X3350 (Q9450) first impressions

JazzJackRabbit · Post by **JazzJackRabbit** » Thu Mar 27, 2008 6:48 am

Got my Xeon X3350 yesterday which is xeon equivalent of Q9450. This is 45nm C1 revision quadcore with 8x multi and 1333MHz FSB running at 2.66GHz. If CoreTemp is to be trusted VID is 1.1125V which I believe is about in the middle of 1.0 - 1.225 range specified for xeon x3350 chips. There are good news and bad news. Good news is that it's running on my Abit IP35 Pro with latest beta 16B10 BIOS, drops multiplier when in idle thus saving energy. The CPU has Ninja RevB installed on it using push pins, there is no fan on the heatsink itself, it is being cooled by residual airflow from the case nexus at 7V. It seems to be running fairly cool at least as far as I can judge using exhaust fan temperature (more on temps later).

Bad news are mostly related to motherboard support. Temperature monitoring is almost impossible to get working. There must be some problems with sensors or motherboard support because in BIOS I've seen temperature jump to 200+ degrees Celsius which obviously can't be true. Latest SpeedFan does not pick up individual core temperatures. It does have "CPU" temp but I don't think it corresponds to real temperature. CoreTemp identifies my CPU as X3250 which is obviously wrong. On the temps it thinks I'm idling at 40+ degrees celsious which also seems high. I finally installed UGURU 3.107 yesterday and it seems to be working fine. It reports 26-29 degrees Celsius which seems about right. Curiously enough, I've never seen CPU temps jump to 200 degrees in uguru like I did in BIOS. I also can't pick any vcore voltage lower than 1.225V, it seems that Abit doesn't support any lower voltages. 1.225V is still within Intel recommended boundaries so I think it will be OK for the CPU.

Originally I wanted to try and overclock the CPU to about 3.2GHz, but I don't want to do it without proper temperature control. Maybe when Abit provides good working BIOS for the motherboard. Overall I like the CPU, it works as expected, drops multiplier when idle, and the air coming out of the exhaust fan seems cool. I am disappointed with lack of proper 45nm chip support from Abit though. That is not to say other manufacturers do not have the same problems, Gigabyte motherboards for example also have to be flashed to work with Yorkfields, but I still lament lack of proper support for new CPUs. SpeedFan is practically useless to me now because it can't read anything of value, so I might as well remove it from startup. For the longest time I had my fans running on sunbeam rheobus, I wanted to put fans on motherboard headers to automatically control them using temperature, but seeing how temps jump to 238 degrees in BIOS, I don't want my fans ramping up and down all the time.

paapaa · Post by **paapaa** » Sun Mar 30, 2008 10:20 am

I'm sure the support for Q9450 is/will be better than for X3350 which is the same processor but the server version (different ID).

Other than that Q9450 and Q9300 look very good CPUs for those who really need 4 cores with low power usage.

Willy Higinbotham · Post by **Willy Higinbotham** » Sun Mar 30, 2008 12:12 pm

JJR,

was there a particular reason why you went with the Xeon equivalent rather than the 'original' one? I recall you mentioning this on other threads as well, but never had the chance to ask.

JazzJackRabbit · Post by **JazzJackRabbit** » Sun Mar 30, 2008 3:15 pm

Several reasons, all fairly minor. One is availability. Q9450 is only getting to retail channels now, x3350 is already available if you look. Other reasons is that xeon versions generally have lower VID which should translate into lower power draw. Xeons also usually have higher operating temperature limits (do not confuse with the actual operating temperature).

ryboto · Post by **ryboto** » Mon Mar 31, 2008 6:55 am

just because Uguru can't select a vcore lower than what you're seeing doesn't mean a program like CrystalCPUID can't adjust lower. If that's what you're trying to do. I use it on an Abit board where I have similar lack of functionality through Uguru.

JLee · Post by **JLee** » Tue Apr 01, 2008 11:22 am

Are these chips identical?

q9450 and x3350

What are the differences aside from the voltage

elec999 · Post by **elec999** » Wed Apr 02, 2008 11:04 pm

JLee wrote:Are these chips identical?

q9450 and x3350

What are the differences aside from the voltage

I dont understand why the xeon version wouldnt be better then the regular cpu. I taught the xeon version was always better.
Thanks

NyteOwl · Post by **NyteOwl** » Thu Apr 03, 2008 2:15 pm

I dont understand why the xeon version wouldnt be better then the regular cpu. I taught the xeon version was always better.

Don't believe everything you read. The P4 Xeons of the same vintage were the same chip, just on a different pinout die so people wouldn't use the cheaper P4's on multiprocessor boards like happened with previous chip families.

I wonder then the x3360 would be a viable choice, since it seems to be cheaper here than an equivalent consumer quad core.

ronnylov · Post by **ronnylov** » Fri Apr 04, 2008 6:25 am

I have also bought myself a X3350 and I hope I will get some time to build my new computer this weekend. I did also choose the X3350 because of availability here in Sweden but if I just had waited a couple of hours I could have bought the Q9450. Well the price was about the same and hopefully X3350 is a little bit better.

Following new components are ready for assembly:
Antec P182 gun metal black
Corsair VX450 PSU
Gigabyte GA-EP35-DS4
Intel Xeon X3350 retail
Thermalright Ultra 120 Extreme heatsink
Scythe S-Flex 120 mm 1200 rpm fan (for the heatsink)
OCZ 2x2 GB PC6400 SLI memory
Gainward Bliss 9600GT (non overclocked version)
Arctic Cooling Accelero S1 heatsink
Three 120 mm Nexus Ultra Silent fans
Western Digital Caviar SE16 WD6400AAKS 640 GB hard drive
Sony NEC Optiarc AD-7200S DVD writer
Sony floppy drive black

My goal is a fast and hopefully silent video encoding machine. I probably will add some hard drives from my older computer because I need 3 drives for fast video editing (one system drive, one video source drive and one video destination drive). The gigabyte boards seems to have good undervolting options and I may try some overclocking too.

I will report back if there are any booting problems (if I need to flash BIOS with another CPU) and how the temperatures are reported.

EDIT: No problems booting with gigabyte GA-EP35-DS4 with original F1 BIOS and then I flashed to latest F3 BIOS. According to the bios settings the default Vcore voltage is 1.225 Volt which seems rather high, maybe I got a bad sample? I have not done any further testing than booting and flashing bios yet. The computer build is progressing slowly because I don't have much spare time as a father to two small children.

The standard cooling of Gainward 9600GT was very noisy so I have to use my Accelero S1. I thought it was a relatively silent card according to reviews but maybe there is something wrong with mine which is screeming like a vacuum cleaner...

badkarma · Post by **badkarma** » Wed Apr 09, 2008 8:51 pm

Have you tried undervolting your X3350? I'm just curious as the Q9450 is rated for 0.85V â€“ 1.3625V and the X3350 is rated for 1.212V. I'm pretty sure it's just because it's a server chip.

I have a X3350 retail sitting in the box. Going to lend it to a friend to review on his site. Actually if MikeC wants a crack at it before next Wed, he's welcome to it.

JazzJackRabbit · Post by **JazzJackRabbit** » Wed Apr 09, 2008 9:08 pm

I haven't because as I said above IP35-Pro wouldn't allow me going any lower than 1.225. I flashed new BIOS release (final 16) but it's still the same.

I will be lapping CPU/heatsink/putting it on TR bolt-through this weekend. After I do that I'll probably start playing with it for real. If I have enough time I may try ryboto suggestion using software tools to try and lower voltage. Although to be honest that was never my intention. I bought the chip so that I could overclock it to 3.2-3.6GHz with minimum voltage bump and minimum heat dissipation.

fwki · Post by **fwki** » Thu Apr 10, 2008 4:28 am

ronnylov wrote:According to the bios settings the default Vcore voltage is 1.225 Volt which seems rather high, maybe I got a bad sample?

My Q9450 defaulted to a high VID, and I thought the same. However it undervolts big time: Prime95 stable at 0.99375 VID (0.960 Vcore idle, 0.944 Vcore load). So don't worry too much about the default, just find your basement VID using Prime95. Here's my setup experience: viewtopic.php?t=47335&highlight=[/url]

ronnylov · Post by **ronnylov** » Thu Apr 10, 2008 5:16 am

Well, my computer build is progressing slowly because I do not have much time left with two small children at home. It seems all X3350 defaults to 1.225 Volts but someone at anandtech said that undervolting to 1.10 Volt worked for him at default speed. I have not tested my lowest Vcore yet. Anyway the gigabyte bios let me use any value from 0.8 Volts and up to over 2 Volts so I can try undervolting when I get more time left. I have finally installed all the hardware except the accelero S1 so tonight I may get time to install Windows.

I started with some memory testing and it seems OK up to at least DDR2 960 MHz in memtest86+ (using 400 MHz fsb, low CPU divider and memory divider for higher RAM speed) so at least my RAM is not limiting any CPU overclocking.

EDIT: I found this:
http://www.kerrywong.com/2008/03/28/tec ... and-x3350/

What does it mean? Can I not use the power saving features or will it make problems if I go to standby mode?

EDIT2: No problems with speedstep because it idles at 2000 MHz and 1,1 V Vcore and full load at 2666 MHz and 1,18 V Vcore (according to CPU-Z). Bios says 1,225 V is normal. CoreTemp reports 1,11 Volts VID and 1.06 V VID depending on load or idle (don't know why it says lower than real voltage).

I have big temperature differencies between the cores. At full load coretemp reports around 70 degrees on two cores and just below 60 degrees on two cores (as reported by coretemp, realtemp is 10 degrees lower). At idle the temperature difference can be 15 degrees celcius.

I tried some overclocking but 3,2 GHz seems to require at least 1,25 V Vcore in bios (I have load line calibration switched off and get 1,20 Volts in CPU-Z). The CPU fan speed increased to above 900 rpm and maximum temperatires (while running OCCT) increased to 73 C (coretemp). I have decided to not overclock and have gone back to stock settings because intel spec is max 71 C. Maybe I will try remounting the CPU heatsink and see if I get better temperatures.

ronnylov · Post by **ronnylov** » Mon Apr 14, 2008 11:50 pm

Now I have tried undervolting at stock frequency. It pass 1 hour OCCT mix autotest with 1,125 V set in bios which results in 1,07 V at load at 2667 MHz. Decreasing to 1,10 V in bios and it is not stable in OCCT anymore.

The undervolting decreased maximum load temperatures by 4 degrees C and the bios controlled CPU-fan (Scythe S-FLEX 1200) dropped speed from 830 to 750 rpm.

fwki · Post by **fwki** » Tue Apr 15, 2008 4:48 am

With the Q9450 Coretemp uses 105 as Tjunction while Realtemp uses 95, so Coretemp reports 10C higher temps. There seems to be a consensus forming that Realtemp and 95C are correct and that's what I think based on waking up the system from C3 (edit: S3) and quickly checking idle temp. If Tj is 95C then Coretemp is 10C high.

Many posters at AT are reporting Coretemp is also showing lower VID than what's set in BIOS, so I am ignoring VID in Coretemp and watching Vcore in cpu-z. Measured Vcore at the cpu is lower than the VID setting due to Vdroop on the MB. The level of difference is MB dependent.

Others are also reporting uneven temps across the cores with many speculating it could be due to uneven application of thermal paste...not sure about that, but my Q9450 is within 1-2C across the cores.

ronnylov · Post by **ronnylov** » Tue Apr 15, 2008 6:56 am

I thought the VID that coretemp reports would be the default Vcore of the processor but then bios sets a higher value so I wonder what it really is? Coincidentally the lowest Vcore that worked when set in bios was about the same as the CoreTemp reported VID.

I am now trying to find lowest stable Vcore at different frequencies, starting with 2000 MHz. Eventually I will get a nice looking graph. I like to explore my hardware so I can understand the relationchip between frequency and required Vcore.

I have also come to the conclusion that realtemp is probasbly closer to the real temps than coretemp. I found a nice guide on tomshardware forum:
http://www.tomshardware.co.uk/forum/221 ... ture-guide

I will remount the cooler and try again. Then I will try to find the highest stable clock at default voltage, lower the frequency a little bit for a stability margin and run with all power savings options enabled.