SPCR

Look!
The LP models ending with a -s are the new 65 W parts.

Sweet stuff. I'm looking foreward to see the Q9550s benchmarked against the Q9550, both power and performance-wise. I wonder if the managed to make a real improvement in the power: performance ratio?

Blue_Sky wrote:I wonder if the managed to make a real improvement in the power: performance ratio?

If they run at the same Vcore, I doubt it. Maybe some small difference.

Overpriced, undervolted CPU's are usually not that interesting, you might as well undervolt a cheaper standard model instead.
The QX9650 uses 65 W according to this site.

That's too bad. I was really hoping that it was a new revision that dropped 5+ watts off load, or something useful like that. Just using up any undervolting headroom isn't helping at all.

Too bad people are still paying a big premium for virtualization technology on the quads.

Blue_Sky wrote:That's too bad. I was really hoping that it was a new revision that dropped 5+ watts off load, or something useful like that. Just using up any undervolting headroom isn't helping at all.

Don't say that, I was just speculating!

Isn't it nice that many quads come from 95W TDP to a whopping 65W TDP? That is only two thirds of the original power! What's wrong with that, really? Or did the 95W TDP quads already perform way below their limit?

juamez wrote:Or did the 95W TDP quads already perform way below their limit?

Yes, did you see my link about the QX9650?
It's rated for 130 W, but it uses less than 65 W.

Mats wrote:

juamez wrote:Or did the 95W TDP quads already perform way below their limit?

Yes, did you see my link about the QX9650?
It's rated for 130 W, but it uses less than 65 W.

I checked your link and it seems that the "load" the cpu's are enduring in that test isn't really full load. According to some of my previous research (viewtopic.php?p=426538&highlight=#426538) Tom's Hardware states that the brisbane BE-2350 pulls 35W at load, while in your link it merely draws 30W. Also a lot of other cpu's are pulling less out of the wall than I would expect. I reckon that test isn't really showing the extremes of power usage that can and sometimes will be the case while using the tested system(s).

I don't know how credible Tom's is in testing load power consumption of cpu's, and certainly not relative to your source, but with this conflicting evidence, I would like to see more results about load power consumption involving enough cpu's to be able to deduct the possible flaws that creep into those tests, like I just did now.

edit_1:
According to this review the Yorkfield cpu's are really low power, so I'm not arguing that. I'm just stating that your "hard" numbers may be a bit off.

It is also weird that Intel rates the QX9650 as 130W TDP while current batches of them draw only half of it under load. Keeping in mind how Intel rates TDP (less conservative than AMD), it is kind of odd to see this. Maybe Intel is just being lazy by not willing to change the specs (TDP) according to what actually comes out of the wafers nowadays instead of what they got in the very beginning?

juamez wrote:I checked your link and it seems that the "load" the cpu's are enduring in that test isn't really full load. According to some of my previous research (viewtopic.php?p=426538&highlight=#426538) Tom's Hardware states that the brisbane BE-2350 pulls 35W at load, while in your link it merely draws 30W. Also a lot of other cpu's are pulling less out of the wall than I would expect. I reckon that test isn't really showing the extremes of power usage that can and sometimes will be the case while using the tested system(s).

A difference of 5 W is totally normal.
That's not really any evidence, there are many factors that affects the power consumption:

1 - Power consumption varies between individual CPU's.
2 - Power consumption varies between different motherboards.
3 - There are different ways to measure power consumption.

Lostcircuit measures before the VRM, meaning the actual power draw of the CPU is at least 10 % lower. If you think that's low, then check out the QX9770 in the same table.

juamez wrote:I don't know how credible Tom's is in testing load power consumption of cpu's, and certainly not relative to your source, but with this conflicting evidence, I would like to see more results about load power consumption involving enough cpu's to be able to deduct the possible flaws that creep into those tests, like I just did now.

There's no point in trying to compare the results, when the methods are different.

juamez wrote:It is also weird that Intel rates the QX9650 as 130W TDP while current batches of them draw only half of it under load. Keeping in mind how Intel rates TDP (less conservative than AMD), it is kind of odd to see this. Maybe Intel is just being lazy by not willing to change the specs (TDP) according to what actually comes out of the wafers nowadays instead of what they got in the very beginning?

AFAIK, neither AMD nor Intel changes the TDP's in that way, so I don't understand what you think is odd about that?

There seems to be a lot of people who still thinks that TDP is an exact number which is valid for every individual CPU with the same TDP, when in fact it's just maximum TDP.
A CPU with a TDP of 130 W can have a power draw that's much less than 130 W.

My guess is the LP quads are just a bin sort at final test...after running scads of parts, Intel probably found they can get enough of a yield to offer these and get a price premium.

I was more excited about the $57 e5200

those new quads are all much more expensive than the old versions though

CA_Steve wrote:I was more excited about the $57 e5200

Reads like 67$ to me... Still that's indeed quite a nice cpu. You don't even get SSE4 with that though (need E7xxx series at least), nor VT (need E8xxx for that, so triple the price).
Those new quads sure come at a high premium. AMD's low voltage parts (for instance the 45W series X2) only had like a 20% price premium or so. Those 40% or so indeed seems a bit high for something you could likely achieve (minus validation) on the standard parts with a little less voltage...

Apparently - I was too excited when I typed it