Sign in with
Sign up | Sign in
Your question

Xeon X3350 OC and Review (WIP)

Last response: in Overclocking
Share
April 11, 2008 8:34:32 PM

This was requested by a few people wanting to see what the Xeon 45nms could do vs the Q9450. While I can't give you the Q side of it, I will do what I can to comprehensively review the Xeon.

First off, BIOS flash to 1004 was necessary on the Maximus Formula SE X38 Mobo, so be ready Asus owners.

On to the OC! For the first test, I just went to 3.2. FSB to 400(x8). Vcore at 1.2 in BIOS, 1.136 under Prime load. I left NB voltage to auto for the moment. Cores are between 65 and 72c during Prime 95 Small FFT. CoreTemp reports that in Vista finally, but I hope it's wrong.




As you can see, the cores are already pretty toasty. I think I have a little more room to play due to the higher Tjunction Max, but it just makes me nervous :(  I hope I have been bitten by the faulty sensor bug :p  I may download RealTemp and compare, Comp says it is the one to trust now.

I will update this as I go, and post the Review portion after locking in the final OC.

Update: I'm retarded and had my fan mounted backwards on my TRUE.....that OC failed anyway, not due to temps.
I'll wait until I get a stable OC to post progress :) 

Update: Can't get 3.6 to load Windows with as much as 1.4 vcore in BIOS.... Sticking with 3.4, which seems stable at 1.3125 in BIOS.

8:24PM - CoreTemp shows a +10c difference to Real Temp. There is no way Core Temp is correct, because as of right now, it's showing cores over 85c. Surely, that would at least throttle the CPU, if not shutdown. I know the shutoff Tcase temp on the 6750 was 72c. I don't know what it is for the Penryn series.

Real Temp seems to be very accurate. Idle temps in the 40s seems possible with a quad at 3.4. Core Temp shows idles in high 50s, can't be true.

I am kind of dissapointed with this chip as far as a great OC'er. I didn't necessarily expect it to be, but I thought I could get at least 3.6 with all of the 3.8s and 4.0s you see everywhere on the QX9650. I wil continue to attempt 3.6 after I make sure 3.4 is stable at lowest voltage needed. I don't wanna have to do this twice if I somehow can't get 3.6.

Interesting side note.. I fail POST at 3.4 with my RAM set to 1066 (EPP Profile stock speed) and stock timings, with proper voltage. I had to slightly underclock the RAM for it to pass POST. Possibly the rumored 4 x 1GB stick is bad for overclocking (on a quad)?

Update in the morning. Stressing overnight.

More about : xeon x3350 review wip

April 12, 2008 1:33:23 AM

Looks like 3.4 is going to be the limit, but not due to overvoltage. If Real Temp is accurate, which I believe it is, at 3.4 my hottest core seems to have peaked at 72c. Pretty warm, but it should come down a bit as the AS5 sets up. We all know that everyday use and most gaming will never heat all 4 cores that warm.

I believe that I'll just stick to 3.4 until I confirm with an IR thermometer the temps I'm seeing.

I'll run 3dMark in the morning and post results. Any other benchies you guys want?
April 12, 2008 2:32:52 AM

Interesting, could you also run Crysis in game benchmark or something similar too?
Related resources
April 12, 2008 12:31:37 PM

Just tried everything I know, aside from jacking the voltage over 1.45 to get 3.6 to load Vista and no joy.

Oh well, on to the Benchmarks! I'll have to reinstall Crysis, but I'll give it a whirl :) 

I can't believe I can't get 3.6 out of this thing with reasonable voltage. Oh well, a 3.4 quad should still be more than enough :) 

Benchmarks soon.
April 13, 2008 3:15:36 PM

3.4 is looking more and more like max for this particular chip. I had some strange problems, getting them all sorted, then I'll run benches. This OC was a little more tricky than the dual core...

Update soon.
April 13, 2008 8:23:34 PM

Keep up the good work.

I suppose you got a lemon chip huh.
April 13, 2008 9:11:32 PM

For the sake of others who ordered this same Xeon, I hope mine is a lemon. I knew anything over 3.6 would be difficult with the 8x multi, but 450 isn't all that high of an FSB.
April 14, 2008 4:15:46 AM

13863 3dMark 06 Default Benchmark

Link was only temporary.... I'll post a screenie when I get home.

Not all that much better than my 6750. It got 13.2k.

Didn't expect CPU to cause that much of a jump tho.

Installing Crysis soon.
April 14, 2008 8:18:45 PM

By the way....for those interested... my VID on this chip according to CoreTemp is 1.0500.......Doesn't seem to make it a great overclocker :) 
April 14, 2008 8:25:38 PM

LOL, thanks! It will be a while before I move on to see if VID matters, but it more than likely will. I know you've been reading the posts about it! And you are right, thats interesting that it cant overclock to good. That doesnt bode well, hehe.

I will be eyeing this to see what else you come up with! What kind of lil things have been going on that you can attribute to the processors' OC?

--Lupi!
April 14, 2008 9:03:54 PM

Nothing that I can attribute directly to the proc. I know that with a quad it is harder to push the FSB, and supposedly with 4 1 gig sticks of RAM. I can only guess those are my main limiting factors if they are indeed true.

Weird problem #1 was the RAM not allowing me to run it at stock when OC'd to 3.4. It may have just been some strange ratio that the mobo didn't like or something, but I've never seen an OC that wouldn't run RAM at the rated speed.

Wierd problem #2 would be Ntune. It worked fine on the 6750 allowing me to change fan speeds for my 8800 GT. When I installed the Xeon, it would blue screen me upon launching it. I removed and reinstalled everything I could think of that would have some residual CPU information that could be fouling it up, but to no avail. I have to use RivaTuner now, which is no big deal, but just another process hogging system resources.

Weird problem #3 would be not being able to load Vista at 3.6 with as much as 1.45 vcore. That really freaked me out. I might send my chip to Cnu to see what he can do with it. Maybe I'm missing some obscure setting that I've never had to change before.
April 14, 2008 9:05:59 PM

I'm about to go home and decode 4 Gigs of iTunes m4ps to mp3s. At least I'll see what this baby can do when you feed it multimedia :) 
April 15, 2008 12:45:55 AM

I agree with you there. Who knows what new settings the new chips will like. I just thought I would mention this, but maybe those chips hit their wall sooner? Thats just blabbing. No real info, but I did notice that in the q6x00 series needed a huge voltage jump to go from 3.6 to 3.8

I was amazed at the amount of VCore to reach 3.8 fully prime stable. It was like 15 notches outside my worse case guess! So from 2.66 to 3.6 was like 8 notches. From 3.6 to 3.8 (prime stable, daily stable is way less than prime stable!) was 26-29 bios VCore notches.

That is one hell of alot! Just a thought!

Hey, do you know if there is just a "standard" test for vid cards? I mean, something that test them to make sure they comply with their settings, and the performance you should get? As in, my Ultra seems a bit slow to me. Is there a way to test it to make sure its even living up to their stock performance?

--Lupi
April 15, 2008 12:56:33 AM

Well... Basically, 3dMark will tell you these things. However, finding systems identical to yours is the key to comparison in that case. Luckily, a single 8800 GT is pretty common on the Orb.
April 15, 2008 1:07:49 AM

I'll have to just run it by 3dmark on my next boot. Maybe its just crysis! But I expected more from an Ultra, lol!

--Lupi!
April 15, 2008 2:30:41 AM

Crysis will bring SLI'd ultras to a crawl on Very High settings :p  Well....40ish FPS, but not "normal" Ultra #s.
April 15, 2008 10:48:33 PM

Heck, even with Tri-SLI'ed ultras won't give you a very good gaming experiences on Very High. Got those benchies in?
April 15, 2008 11:04:33 PM

Did you poke around for your FSB wall by chance? Have you tried to remove two RAM modules? What RAM timings are you using?
April 16, 2008 3:45:49 AM

I didn't try for 3.5 or anything really. If the core temps at 3.4 are accurate according to RealTemp, then 3.6 probably wouldn't be a good idea anyway. I didn't try to remove any RAM, because I wouldn't run this machine with 2 gig so why even bother :) 

RAM is set at 5-5-5-18.
April 16, 2008 3:47:00 AM

@Evil - Haven't even installed Crysis yet. Been a crazy week at work. Having to stay late an aweful lot. I'll get on em ASAP :) 
April 17, 2008 3:51:07 AM

Running CPU benchmark 1
Results will depend on current system settings
Press any key to continue . . .
Running...
==============================================================
TimeDemo Play Started , (Total Frames: 1500, Recorded Time: 44.62s)
!TimeDemo Run 0 Finished.
Play Time: 50.87s, Average FPS: 29.49
Min FPS: 14.82 at frame 196, Max FPS: 40.93 at frame 105
Average Tri/Sec: -38446408, Tri/Frame: -1303844
Recorded/Played Tris ratio: -0.55
!TimeDemo Run 1 Finished.
Play Time: 47.68s, Average FPS: 31.46
Min FPS: 14.82 at frame 196, Max FPS: 40.93 at frame 105
Average Tri/Sec: -40857908, Tri/Frame: -1298800
Recorded/Played Tris ratio: -0.55
Press any key to continue . . .

Not sure how to tell which res/AA that was...I'm guessing 1680x1050 No AA, but not sure about quality.
April 17, 2008 4:09:48 AM

This is Benchmark_CPU2.bat

I noticed it's at 1680x1050, but still not sure how to tell AA and quality levels.

Running CPU benchmark 1
Results will depend on current system settings
Press any key to continue . . .
Running...
==============================================================
TimeDemo Play Started , (Total Frames: 1500, Recorded Time: 127.52s)
!TimeDemo Run 0 Finished.
Play Time: 73.09s, Average FPS: 20.52
Min FPS: 11.68 at frame 453, Max FPS: 26.46 at frame 104
Average Tri/Sec: -3448770, Tri/Frame: -168050
Recorded/Played Tris ratio: 4.97
==============================================================
TimeDemo Play Started , (Total Frames: 1500, Recorded Time: 127.52s)
!TimeDemo Run 0 Finished.
Play Time: 73.47s, Average FPS: 20.42
Min FPS: 9.75 at frame 477, Max FPS: 26.48 at frame 105
Average Tri/Sec: -3532312, Tri/Frame: -173020
Recorded/Played Tris ratio: 4.83
Press any key to continue . . .

GPU Benchmark is irrelevant unless you just wanna see what an 8800 GT can do :p 

I did notice that in the upper right, it read "VeryHighSpec HDR Devmode Streaming" Not sure if that means anything, but I guess those could be "Very High" settings, No AA.

Also, this was 32-bit DX10 mode. 64-bit DX10 would freeze on load screen. 64-bit DX10 never worked for me on the dual core either tho....can't fault the Xeon for that one.
April 17, 2008 4:37:05 AM

Confirmed, this was at 1680x1050, Very High, No AA. 32-bit DX10.
April 22, 2008 6:11:09 AM

I found mine to be FSB wall. Tried 450x6 with as much as 1.65v on the NB. No dice. Dual core was able to push higher FSB, so I have to blame this on the additional cores. I knew this was the case, but didn't expect it to give up so early. I'm done tryin to get anything higher. 3.4 screams thru whatever I toss at it. Even if I could get 3.6+, I wouldn't leave it there except to run a few benchmarks for comparison.
April 22, 2008 6:17:57 AM

Ninjawithagun said:
I still find it amazing that the X3350 adjusts the voltage automatically up to around 3.6Ghz. Gotta love that :D 



I have to ask what you mean by that? Do you leave your vcore on auto??
April 22, 2008 11:54:23 PM

A setting of auto will always increase the voltage along with the cpu clock speed (or fsb?). It is generally a poor idea to leave voltages on auto because more often than not, the bios always overshoot, giving the cpus more than voltage than necessary.

Ex:
Auto voltage for 3Ghz Q6600: ~1.40v

Minimized voltage for 3GHz: 1.30v
April 22, 2008 11:55:09 PM

A setting of auto will always increase the voltage along with the cpu clock speed (or fsb?). It is generally a poor idea to leave voltages on auto because more often than not, the bios always overshoot, giving the cpus more than voltage than necessary.

Ex:
Auto voltage for 3Ghz Q6600: ~1.40v

Minimized voltage for 3GHz: 1.30v
April 23, 2008 1:48:24 AM

Agreed. Auto will always show green on your onboard LEDs if that's what you're going by. But as Evil said, it will always pump way too much voltage into the proc for the desired speed.
April 23, 2008 1:54:45 AM

Everyone that seems to be getting great OCs on the Penryn quads are (in my experiences) using Nvidia 780/790i chipsets. Seems Nvidia scored on that one very well.

I would be interested to see anyone out there with X38/X48 chipsets OC for the Penryn quads. If it's just a chipset limitation, I may splurge and buy one. I have an Nvidia card anyway and wouldn't cost much to SLI the 8800 GT. Might be an option!
June 6, 2008 10:17:51 PM

I am running:
Gigabytes EP35-DS3R Motherboard.
4Gb G.Skill DDR2-1066 5-5-5-15 Memory
Geforce 8800GTS 512Mb 730MhZ / 1065Mhz Mem.
Xeon X3350 12Mb Core 2 Quad 2.67 > @ 3.42Ghz With out even trying to push it. (I am using Stock CPU Cooler off the 65Nm Dual Core Proc. =)

I get 43.22 Frames per Sec AVG on Crysis 1280X1024 High Detail No AAs

Runs very stable and smooth.
I only got $1,200.00 invested in entire computer.

Thanks plenty good enough for me.

I have been running World of war Craft for hours on end with no issues plus my computer is so amazingly quiet I cannot hear it.
June 6, 2008 10:35:51 PM


Also, I get 15,636 on 3D Mark 06 at 1280X1024

CPU Score is 5617
Sm2.0 Score 6510
SM3.0 Score 6254


GT1 Return to P.. 54FPS
GT2 Forest... 54 FPS
Canyon... 61 FPS
Deep Freeze 63 FPS
CPU1 1.86FPS
CPU2 2.71 FPS

Also I only run Window XP Pro.

!