Or at least could you slower the multiplier keeping 400 FSB and doing the inverse droping fsb increasing the multiplier, in order to have the same cpu frequency but a differnt fsb on board to see if it's really related to the cpu?
[quote]
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.
[/quote]
quote: The performance gains are impressive, given that we're only looking at a 6.6% increase in clock speed. The 1600MHz FSB does seem to do a bit, giving us 7 - 8% performance boosts in a couple of instances.
Was Anand being facetious with this conclusion? A 6.6% increase in clockspeed gives a 7-8% increase in performance? That means the 1600MHz FSB did almost nothing to improve performance! We're talking a gain of 1% due to the FSB increase? Not to mention that 1% is probably within the margin of error for those benchmarks! What gives?
It seems like that kind of FSB to start off with is not worth it. It seems to me that Intel is actually trying to curtail the vast overclockability of it's lineup in the future.
I imagine they would love to reestablish some reason to spend $600+ on a CPU. Seems like starting at a low factory FSB makes overclocking too easy.
I feel more and more so every passing day that I made a big mistake by not getting a Q6600 when they came out. If the word on the street is right, you can get those up to a steady 3.0Ghz for under three bills. That performance at that price is going to be hard to beat even by Intel it looks like.
It does not explain, however, the increased power usage. In any case, the HSF mounting mechanisms on Intel's chips since the P4 have been subpar at best. First, they post warnings about how they can bend but not damage motherboards, then they switch to the current scheme. AMD's system since the K8 is simple: provide proper bracing on and under the mb and you are set.
Chips can have varying levels of leakage current from batch to batch or even within the same batch. The difference between the lowest leakage chip vs. highest leakage chip can be quite large. Chips with high leakage are usually faster than chips with lower leakage. The power draw is probably due to high leakage because of the large increase in idle power usage of QX9770 vs QX9650 (58W). Subtracting 58W from QX9770's full load power consumption yields 209W. This is 9% more than QX9650's power consumption under full load. This is a more reasonable difference in power consumption given the increase in clock speed and FSB. Leakage also increases at higher temperatures. A very fast part on the borderline of acceptable leakage may fail at lower temps than a slow chip with more leakage margin.
Could you do some double check the CPU voltage?
How the power consuming grow up so much for just 200Mhz increase?
(It’s 4 x 200Mhz) but even so.
Is Intel getting the same P3 1.133MHZ bug? Where the Pentium 3 couldn’t be clock more than 1.000MHZ? But this time hit at 3.2GHZ.
Or is this CPU manufactured in some other less "qualified" Intel factory?
The voltage is definitely hig. Default should be around 1.25v, and that's for the 65nm Q6600. I see 1.336v being reported. That's where the power consumption is coming from.
I think there was some other issue going on here that should have been looked into before posting the article. Most reviews of the 3ghz/1333fsb model are clocking up to 4ghz on air - now we are saying the 3.2 has issues with heat? come on, get real.
Obviously it's true, since they did measure a significant increase in power consumption which has to end up as heat. Granted some comes from the higher FSB, a very small amount only!
I will speculate that one of two things is true:
1) This sample needed higher voltage to remain stable than what they eventually plan on selling through channels. They'll release their spec sheet and then we'll see, but really the numbers don't look too outlandish given the clock speed and quad cores, if the vcore is raised a bit.
2) The motherboard VRM subcircuit isn't able to handle this current level so well and excessive ripple causes higher consumption.
What would be interesting is to use the better heatsink and lower the vcore a notch to see if that regains a bit of stability, or of course to try it on more motherboards.
it shouldnt be that much hotter - people are running kentsfield at way higher than that on air - many approaching 500mhz bus. 400 is achievable on basic cheap motherboards with minimal cooling solutions and has been for over 1 year.
quote: Could somebody explain why after only 2 years after the death of netburst and its super high power envelope, we are stuck in the same situation again?
no one is stuck with that situation. you can't even pre-order this chip and it obviously not a retail part. also it wasn't netburst's "high power envelope" that was the problem, it was the actual high power draw of the prescott core that was the problem.
So going from 1333Mhz FSB to 1600Mhz FSB gives an increase in 58W at idle and 75W at full load...
Then maybe it’s better not release 1600 FSB cpus at all
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
27 Comments
Back to Article
cVEcOC - Monday, December 3, 2007 - link
hello for Overclocking this new cpu i can up to 4.5ghz like to QX9650 at 1600FSB?nemrod - Wednesday, November 21, 2007 - link
I really don't understand. You've check X48 chipset:http://www.anandtech.com/cpuchipsets/showdoc.aspx?...">http://www.anandtech.com/cpuchipsets/showdoc.aspx?...
with a Intel Core 2 Quad QX9650
You have had to problem to run the cpu at 4GHz 400MHz x 10
And now you say that a cpu at 400MHz x 8 is not stable? And you assume extra power is due to the cpu...
Have you try to measure the power of the cpu? (like there for example:
http://www.matbe.com/articles/lire/521/intel-qx965...">http://www.matbe.com/articles/lire/521/...45nm-12-... not the QX9770 yet on this site)
Or at least could you slower the multiplier keeping 400 FSB and doing the inverse droping fsb increasing the multiplier, in order to have the same cpu frequency but a differnt fsb on board to see if it's really related to the cpu?
nemrod - Wednesday, November 21, 2007 - link
qx9650http://www.matbe.com/articles/lire/521/intel-qx965...">http://www.matbe.com/articles/lire/521/...45nm-12-...
CPU power measure
idle: 12W
full: 63.6W
so a change of 52.6W
http://enthusiast.hardocp.com/article.html?art=MTQ...">http://enthusiast.hardocp.com/article.html?art=MTQ...
power measure behind the psu, with "small system"
idle:82W
full:135W
so a change of 53W
http://techreport.com/articles.x/13470/15">http://techreport.com/articles.x/13470/15
idle: 160W
full: 215W
so a change of 55W
But in your mesurement
idle:151W
full:192
a change of only 41W!
nemrod - Wednesday, November 21, 2007 - link
http://www.sharkyextreme.com/hardware/cpu/article....">http://www.sharkyextreme.com/hardware/cpu/article....power load of QX6770 below QX6700 and QX6850
On this one same kind of result than you
http://techgage.com/print/intel_core_2_extreme_qx9...">http://techgage.com/print/intel_core_2_extreme_qx9...
look like there is some differencies between the ES
nemrod - Wednesday, November 21, 2007 - link
or motherboards...[quote]
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.
[/quote]
magreen - Tuesday, November 20, 2007 - link
Was Anand being facetious with this conclusion? A 6.6% increase in clockspeed gives a 7-8% increase in performance? That means the 1600MHz FSB did almost nothing to improve performance! We're talking a gain of 1% due to the FSB increase? Not to mention that 1% is probably within the margin of error for those benchmarks! What gives?
Toferman - Tuesday, November 20, 2007 - link
Guys,The CPUZ screenshot shows too much Voltage for stock speed. Change it to Default, and your temps shouldn't be so high.
:)
Tofer
gochichi - Tuesday, November 20, 2007 - link
It seems like that kind of FSB to start off with is not worth it. It seems to me that Intel is actually trying to curtail the vast overclockability of it's lineup in the future.I imagine they would love to reestablish some reason to spend $600+ on a CPU. Seems like starting at a low factory FSB makes overclocking too easy.
I feel more and more so every passing day that I made a big mistake by not getting a Q6600 when they came out. If the word on the street is right, you can get those up to a steady 3.0Ghz for under three bills. That performance at that price is going to be hard to beat even by Intel it looks like.
supremelaw - Monday, November 19, 2007 - link
http://www.supremelaw.org/systems/heatsinks/warnin...">http://www.supremelaw.org/systems/heatsinks/warnin...Sincerely yours,
/s/ Paul Andrew Mitchell
Webmaster, Supreme Law Library
http://www.supremelaw.org/">http://www.supremelaw.org/
Etern205 - Monday, November 19, 2007 - link
You must study law, well then good for you.Just don't abuse it!
supremelaw - Monday, November 19, 2007 - link
http://www.anandtech.com/cpuchipsets/intel/showdoc...">http://www.anandtech.com/cpuchipsets/intel/showdoc...
"... switching to a newer Intel retail heatsink/fan
fixed my issues well enough for me to run through all the tests."
Yep! Same 4 defective fasteners:
http://www.anandtech.com/cpuchipsets/intel/showdoc...">http://www.anandtech.com/cpuchipsets/intel/showdoc...
Sincerely yours,
/s/ Paul Andrew Mitchell
Webmaster, Supreme Law Library
http://www.supremelaw.org/">http://www.supremelaw.org/
sprockkets - Tuesday, November 20, 2007 - link
It does not explain, however, the increased power usage. In any case, the HSF mounting mechanisms on Intel's chips since the P4 have been subpar at best. First, they post warnings about how they can bend but not damage motherboards, then they switch to the current scheme. AMD's system since the K8 is simple: provide proper bracing on and under the mb and you are set.jones377 - Monday, November 19, 2007 - link
?acx - Monday, November 19, 2007 - link
Chips can have varying levels of leakage current from batch to batch or even within the same batch. The difference between the lowest leakage chip vs. highest leakage chip can be quite large. Chips with high leakage are usually faster than chips with lower leakage. The power draw is probably due to high leakage because of the large increase in idle power usage of QX9770 vs QX9650 (58W). Subtracting 58W from QX9770's full load power consumption yields 209W. This is 9% more than QX9650's power consumption under full load. This is a more reasonable difference in power consumption given the increase in clock speed and FSB. Leakage also increases at higher temperatures. A very fast part on the borderline of acceptable leakage may fail at lower temps than a slow chip with more leakage margin.MrKaz - Monday, November 19, 2007 - link
Could you do some double check the CPU voltage?How the power consuming grow up so much for just 200Mhz increase?
(It’s 4 x 200Mhz) but even so.
Is Intel getting the same P3 1.133MHZ bug? Where the Pentium 3 couldn’t be clock more than 1.000MHZ? But this time hit at 3.2GHZ.
Or is this CPU manufactured in some other less "qualified" Intel factory?
Could you under clock it to make sure?
swaaye - Tuesday, November 20, 2007 - link
The voltage is definitely hig. Default should be around 1.25v, and that's for the 65nm Q6600. I see 1.336v being reported. That's where the power consumption is coming from.xsilver - Monday, November 19, 2007 - link
Could somebody explain why after only 2 years after the death of netburst and its super high power envelope, we are stuck in the same situation again?Also is there anything being done for future processors to rectify this problem? (besides die shrink that is)
retrospooty - Monday, November 19, 2007 - link
I think there was some other issue going on here that should have been looked into before posting the article. Most reviews of the 3ghz/1333fsb model are clocking up to 4ghz on air - now we are saying the 3.2 has issues with heat? come on, get real.mindless1 - Monday, November 19, 2007 - link
Obviously it's true, since they did measure a significant increase in power consumption which has to end up as heat. Granted some comes from the higher FSB, a very small amount only!I will speculate that one of two things is true:
1) This sample needed higher voltage to remain stable than what they eventually plan on selling through channels. They'll release their spec sheet and then we'll see, but really the numbers don't look too outlandish given the clock speed and quad cores, if the vcore is raised a bit.
2) The motherboard VRM subcircuit isn't able to handle this current level so well and excessive ripple causes higher consumption.
What would be interesting is to use the better heatsink and lower the vcore a notch to see if that regains a bit of stability, or of course to try it on more motherboards.
murphyslabrat - Wednesday, November 21, 2007 - link
OMG, can we only hope?!?If Intel responded this quickly, with a shoddy product to boot, could this mean that they are afraid of Barcelona/Agena?
retrospooty - Monday, November 19, 2007 - link
I meant to say "Most reviews of the 3ghz/1333fsb model are clocking up to 4ghz on air -" including the one right here at Anandtech..."Our unlocked QX9650 had no problems hitting 333MHz x 12.0, for a final clock speed of 4.0GHz"
http://www.anandtech.com/cpuchipsets/intel/showdoc...">http://www.anandtech.com/cpuchipsets/intel/showdoc...
Whats the deal?
JarredWalton - Monday, November 19, 2007 - link
12x333 is quite different from 8x400. That might be the problem.nemrod - Wednesday, November 21, 2007 - link
but they have done 10 x 400 on the X48 testretrospooty - Monday, November 19, 2007 - link
it shouldnt be that much hotter - people are running kentsfield at way higher than that on air - many approaching 500mhz bus. 400 is achievable on basic cheap motherboards with minimal cooling solutions and has been for over 1 year.semo - Monday, November 19, 2007 - link
no one is stuck with that situation. you can't even pre-order this chip and it obviously not a retail part. also it wasn't netburst's "high power envelope" that was the problem, it was the actual high power draw of the prescott core that was the problem.
JarredWalton - Monday, November 19, 2007 - link
The chipset and mobo have to run at higher clocks as well. Maybe that's the problem?MrKaz - Monday, November 19, 2007 - link
So going from 1333Mhz FSB to 1600Mhz FSB gives an increase in 58W at idle and 75W at full load...Then maybe it’s better not release 1600 FSB cpus at all