Couple of questions (Nick   )

I recently exploded my overclocked Q6600    (don't ask)

And grabbed a replacement (until I'm ready for i7)CPU and new M/B..

I got a deal (meaning they had it on the shelf and I didn't have to wait)on a Q9550 / EP45-DS3R.. and have it running at a timid 3.11Ghz, and it runs FSX noticeably better than than the Q6600 @ 3.2Ghz..

So... I figure it's gotta be the L2 ?

And to top it off.. it runs ice-cold (30C idle ).. Is that 'cause of the 45nm ?

LOL

I love this report because it flies in the face those who claim such an upgrade is not worth the investment

What I am trying to figure out is how you killed a Q6600.. thats pretty hard. I have turned a system on with no heatsink on Core2 quads and they simply shut down without being fried.. not something I suggest people do though.  Sorry you lost your proc.

You are seeing the result of 2 things

First, yes.. the cache.

Second is the 45nm design and the 10-15% increase in optimized internal communication of the proc

Due to 45nm design the Vcore is lower and therefore the chip runs cooler at the same CPU speed


Now, there are RULES around 45nm which are not the same as 65 in CPU VOLTAGE. A 45nm CPU even LN2 cooled can not run a CPU VOLTAGE of higher than 1.40 or damage may ensue. @ 1.50 and above I would say within 6 months the proc is toast if not sooner where with 65nm one could LN2 cool on 1.8v and never see any problem


As for clocking it.. .the Q9550 has a 8.5x multiplier and if it is E0 stepping it will make it to 4.0GHz on about a 1.32-1.36 CPU VOLTAGE and a VTT or FSB TERMINATION VOLTAGE of 1.28-1.30 safely and usually a northbridge voltage of about 1.42-1.55. That high clock does require a proper clocking heatsink to accomplish and a motherboard able to run 475-480MHz

Most Q9550 chips are not E0 stepping and they can be hard to find. If you are lucky to have one it is very easy and stable to clock. The Stepping will be presented on the CPU tab of the software CPUz.

http://processorfinder.intel.com/List.aspx?ParentRadio=All&ProcFam=0&SearchKey=Q...

There are 3 Q9550 steppings and models.. C1, E0 and the Q9550s with a stepping of E0 as well.

C1's are not as easy to clock as E0's

The main difference in the Temp spec is with the Q9550s compared to the other two.

The max temp as read by CoreTemp under a FULL LOAD test using the software OCCT v2.01 is 78c (could be pushed to 80c without problems) however with the Q9550s (ONLY the S version) that expands to 85c

The Tjunction Max or the point of damage is 100c on all but the S chip and 105c on the S model

Brett

make sure you use CORETEMP to read the processor temps.. you do not want to read the wrong value. Coretemp will read the correct value

RealTemp can also be used but you may need to manually select and input the correct TJMAX value for it to display correctly.

Brett_Henderson wrote on Jan 23^rd, 2009 at 10:48am:



RealTemp will work.. I dont know what version you are on but the newer one has a section for inputting the TJMAX in your case its 100c unless you are running a Q9550s

Right now I am on 180.48 however I do use XP too... there have been mixed reports about it with Vista 64

Before that I held fast to the 174.74 beta driver..

..all of which are available at Nvidia in their driver download area. You may need to input for archive drivers and allow beta drivers to appear too.

Brett_Henderson wrote on Jan 23^rd, 2009 at 10:54am:


If you download and run CPUz.exe it will show you the stepping and processor.. it will also show you the memory speed and timing onthe memory tab

If you give me links to the motherboard and memory product I can give you clocking values within reason of your heatsink but you will still need to check the temps under load

You do want the CPU STRAP set to 333..   no exceptions and them set the memory speed based on that divider in use


http://www.cpuid.com/download/cpuz_149.zip

OK.. "Revision"  is  "E0"   

I'll get those links up ASAP.

THANKS !

Motherboard..

http://www.gigabyte.us/Products/Motherboard/Products_Overview.aspx?ProductID=284...


The RAM I recycled from the dead computer.. here's the newest link:

http://www.newegg.com/Product/Product.aspx?Item=N82E16820145215

Brett_Henderson wrote on Jan 23^rd, 2009 at 11:17am:




OK.. give me sont time to look this over while I sip coffee and I will throw out some values to work with. The primary concern is to not exceed 2.1v on the DRAM VOLTAGE, 1.36v on the CPU VOLTAGE and as long as the temp of the processor remains below 80c, you are good to go

How high you can go will depend on that temp and remaining at or below the max Vcore, which should not be a problem. The advantage of the Q6600 is its 9 multiplier.. you are losing some ability there but gaining on another end



you have a vendetta2 ????

Let the good times roll

damn it!

Its a P45 chipset.. you may be limited on FSB.. checking now. Some of them would run high FSB and others would not

the right board with the x38 or x48 chipset would not be limiting

Its probably going to be some trial and error in this with the P45

you got the right heatsink.. thats for sure.. the Zalman cant really handle the better quads and is old news

Brett_Henderson wrote on Jan 23^rd, 2009 at 11:35am:



Im expecting 3.8 when we are finished if the P45 will comply

and thats not outrageous

Yeah.. that's what happenes when you buy what's on the shelf. Normally I research (and ask) before buying.. and will when I go i7.. but this was a panic buy..   

Don't spend too much time on this..  I'll grope my way through it..  I'll likely be moving up in 6 months, anyway.. I'm already getting improved performance.. even at 3.11Ghz

Not sure what BIOS you are on but there have been revisions

http://www.gigabyte.us/Support/Motherboard/BIOS_Model.aspx?ProductID=2842

I would not mess with it unless you have a problem clocking however they do post enchanced overclocking ability and stability has been added

BIOS F10 is the latest...

According to CpuZ, it's F10..

It seems like an easy to use BIOS.. I mean, even "I" just went in,, boom boom boom and upped it 10% without even thinking about it.

I'll likely not push it to where I'm cranking voltages up (much)..

It's just fun top play with it 

So... don't cut your day in 1/2 on this.. I'm in no hurry..  I'm gonna play with it all night..lol

Brett_Henderson wrote on Jan 23^rd, 2009 at 12:14pm:




ahhhh OK you are just playing with the automated overclocking features and boosting %...   we are going to get you off that system and put you in behind the yoke

If this motherboard manual ever finishes downloading


EDIT: Ahhh   finally here..

I will read this over and post back in a bit

Brett_Henderson wrote on Jan 23^rd, 2009 at 12:29pm:



Yaaaa

You are reading the CPU with SPEEDSTEP and CE1 support enabled.. those REDUCE the CPU multiplier automatically till a load is sensed..   and we will be shutting those down too




EDIT: Oppppsss...   I read that wrong.. It reads 3.4?


OK.. thats 8.5x Brett, not 8.. so its reading the full 8.5 multiplier

OK  gig is using a slightly different method of CPU STRAP to memory..

There are differences in how this board will work compared to Asus boards.

Here is what I would like you to do in order to investigate a bit and confirm a few things since what they say in the manual and what really appears on the screen in the BIOS can be different especially if they have upgraded the BIOS since the manual and changes the programming math of the BIOS

You will not SAVE these settings at this time.. I just want to know if they are working the numbers as the manual suggests...

Enter the BIOS and find:

CPU Host Frequency (Mhz) and raise it to 450

Now.. the available Memory Frequency (Mhz) at this FSB will be based on

(G)MCH Frequency Latch - Same as CPU STRAP in ASUS

System Memory Multiplier - Available values depend on Latch setting


Here is the deal.. we are looking for 1066... those sticks you have are not clocking sticks however they may have some overhead in them. If they do I doubt it will be much higher than 1090 @ 5-5-5 so what you are looking for is DDR2 1066 or just slightly above in Memory Frequency (Mhz) @ 450-460FSB

When looking for this value, first lock the

(G)MCH Frequency Latch to 333

Latch freq is the NB strap, basically, internal NB timings. Lower is tighter. The separate MCH Latch settings in the Giga boards are are not redundant you can pick mem ratio and the strap in the mem section at the same time but with the LATCH set to a value it limits the values under System Memory Multiplier to ones that only function with the LATCH of 333.

(G)MCH Frequency Latch to 333 - remains a constant

Now, work the CPU Host Frequency (Mhz) (start at 450 and go up but try to keep it as close to 450 as possible without going UNDER), the System Memory Multiplier and check the Memory Frequency (Mhz) ... you are looking for a Memory Frequency of 533MHz or also known as DDR 1066 (533x2)

Now.. once you find the FSB and memory multiplier that gets you as close to 1066 as possible without going UNDER but without going too far OVER (try for 1080max if that is the best you can do) then verify the following


Standard Timing Control
CAS Latency Time: 5
tRCD: 5
tRP: 5
tRAS: 15



If this all checks out and you have established a 1066-1080 memory speed @ that timing.. then you are ready to clock

At this point you can cancel out of the BIOS and remember the settings you found above and let me know what values you ended up to get 1066 (or close) with for


CPU Host Frequency (Mhz)

System Memory Multiplier

and what the final Memory Frequency (Mhz) ended up at

This will tell me what to post for voltages to stabilize the clock attempt and keep you safe


To recap, the goal in this exercise is to find the Memory Frequency (Mhz) which will place the memory as close to DDR2 1066 (533Mhz) as possible on 5-5-5-15 timing with a locked (G)MCH Frequency Latch set to 333 and a CPU Host Frequency (Mhz) of at LEAST 450Mhz or slightly higher


NOTE 1: If you find the memory speed is sitting at 1080-1090 @ 450MHz and you can not get it lower without reducing CPU Host Frequency (Mhz) to below 450.. tell me where FSB sits below 450 to place memory at 1066 as I may go that route instead


NOTE 2: Setting (G)MCH Frequency Latch to AUTO instead of 333 may allow a 1066 memory speed at 450MHz.. check this too although I would prefer (G)MCH Frequency Latch be locked at 333

Brett_Henderson wrote on Jan 23^rd, 2009 at 2:09pm:



Windoze is reading the full 8.5x multiplier... probably a BIOS glitch

I need to remember how to activate that 1/2 multiplier and lock it on.. not sure if Gigabyte does it automatically or not



Just to make sure... you ARE using this CPU cooler, correct?

http://www.ocztechnology.com/products/cooling_products/ocz_vendetta_2_cpu_cooler

Thats the Vendetta2

or is it just the 1st release Vendetta?

http://www.ocztechnology.com/products/cooling_products/ocz_vendetta_cpu_cooler


The primary difference is the fan size.. as I recall the V1 is a 92mm fan and the V2 is a 120mm and there were a few other minor tweaks. So let me know about that because that may affect the settings.. you can always strap a 120mm PWM 4 pin fan to the Vendetta1 and get basically the same cooling result down the road

OK so 444MHz CPU Host Frequency (Mhz) is 1066

OK so here are the options.. and its based on if the memory will run 1080 stable

All of the tests will have the following DISABLED

Limit CPUID Max. to 3
CPU Enhanced Halt (C1E)
C2/C2E State Support
C4/C4E State Support
CPU EIST Function
Virtualization Technology

I left the TM function enabled for you for thermal protection. What I posted should lock the processr at the full 8.5x multiplier

Same settings @ 450 placing the memory @ 1080

The CPU Vcore will be problematical and you may need to work this a bit. I estimate 1.28-1.32 will be the range however that may be stable a bit higher or a bit lower than the range

Lower is cooler and always better. 1.38v will be your max and do not exceed it

Start at 1.28v

CPU Termination will be set to 1.25v. This can end up a bit higher as well with a max on 1.35 however I would work that in up by starting at 1.25 then move to 1.28, 1.30, 1.32, 1.34 testing for stability. To be quite honest I dont see this needing to be any higher on the clock we are looking at than 1.28v

CPU PLL will be set to 1.50v and leave it


MCH Core: This is the northbridge voltage and may need to be adjusted from AUTO for higher FSB. Lets start at 1.3v with the possibility of this needing 1.35 or 1.40, or 1.45

start at 1.30

DRAM Voltage will be set to 2.1v and no higher as per the spec from Corsair

Leave everything else on AUTO



Ok.. F-10 save and exit.. she should boot


NOTE: it is possible the system may boot into Windows and be unstable in load tests. At that point you need to do some voltage tweaking.. the same tweaking you would do if the system did not boot at all as I outline below:

If it fails to boot or fails in Windows under stress then enter the BIOS and bring up the Vcore and the MCH Voltage slightly and try again

VCORE 1.28

and

MCH CORE 1.35

If still fails in load tests or no boot it is very possible this memory simply will not do 1080 and that is the problem. To do a quick check enter the BIOS at that point and raise DRAM VOLTAGE to 2.2v leaving everything else as it was and F-10

If it boots then the memory is the issue. At 2.2v you are pushing the outer limits of the spec and although it can be run like that it is past what Corsair specs for the memory and could shorten its life however you can NOT go above 2.2v

If you would rather run the 2.1 spec then the only option is to back down the CPU Host Frequency (Mhz)

At that point lest set the voltages back to the original list I posted above and set the CPU Host Frequency (Mhz) to 444 which makes the memory speed 1066 and boot up

I would think that would be stable but if not, its the same drill as above.. start with Vcore and MCH CORE first and if after you have raised those to to VCore 1.32 and MCH CORE 1.40 you still are unstable post to me and let me know


Now, when in Windows do not use the system properties to see the CPU Speed and the memory speed and timing, use CPUz

At 450x8.5 you should have a CPU speed of 3825MHz

At 444x8.5 you should have a CPU speed of 3774MHz

Both are nice clocks


but





This would be a setting of 471MHz CPU Host Frequency (Mhz) and adjusted memory speed for as close to 1066 as possible using the method mentioned above. Check your multiplier ranges at that HOST speed and see if 1066 comes up close to or around 471MHz. Even if you end up a touch slower than 1066.. like 1045 it may be worth the 4GHz clock.


Same deal on the voltages.. you may find you will run just fine at 4003.5MHz

So thats how its done


Now, temps

no matter what speed...

http://www.ocbase.com/download.php?fileext=exe

You need to check for temps and stability in such clocks. I typically use the software OCCT and run the 1hr automatic test. I enter the settings for OCCT and make sure the CRASH or FAIL temp is set to the Tjunction of the processor.. in your case this is 80c

Fire up the test with ReatTemp or Coretemp up on the screen.. there will be a 1 minute prep period then the test will start.. temps will go sky high and that is what we want

As long as the 1hr test passes and your temp remains below 80c you are set on that clock

Now, if you are stable and temps look good you do not need to make any changes however.. if Vcore will drop a tad and stability remain a constant then the proc will run cooler. I estimate you wont get much lower than 1.25 no matter what and in the case of 4GHz it may or may not run less than 1.3v. Some do, some dont.

The next step is to run FSX and make sure you are stable in 3D. If not let me know and we can go from there

The 2 primary settings you will be messing with for stability are

VCORE

On the 4GHz clock this may need to be 1.36 but many manage a 4Ghz clock at 1.28v. do not exceed 1.38 and temps must be under 80c in load tests no matter what voltage

MCH CORE

MCH may need to come up to as high as 1.50 but do not exceed that without checking with me

you get the idea now...


when you get that stable and its up in the 3.7 to 4GHz range.. FSX will sing for you

I just noticed this board may have these settings


CPU Clock Ratio
Fine CPU Clock Ratio

If it does make sure it is set to

CPU Clock Ratio [8X]
Fine CPU Clock Ratio [+0.5]


Sorry I missed that!


You can also just work with 8x..   that is a way you can find stability as well by reducing CPU speed if you find the CPU is unstable


The trick is figuring out if its

the CPU

the Northbridge

or
the memory

causing stability... thats why we work the voltages back and forth, up and down to find the stable lowest point for each

Dont get frustrated.. it may take a bit of playing and a few posts to lock it down


As for the memory, I just checked to be sure.. Corsair only warrantys those sticks to 2.1v so anything above that is pot luck. Some run them at 2.2 (never higher) and never have a problem but that is a risk. running 2.2 for a boot and stability check is fine.. running it long term is what can be dangerous

QUACK QUACK QUACK

gets 'em every time

he he he


Yea.. I left that out on purpose   sorry, call it a hazing for the  clocking fraternity



Thats fine Brett.. as long as voltages are within the limits I posted and the temps are solid.. you are fine and I know you can do better on the clock.

When you are ready for a memory purchase let me know. Those hybrid DDR2/DDR3 boards can be a bit finicky about memory though

You should be stable as long as the memory is @ 1066 and 2.1v.. but it may require higher MCH Core and Vcore with a touch of VTT to run the higher FSB

Those E0 95 and 9650 chips will do 4GHz on the right boards pretty easy

DDR3 1600
Brett_Henderson wrote on Jan 24^th, 2009 at 4:44pm:



LOL!!


i7 DDR3 is not the same as the 1st generation DDR3 used in that board... and the memory you posted is a bit too bold for that board

If you purchase DDR3 now it can not be used for i7 so dont go spending a bundle.

In DDR3 @ 400MHz I would go with these:

http://www.newegg.com/Product/Product.aspx?Item=N82E16820146777


@ 1.9v they will also do 1T instead of 2T COMMAND RATE (faster) and I can show you a way to pop a bit more out of them too using a tool called MEMSET which will access the northbridge and set the tRD (MCH READ DELAY) down to 6 which will boost their ability again..


That board is a hybrid so I can not be confident of its stability tweaking as mentioned however those sticks are really nice and even if you could not tweak them as I posted should do the job a lot cheaper





EDIT: Posted the wrong memory... edited the link for 2x2GB

Brett_Henderson wrote on Jan 24^th, 2009 at 5:11pm:


If you have any questions while toying with settings shoot them to me


Also.. in case you have not seen it.. I did post a FSX tuning guide for settings and config tweaks

http://www.simforums.com/forums/forum_posts.asp?TID=29041

It covers just about all the bases including the use of Nhancer


hmmmmm.... in your area I can think of a few well known NFL boys who I would not want to upset