtech, PetNorth's post shows other stuff and so does the 130W TDP statement. Only 10% power usage, against what? QX9775 x2?

To go a tad OT and to be used as a little side note only to 'underline' Mechro's post, K10 did hardly add power consumption to K8's fastest CPU and yet it gained IPC wise improvements upto IIRC 40%.

Anyway, thx for explaining. Somehow I failed hard at finding that:p:.

Originally Posted by Anandtech
And reasonable they are; for a 20 - 50% increase in performance, total system power consumption only went up by 10%.

Its stated as an "average" not an absolute. 10% after adding Northbridge and "NO" PCI-E for Bloomfield is hardly a "Show stopper" or a negative. It's still all good:up:

Later steppings will improve the process as they (AMD and Intel) always have. Didn't take long for Current E0 to improve on C0 now didn't it? Posters have already said they'd wait for the second stepping anyway.

[savantu]

The level of BS in this thread is astounding.

Folks , relax for a second and then repeat after me : Intel uses family TDPs.Bloomfield targets enthusiasts and all CPUs for that segment will have a 130w TDP irrespective of actual power per bin.

It has always been like this ; all XE CPUs have ~130w rating , from QX6700 to QX9770.Only an idiot would think they burn the same power.That 130w TDP is meant to show what kind of power delivery and cooling you need.

Instead of issuing dozens of different specs ( as many bins there are ) Intel supplies typically 3 : 65w , 95w and 130w.Motherboard designers build boards according to this specs ; so do the cooling people.Instead of having different coolers for a 2.66GHz ,2.93GHz and a 3.2GHz Nehalems you have only 1 , designed to dissipate at least 130w , regardless of what the actual power consumption of a specific bin is.

If somebody bothers to dig in Intel's tech docs they will find this explained :

http://download.intel.com/design/processor/datashts/318726.pdf

Page 76 , note 3 : "This table shows the maximum TDP for a given frequency range. Individual processors
may have a lower TDP."

[Orthogonal]

Savantu is correct, and to add a little color to his statement. Intel always reserves the highest TDP bin for the top tier parts regardless of their actual power consumption. Bloomfield is an Enthusiast platform that is designed to be overclocked, therefore, Intel puts all chips in the highest TDP rating to ensure that OEM's and Mobo manufacturer's design a thermal solution to support it. This will become even more important now that Nehalem offers "Turbo Mode" allowing chips to self overclock up to the TDP when under heavy load.

[Helmore]

Originally Posted by Anandtech
And reasonable they are; for a 20 - 50% increase in performance, total system power consumption only went up by 10%.

Its stated as an "average" not an absolute. 10% after adding Northbridge and PCI-E for Bloomfield hardly a "Show stopper" or a negative. It's still all good:up:

Later steppings will improve the process as they (AMD and Intel) always have. Didn't take long for Current E0 to improve on C0 now didn't it? Posters have already said they'd wait for the second stepping anyway.

Bloomfield does not have integrated PCIe, that will be on Havendale or whatever it was and those chips will see the light of day within 1 year from now.

[Donnie27]

Bloomfield does not have integrated PCIe, that will be on Havendale or whatever it was and those chips will see the light of day within 1 year from now.

I meant it didn't, my Bad!:yepp: The others will and have it and IGP. You're right though, who'll wait on that?

[JumpingJack]

Rammsteiner -- write out 100 times -- "TPD is not the same as actual consumption." Then chant it to yourself before you go to sleep...

Savantu nailed it in this case....

[JumpingJack]

I might be completely retarded for this so correct me if Im wrong, it seems Bloomfield's reference clock is 133Mhz right? So 266Mhz would basicly double it's clock speed, although be it for only a few cores and not all to increase single threaded performance.

Thx in advance:up:

Well, I don't think you are retarded ... but you are wrong.

Reference clock is indeed 133 Mhz, just as for AMD the reference clock is 200 Mhz, and for today's 45 nm Intel quads it is 333 Mhz.

The clock speed is determined by the PLL and the phase lock on that reference clock (you will understand this as a multiplier). So at 2.66 GHz, Intel's multiplier is 20, when turbo mode kicks for the 1 or 2 used cores, it will bump this multiplier up as high as 22 (or 2 over it's stock multiplier), hence the speed will jump 266 (133 x 2) MHz to 2.93 GHz.

This is also probably a good reason to place the thermal envelop requirement high, in order to enable good clocking on 1 or 2 cores and boost single threaded performance significantly.

I always wondered why a 130 W QX9650 was coming in in the 70-80 W range, Intel's 45 nm process is easily capable of specing out a 95 W quad at 3.2 GHz. It would appear that they want their OEM infrastructure and cooling solutions build to support this new feature for the top bins.

[ryboto]

I always wondered why a 130 W QX9650 was coming in in the 70-80 W range, Intel's 45 nm process is easily capable of specing out a 95 W quad at 3.2 GHz. It would appear that they want their OEM infrastructure and cooling solutions build to support this new feature for the top bins.

They *might* be able to do that with their new stepping, but look here,
Lostcircuits (http://www.lostcircuits.com/cpu/amd_phenom9350/6.shtml)

The QX9770 is above 110W. If we consider the VRM to be ~85% efficient, that gives you 96W, so it wouldn't quite fit into the envelope, then again, we have no data on VRM efficiency, but the discussion on LostCircuits mentions they're quite efficient.

[Flambo]

Please let's not turns this into another Intel vs. AMD thread. There are enough of those already.

[JumpingJack]

They *might* be able to do that with their new stepping, but look here,
Lostcircuits (http://www.lostcircuits.com/cpu/amd_phenom9350/6.shtml)

The QX9770 is above 110W. If we consider the VRM to be ~85% efficient, that gives you 96W, so it wouldn't quite fit into the envelope, then again, we have no data on VRM efficiency, but the discussion on LostCircuits mentions they're quite efficient.

What do you mean they might? This has been demonstrated at many different sites. The did do it with the old stepping, there are multiple sites which measured at the socket QX9650 and got ~1/2 the rated TDP type numbers. The 9770 is an odd beast, also the only on with a 1600 MHz FSB supported in the package. The same site you quote also meaured as low as 64.8 W for the QX9650 at 3.0 GHz, he later tested the QX9650 @ 3.2 Ghz and got in the mid 70's as I recall.

But I agree with you to an extent, there are conditions where it is just under the cutoff, but due to margin -- both Intel and AMD bump it up to the next power bin.

It is important to separate the concept of TDP vs actual consumption and realize TDP refers to the rate of power dissipation of the cooling solution and not the actual consumption of the processor. Both AMD and Intel must spec TDP above worst case conditions, otherwise, OEMs would extract money from them when systems failed for incorrect specs. nVidia is feeling this pain right now.

The other problem is how is the power measured ... there is so much variability....

Here is digit-life measuring 86 W for a QX9770: http://www.digit-life.com/articles2/cpu/intel-c2x-qx9770-page1.html

Anandtech noted weirdly high powers and got information from Intel that the BIOS was setting VIDs too high:
Updated: We're working with Intel on the source of the thermal problems we mentioned in this review, it looks like the culprit was our ASUS P5E3 Deluxe motherboard. ASUS has since released an updated BIOS intended to address the power consumption issues we faced, you can read more about it here while we continue with our testing.
http://www.anandtech.com/showdoc.aspx?i=3154&p=1

Frankly, I am not too sure I trust the LC 9770 numbers, especially if there were BIOS problems as reported at Anand's when the processor was first launched.

[ryboto]

Frankly, I am not too sure I trust the LC 9770 numbers, especially if there were BIOS problems as reported at Anand's when the processor was first launched.

I didn't say anything about a QX9650 at 3.2ghz, if he did test that, and the power was as low as you're suggesting, that's great. I'm just posting the info he put on his site. As for the reliability of the numbers, the guy seems to know a great deal about power testing, look into his testing methodology, or the forums for some examples. I agree the 9770 looks off, but that's one of the few sites that reports the power, maybe Intel is giving them a higher VID to make sure they're stable at 3.2? Unless of course someone here with an ammeter and a 9770 wants to test it for us.

[loutsos]

I think that people are paying a lot more attention to fudzilla than they should. :shakes:

[JumpingJack]

I didn't say anything about a QX9650 at 3.2ghz, if he did test that, and the power was as low as you're suggesting, that's great. I'm just posting the info he put on his site. As for the reliability of the numbers, the guy seems to know a great deal about power testing, look into his testing methodology, or the forums for some examples. I agree the 9770 looks off, but that's one of the few sites that reports the power, maybe Intel is giving them a higher VID to make sure they're stable at 3.2? Unless of course someone here with an ammeter and a 9770 wants to test it for us.

Well you really cannot use a direct current measurement on the higher wattage CPUs except, perhaps/probably, with a toroid.

Most people shunt it.

Anyway, does it matter re 9770 or 9650 or what you said ... a yorky at 3.2 Ghz is a yorky at 3.2 Ghz, they are from the same architcture same process. When you see this discrepancy, you ask the question. Why? Anand seems to indicate that the first review samples shipped with BIOS that were incorrect, not that the VID was too high, rather the BIOS was forcing the voltage too high (VID and BIOS set voltages are different, I mis-spoke above). There will also be some higher consumption related to the higher IO FSB clock speed so I was not expecting 9770 to do better than 9650.

Digit-life produced their results a month or two after launch (check the dates) which may include a more updated BIOS (we are both speculating)... so with Anand's comments and the digit-life measurement, I tend to find the 86 W number more in line with expected.

What I can do is verify roughly, the 64.8 W measurement for the QX9650, and less than 75 W at 3.2 Ghz... I can also get to 3.8 Ghz on stock voltage on the QX9650 and this is on C0 silicon. Considering the headroom on these chips, I don't think Intel 'needs' the higher VID to get it stable as I can easily go 3 bins higher without touching the voltage.

http://www.lostcircuits.com/cpu/intel_yorkfield9770/4.shtml He did it here, at 3.2 he measured 78 W so I was off a bit in my recollection.

Jack

[Macadamia]

The level of BS in this thread is astounding.

Folks , relax for a second and then repeat after me : Intel uses family TDPs.Bloomfield targets enthusiasts and all CPUs for that segment will have a 130w TDP irrespective of actual power per bin.


Bull. X6800 = 75W.

If their bins were consistent they would brand it 95W since obviously the 2.66Ghz is going to take about 60-70.

[gOJDO]

Somehow I did not even find 20% IPC improvement in all benchmarks. In quite a few benchmarks Bloomfield did awesome, but not all and also not awesome. In some benchmarks(mostly ALU dependent) the IPC gain is below 20%, but in average there is going to be above 20% for desktop and a lot more for multi socket servers.

And so much for Anandtech, PetNorth's post shows other stuff and so does the 130W TDP statement. Only 10% power usage, against what? QX9775 x2?It seems that you and PetNorth don't understand what TDP is. It has nothing to do with the power consumption.

To go a tad OT and to be used as a little side note only to 'underline' Mechro's post, K10 did hardly add power consumption to K8's fastest CPUAlthough the fastest K8(6400+) is a 90nm CPU, the Phenom 9850 is consuming at least 20% more.
http://www.anandtech.com/showdoc.aspx?i=3272&p=13
A 2.5GHz Brisbane consumes half of the Phenom 9850 power consumption.

and yet it gained IPC wise improvements upto IIRC 40%.Can you point at any app where Phenom has 40% higher IPC than K8?
BTW, performance = IPC * frequency; energy efficiency = performance/power consumption

I know Bloomfield is basicly what Phenom is now:rofl:

yeah, it seems Nehalem is power hungry. OBR has tested it, and thats what he stated here http://www.xtremesystems.org/forums/showpost.php?p=3263689&postcount=1196TDP has nothing to do with power consumption. Also, the chip that OBR was using is an ES, so you can't make a conclusion about the retail chips.

[Movieman]

Gentlemen;
I always think in generalities and it gets me the working info I need for my projects.
Say just for the hell of it that all the nehalems are 130w chips.
That gives you a ballpark area for your planned builds whether on air,water or a chunk of ice sitting on the MB..:D
130w is controllable on top air easily.
I had 150w X5482's here and although tough I handled them on air.
The X5470's I have now are 120w and OC'd to 3750 I can run at 100% load under 50C on air..
This settles it in my opinion.
Knowing what is needed to run the chips is all that matters, not whether one is 10TDP higher than another.

[savantu]

Bull. X6800 = 75W.

If their bins were consistent they would brand it 95W since obviously the 2.66Ghz is going to take about 60-70.

That's the exception from the rule.Even in Netburst time , all XE had 130w TDP or a bit more.
That's less relevant anyway.

The point is Intel uses 130 for high performance ; 95w for mainstream QC ( all Q9000 have 95w TDP ) and 65w for DC.

[Epsilon84]

Somehow I did not even find 20% IPC improvement in all benchmarks. In quite a few benchmarks Bloomfield did awesome, but not all and also not awesome.

Using Anandtech's preview, which was early stepping + single channel RAM btw...

DivX 6.8 - 27.7% faster
x264 - 44.3% faster
3dsmax 9 - 42.4% faster
CB10 MT - 20.6% faster
CB10 ST - 2.9% faster
POV-Ray - 36.4% faster

So basically anything multithreaded Nehalem buries Penryn.

As for the 130W TDP, seriously Fuddo has no clue what it even means... as others have pointed out many times already in this thread, TDP does NOT equal power consumption. Its more a thermal guideline for OEMs to adhere to when designing cooling solutions.

[Donnie27]

Gentlemen;
I always think in generalities and it gets me the working info I need for my projects.
Say just for the hell of it that all the nehalems are 130w chips.
That gives you a ballpark area for your planned builds whether on air,water or a chunk of ice sitting on the MB..:D
130w is controllable on top air easily.
I had 150w X5482's here and although tough I handled them on air.
The X5470's I have now are 120w and OC'd to 3750 I can run at 100% load under 50C on air..
This settles it in my opinion.
Knowing what is needed to run the chips is all that matters, not whether one is 10TDP higher than another.

Certain folks will have proof you provided here but pretend you never posted it.

[Rammsteiner]

! Too many people to quote (can do but I dont feel like making a 10 page post).

Anyhow, JJ and sav, thx for all your explanations. It makes a whole lot clear.

Also I do know that TDP is for the cooling and motherboards and not specially for power consumption, but used for other things, no idea. Ive to admit I confused TDP and power consumption a bit earlier on though:shocked:

Donnie, like who? I think it's quite obvious movieman posted there:shrug: Also, let me make a thingy clear, Ive nothing against power consumption and/or TDP. If I had I should have commited suicide, it's plain hypocrit to care and yet have a PC in the first place. And if the TDP doesnt suit someone, then dont bother:up:

A few remarks though, Epsilon, there were a few previews earlier and they didnt show all 20% either. I know it's not a lot, but then dont call 'between 20~50%'. But we all know those were ES's etc, so whether this is going to change I dont know, think it will.

gOJDO, I dont know what program that was anymore:(. Could have been Barcelona too to be honest. But it was upto 40% as in their statement, but it was only one application though:ROTF:. But hey, marketing is marketing;). Also, Bloomfield being like K10 was based on native quad core and the seperate NB/IMC etc;). Not performance wise, thought that would be clear but looks like it's not:rolleyes:.