First siting - 45nm Quad
- Thread starter P5-133XL
- Start date
ddrueding
Fixture
Awesome. What clock range are they hoping for in this series?
udaman
Wannabe Storage Freak
- Joined
- Sep 20, 2006
- Messages
- 1,209
http://www.intel.com/products/processor/core2quad/specifications.htm
Also at Tigerdirect, same price
http://www.tigerdirect.com/applicat...EdpNo=3584440&sku=CP2-DUO-Q9300&CMP=ILC-FPM05
Not very awesome, IMO. Xeon L5420 just announced at similar price range ($380/1000 OEM) does the same with 2x cache, and nearly 1/2 the TDP @50w.
http://www.intel.com/pressroom/archive/releases/20080325comp.htm?iid=pr1_releasepri_20080325m
Stereodude
Not really a
And I only just got my Q6600. 
ddrueding
Fixture
With a voltage bump my month-old Q6600 is quite happy @ 3.5Ghz. But I do feel your pain, Stereodude.
The problem with Xeon processors are FB-RAM (very hot, slow, and costly) combined with Xeon server motherboards ($$$) and thusly, the system cost will be far higher. Then there is that the energy cost for FB-RAM will be far higher than the energy savings of the low wattage processor. FB-RAM really sucks big ones, unless you need lots and lots of ram and then you just have to suck it up.
Stereodude
Not really a
Are you sure the Xeon requires FB-RAM? Some of the Xeons are basically identical to the standard Core 2 chips. Fro example, the E8400 and the Xeon E3110 are basically the same chip and you can run the E3110 with standard DDR2.
Fushigi
Storage Is My Life
Other than heat/power, I don't know that the 9300 will be different from the the 6600. The 9300 is .1 GHz faster and has the faster bus but it only has 75% of the cache of the 6600. I'm guessing it's close to a wash.
Stereodude
Not really a
Yes, the L5420 requires FB-RAM. It has a LGA771 pin out (which is Xeon requiring FB-RAM). The LGA775 pin out is the one that allows standard DDR2/DDR3 RAM.
I will universally agree, that the Q9300 is better than the Q6600.
The issue brought up by Udaman was that he thought the Xeon L5420 was better than the Q9300 and with that I disagree. The Xeon has a bigger cache and a lower power envelope. However, it costs more, and will require a costly server motherboard. The Xeon also requires FB-RAM which is very slow, costs much more than DDR2, and uses up power like mad.
The issue brought up by Udaman was that he thought the Xeon L5420 was better than the Q9300 and with that I disagree. The Xeon has a bigger cache and a lower power envelope. However, it costs more, and will require a costly server motherboard. The Xeon also requires FB-RAM which is very slow, costs much more than DDR2, and uses up power like mad.
udaman
Wannabe Storage Freak
- Joined
- Sep 20, 2006
- Messages
- 1,209
Care to post up some facts to back your claims? "Uses up power like mad", wouldn't make any sense to have a 50w CPU that had 45w RAM. Q9300 has a TDP of 95 watts. Sorry, it can't come close to the L5420- 50w is not merely an easily brushed aside *minor* reduction, "lower power envelope"; it is almost ONE HALF the amount of power. Bigger cache = higher performance for same clock, probably could out perform Q9300 w/ slightly lower cost, faster DDR2. I'll take a nearly 50% power consumption reduction, over minor speed bump in clock speed...any day, even if it costs a little more! Want to take a guess as to which one will OC more, rendering 'slow' FB-RAM moot?
BTW, DDR3 is more power efficient than DDR2, even if not currently up to DDR2 performance levels. As a self-admitted 'tree hugger', have to say I'm disappointed at such extreme cheapness. After all, we are not talking about Core Extreme quad 45nm CPU's which cost ~3x as much, suck power like the old space heater CPU's.
Chewy509
Wotty wot wot.
Xeons can use normal DDR2 as well as FB-DIMMs, it all depends on the chipset on the mainboard.
Xeons paired with the 5000, or 5400 series chipsets require FB-DIMMs, however when paired with the 5100 or the 3000/3200 series chipsets normal Reg ECC DDR2 DIMMs are all that's required.
eg. http://www.supermicro.com/products/motherboard/Xeon1333/5100/X7DCA-3.cfm vs http://www.supermicro.com/products/motherboard/Xeon1333/5000X/X7DAE+.cfm
Now back OT, with the 9300 being a die shrink of the 6600 and most people are getting 3GHz+ from the 6600, how far will the 9300 go when overclocked?
Fushigi
Storage Is My Life
And what would be a recommended motherboard for a Q9300-based system? Do current 775 boards need a BIOS update for the new CPU?
Stereodude
Not really a
Some will, some won't. The real question is when you get a board that doesn't support the chip until you flash the BIOS, how do you flash the BIOS unless you already have another CPU to use during the flash?
Fushigi
Storage Is My Life
Well I already have 3 Q6600s so I could just bring down one of those machines. But I agree with the chicken & egg dilemma. I suppose worst case you buy the cheapest 775 Celeron, like this, and use it for the build. Then either return it or leave it on the shelf as a spare in case something blows up.
ddrueding
Fixture
Wow, that is a cheap chip.
Here is a system review comparing two low power systems: a Xeon vs. an Opteron system which breaks down the power usage of different components. What they were trying to do was compare the low-power offerings of the two companies. Specificly, for my purposes the two machines have 8 GB of different RAM ( 667MHz FB-RAM vs DDR2-667 RAM). Note that the FB-RAM is using aprox 70W vs 8W or 862% more power than the DDR2 at idle.
My point is that yes the low power l5420 Xeon uses less power than the Q9300. However, that power savings is totally used up by the required FB-RAM producing a higher total system power usage.
Now going into more detail concerning FB-RAM's power usage:Micron FB-RAM technotes. It appears to be quite complicated to determine the amount of power will be used. Variables include the ram BW, the number of channels, the number of slots used, even which slots are used. However, it will always be significantly higher than simple DDR2/3 because of the advanced memory buffer (AMB) which uses quite a bit of power in addition to the DRAM chips.
Here is a system review comparing two low power systems: a Xeon vs. an Opteron system which breaks down the power usage of different components. What they were trying to do was compare the low-power offerings of the two companies. Specificly, for my purposes the two machines have 8 GB of different RAM ( 667MHz FB-RAM vs DDR2-667 RAM). Note that the FB-RAM is using aprox 70W vs 8W or 862% more power than the DDR2 at idle.
My point is that yes the low power l5420 Xeon uses less power than the Q9300. However, that power savings is totally used up by the required FB-RAM producing a higher total system power usage.
Now going into more detail concerning FB-RAM's power usage:Micron FB-RAM technotes. It appears to be quite complicated to determine the amount of power will be used. Variables include the ram BW, the number of channels, the number of slots used, even which slots are used. Looking at figure 12, it appears that FB-RAM will use between 9W and 15W per slot used. and in the above review with 8x1GB FB-RAM using the 9W number produces 72W which is in line with the above review.
Fushigi
Storage Is My Life
Interesting review. While the CPU numbers look reasonable I call into question how a dozen 15K RPM drives can suck a total of only 9 watts of power. And the drives must be internal or the power consumption would be zero.
time
Storage? I am Storage!
Not sure where you got that from, Fushigi.
Fushigi
Storage Is My Life
"RAID Storage
LSI Logic 8480E MegaRaid Controller
Promise VTRAK J300s SAS Chassis
12 x 146GB Fujitsu 15,000 RPM SAS Drives configured in RAID 0"
Sorry, I was mostly just skimming the specs and saw the above. I didn't fully read the paragraphs of each system.
Considering they didn't quote power/performance for the RAID setup it still invalidates their server power/performance specs for any workload that touches the array. They should have either left it out of the server config or counted it; not a little of each.
I disagree, because the array is common to both they don't need to contrast between the two systems. The point of the array, is to help judge performance differences between the two servers and using the same array removes one major variable. I will agree, that it would have been a good idea to list the power usage of the array but it does not bother me excessively that it is not there. Perhaps, the reason it is missing is that the amount of power used was highly variable depending upon what it was being asked to do and thereby hard to measure accurately and give a easily comparable number.
Fushigi
Storage Is My Life
The RAID card & case fans were common to both systems and their power was quoted. Again, they were not consistent.
ddrueding
Fixture
They were consistent in quoting the RAID card and case fans for both systems, which is where it matters IMHO.
I don't think that issue makes much difference in Mark's argument; that socket-771 CPUs require FB-DIMMS and they suck huge amounts of power, removing the benefit gained by the more energy efficient CPU.
I don't think that issue makes much difference in Mark's argument; that socket-771 CPUs require FB-DIMMS and they suck huge amounts of power, removing the benefit gained by the more energy efficient CPU.
Just as a side note:
It appears that Intel will be releasing a XEON 5100 MCH chipset that will support DDR2 with the LGA771 socket for the l5400 series CPU's. When you see XEON motherboards with that chipset then you will not be forced into FB-RAM making my whole argument moot and Udaman correct. However, till then, I'm still right.
It appears that Intel will be releasing a XEON 5100 MCH chipset that will support DDR2 with the LGA771 socket for the l5400 series CPU's. When you see XEON motherboards with that chipset then you will not be forced into FB-RAM making my whole argument moot and Udaman correct. However, till then, I'm still right.
Chewy509
Wotty wot wot.
For those that may of missed it: Q6600 vs Q9300 @ anandtech.
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=3272&p=5
(Review is of the new AMD Phenom X4 and new Intel 45nm offerings).
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=3272&p=5
(Review is of the new AMD Phenom X4 and new Intel 45nm offerings).
Fushigi
Storage Is My Life
They were inconsistent. The storage subsystem was only partially quoted. Storage Adapter: Quoted. Storage Itself: Not. If the performance tests included the external storage it should have been accounted for. If the tests didn't the storage adapter should have been eliminated.
If you're trying to determine the power draw of delivering a service you have to include everything. Running a benchmark is delivering a service.
Look, I'm not arguing that the test methods are flawed. Or that the results - winner, loser - would be different. But if they had included all aspects of the environment being tested the graphs would have been flatter; the differences in performance & power consumption would have been the same but the relative importance of the differences would have been lessened. It would have looked less dramatic but been more realistic.
From the Power Analysis on page 5, under load the solutions are using roughly 308, 277, and 253 watts. The highest draws roughly 22% more juice than the lowest. Now add 120W for the drive array (just a wild number; 10W/drive). The results would then be 428, 397, and 373 and the difference would be only about 15%.
If you're trying to determine the power draw of delivering a service you have to include everything. Running a benchmark is delivering a service.
Look, I'm not arguing that the test methods are flawed. Or that the results - winner, loser - would be different. But if they had included all aspects of the environment being tested the graphs would have been flatter; the differences in performance & power consumption would have been the same but the relative importance of the differences would have been lessened. It would have looked less dramatic but been more realistic.
From the Power Analysis on page 5, under load the solutions are using roughly 308, 277, and 253 watts. The highest draws roughly 22% more juice than the lowest. Now add 120W for the drive array (just a wild number; 10W/drive). The results would then be 428, 397, and 373 and the difference would be only about 15%.
Or if they added 100 more drives the percentage would be even less.
Fushigi
Storage Is My Life
That's correct. And if the test bed had included as much then it would be important to note them. As part of a system, the power draw of the CPUs & FB-RAM would be trivialized by the power draw of the storage array making the power analysis of less relative value compared to, say, using 10K or 7200RPM disks instead of 15Ks.
The article was not a review of low power CPUs. It was a review of a Low Power Server.
The article was not a review of low power CPUs. It was a review of a Low Power Server.
Fushigi,
You definitely have a valid point. I agree, that it would have been better to include the power usage of the array. The missing data, distorts the performance/power results.
However, it isn't something I can't deal with. If you recognize it, the you can insert your own rough estimate for the missing data and get reasonably close to real data. Just as you did, above. The real harm done is the destruction of the confidence of the reader caused by the idea that if they didn't include this data, then what else are they not telling me that does matter.
Regardless of the above, I still found that the review supplied, to me, insight that I found valuable. As such, I used it for my own purposes: Specifically, the comparison of FB-RAM and DDR2 ram and their respective power usage. It made a big enough impression to be that when Udaman challenged me, I went searching for it to support my case. I have never seen another review that tried to seperate out the power-usage of the components, like this one did. Note, I did not use it as my sole support, but I also supplied a Micron technote that verified what was being seen in the review as to the power usage of FB-RAM. The review was offered solely to support my case and not as a general review.
You definitely have a valid point. I agree, that it would have been better to include the power usage of the array. The missing data, distorts the performance/power results.
However, it isn't something I can't deal with. If you recognize it, the you can insert your own rough estimate for the missing data and get reasonably close to real data. Just as you did, above. The real harm done is the destruction of the confidence of the reader caused by the idea that if they didn't include this data, then what else are they not telling me that does matter.
Regardless of the above, I still found that the review supplied, to me, insight that I found valuable. As such, I used it for my own purposes: Specifically, the comparison of FB-RAM and DDR2 ram and their respective power usage. It made a big enough impression to be that when Udaman challenged me, I went searching for it to support my case. I have never seen another review that tried to seperate out the power-usage of the components, like this one did. Note, I did not use it as my sole support, but I also supplied a Micron technote that verified what was being seen in the review as to the power usage of FB-RAM. The review was offered solely to support my case and not as a general review.
CougTek
Hairy Aussie
If I understand it right, this thread is about you guys dreaming about holding one of the new 45µm quad cores from Intel. Like what I did when I took those two poor pictures of my new Xeon X3350 below :
Someday, you'll have yours too...
Someday, you'll have yours too...
mubs
Storage? I am Storage!
Yay, Coug is back! Now stay, Coug, don't you ever disappear again!
paugie
Storage is cool
Yeah welcome back Coug!
(hmmm the avatar is familiar. but it seems the Coug we've known handled hardware with heavy boots!)
probably an incarnation. it's been enough time for a new terminator sequel
(hmmm the avatar is familiar. but it seems the Coug we've known handled hardware with heavy boots!)
probably an incarnation. it's been enough time for a new terminator sequel
![[Edit]](/forum/data/avatars/m/0/508.jpg?1546151166){kind=avatar}
Hmmm...
Seems as if I handled some of these 45 nanny quaddy Xeons also, but that was way back in December 2007... and it was a dual-socket server package... and its name rhymed with Hell.