Intel Q9550S: A Greener Quad Core?

jessekopelman · Post by **jessekopelman** » Wed Jun 24, 2009 10:38 am

MikeC wrote:Paying more for higher tolerance/performance parts is nothing new or abnormal. It's just the way of the world.

This is true for all kinds of technological products (as well as farmed, grown or harvested items). In test microphones, for example, a mic calibrator with a NIST certification and guaranteed 0.25 dB precision under a wide range of temp, humidity and atmosphere pressure (altitudes) might fetch $3000. A virtually identical instrument, but not NIST certified, with only 0.5 dB assured precision, and not as rigorously tested under wide conditions, might only be $1000. Most users will not pay the 200% surcharge, but there are circumstances and applications in which the pricier item is preferred.

CPUs have been binned since the very start. Most C2Ds (and A64 x2s) come off the same assembly line. Binning separates the high end parts from the mids and the lows.

Nobody's complaining about the binning, Mike. They're complaining about the huge price premium Intel is charging for the "S" chips. It should be a 10-20% price premium, not a 30+% premium. That there were so many of these chips available that they can be offered through the retail market, argues further against the high premium.

Post by **MikeC** » Wed Jun 24, 2009 10:57 am

jessekopelman wrote:Nobody's complaining about the binning, Mike. They're complaining about the huge price premium Intel is charging for the "S" chips. It should be a 10-20% price premium, not a 30+% premium. That there were so many of these chips available that they can be offered through the retail market, argues further against the high premium.

Like any good corporation, Intel charges what they figure the market will bear. For the enthusiast w/ knowhow, for those who read articles like this one, it's too much.

I'm guessing that a large percentage of the S variants go to the corporate/OEM sector where in 1000 lots, the prices are $266 vs $320 for the 2 versions of the Q9550. $56 is not that much in the context of a large lot of systems, and if it helps with "greening" -- either in reality or in perception -- for some buyers & sellers, it's worthwhile.

Olle P · Post by **Olle P** » Wed Jun 24, 2009 2:00 pm

I'm thinking about the slower versions...

The basic models should consume less power than the higher clocked 9550, right?
Then the equivalent S models should also have a proportionally smaller gap in power consumption compared to the basic models, right?
So the actual power difference in watts between Q8200 and Q8200S is smaller than the shown difference between Q9550 and Q9550S, meaning that the lower difference in retail price still result in a long payback time.

Cheers
Olle

Monkeh16 · Post by **Monkeh16** » Wed Jun 24, 2009 2:54 pm

rpsgc wrote:
Monkeh16 wrote:Don't forget the CPUs are binned. A Q9650 will run faster than a Q9550, a QX will run a hell of a lot faster than either.
Uh... what?!

Perhaps it runs faster because it has a higher clock speed...?! A QX is nothing more than a Q with an unlocked multiplier (and higher clock speed as well where applicable). I don't know where you got those crazy ideas from.

... I'm talking about maximum clock here, not stock.

bradc · Post by **bradc** » Sun Jul 05, 2009 11:48 am

Sorry to bump this up from the dead, but I have to ask, why not test using a clamp meter over the 12v line going to the cpu. Virtually every motherboard with a P4 power connector (4pin or 8pin) will get it's power for the cpu through that power connector and nowhere else.

While a test like this obviously lets us compare the standard versus the S, we don't actually know how much power either cpu on it's own draws which makes comparisons to AMD's a bit difficult.

The 95W TDP is very conservative on the normal ones. In fact I'd suspect the normal CPU is in the 65-75w region, while the S model is around 50-60w.

If I'm honest, this is the only thing that annoys me about the entire spcr review process! Now isn't that great praise

jessekopelman · Post by **jessekopelman** » Mon Jul 06, 2009 3:38 pm

bradc wrote: While a test like this obviously lets us compare the standard versus the S, we don't actually know how much power either cpu on it's own draws which makes comparisons to AMD's a bit difficult.

But it doesn't actually matter whether the AMD or Intel draws more power in and of itself, since you can't use either without a MB and you can't use exactly the same MB for both. So, the only power consumption figures that matter for deciding Intel vs AMD are CPU+MB. This why it's unfortunate that SPCR has the means/opportunity to test so few MB. It would be nice to have a better idea what the best power consumption MB was for any given application. The good news is that there is a good amount of reporting going on in the forums -- would be an interesting project for someone to mine the data a put together some tables.

Cov · Post by **Cov** » Thu Jul 09, 2009 6:08 pm

Hi, I wanna pair up a Mini iTX board with the Q9550, so I wrote an email to Zotac, asking whether they plan in one of their future BIOS updates, to implement the undervolting / underclocking feature.

They replied, saying that this is not going to happen to "ensure a stable running system."
I think that was a lame excuse since SPCR already showed that 1.024V (CPU-Z) / 1.03125V (BIOS) is ok for the that CPU.

bradc · Post by **bradc** » Fri Jul 10, 2009 12:45 am

I know what you are saying (it is an obvious point and I already realised that

) but for OCD people like myself it would be a nice thing to have.

jessekopelman wrote: But it doesn't actually matter whether the AMD or Intel draws more power in and of itself, since you can't use either without a MB and you can't use exactly the same MB for both. So, the only power consumption figures that matter for deciding Intel vs AMD are CPU+MB. This why it's unfortunate that SPCR has the means/opportunity to test so few MB. It would be nice to have a better idea what the best power consumption MB was for any given application. The good news is that there is a good amount of reporting going on in the forums -- would be an interesting project for someone to mine the data a put together some tables.

Chrissicom · Post by **Chrissicom** » Thu Sep 03, 2009 10:52 am

Really great article

I am just building some energy efficient system and this article quite helped me a lot in a decision!

I just find it a pity that the conclusion is, the processor costs 115 USD more and in energy price it takes 12 years to redeem that. Using energy efficient components and protecting the environment is not just about saving money. The CPU requires ~80 kWh less power within a year. Put that in a small data center with 1000 of these CPUs and I save 80000 kWh which is a considerable amount of CO2. We also use green energy which actually costs 27 cents / kWh (excluding basic fees)

the US average really isn't comparable here to the European average, because there is more coal, nuclear and oil power produced.

Anyway, I know you might not get your money back, but you save the planet

Another thing I find strange is that these things can't be undervolted to below 1V. I am running a few Core i7 920s @ 0.992V with Full Load and HT enabled, TM disabled on a Asus P6T WS Pro (Prime95 Stress Tested). Some of them even go to 0.89V @ Full Load but I have no time testing each individual CPU and I found 0.992V (1V Bios Setting) to be working on any i7 920.