1075T Unlocked Multiplier to 17.5x

Hey everyone I just wanted to say a big hello to everyone on the board. Your answers have been a great help to me. I really appreciat all of your expertise.

Here's my question or rather an interesting observation I made.
I'm building a system for a friend of mine with GA-890FXA-UD5 and a 1075T. I'm using a Hyper 212+ for the HSF. The case is a HAF 912. I'm using G.SKILL Ripjaws 4GB DDR3 DDR3-1333 for the system memory.

What I noticed today was when I went into the M.I.T(overclocking portion in the Gigabyte Bios) in my bios I was able to change my CPU Clock Ratio to 17.5x where it was normally at 15x. This bumped my speed up to 3.5GHz rather than 3 GHz which was the same as my Turbo, so I just turned it off.
I am currently running Bios F3. When I updated to Gigabyte's newest bios (F5) the option was gone and I could only get to 15x for the CPU Multiplier. I downgraded my bios back to F3 just to make sure I wasn't crazy and sure enough there was my option for 17.5.

My question is should I just stick with the F3 bios and get an easy OC to 3.5GHz? At this point I have Turbo disabled since it's not going any higher than 3.5GHz.
My friend isn't a huge fan of OCing, but if I can do this little tweak and pretty much have it run in turbo all of the time I'm sure he'll be fine with it. He just doesn't want me to do anything with uping the VCore and I'm fine with that. The idea is to have a 24/7 stable machine.

Let me know what you think.
Thanks in advance!

Bondo said:

Hey everyone I just wanted to say a big hello to everyone on the board. Your answers have been a great help to me. I really appreciated all of your expertise.

Here's my question or rather an interesting observation I made.
I'm building a system for a friend of mine with GA-890FXA-UD5 and a 1075T. I'm using a Hyper 212+ for my HSF. The case is a HAF 912. I'm using G.SKILL Ripjaws 4GB DDR3 DDR3-1333 for the system memory.

What I noticed today was when I went into the M.I.T in my bios I was able to change my CPU Clock Ratio to 17.5x where it was normally at 15x. This bumped my speed up to 3.5GHz which was the same as my Turbo, so I just turned it off.
I am currently running Bios F3. When I updated to Gigabyte's newest bios (F5) the option was gone and I could only get to 15x for the CPU Multiplier. I downgraded my bios back to F3 just to make sure I wasn't crazy and sure enough there was my option for 17.5.

My question is should I just stick with the F3 bios and get an easy OC to 3.5GHz? At this point I have Turbo disabled since it's not going any higher than 3.5GHz.
My friend isn't a huge fan of OCing, but if I can do this little tweak and pretty much have it run in turbo all of the time I'm sure he'll be fine with it. He just doesn't want me to do anything with uping the VCore and I'm fine with that. The idea is to have a 24/7 stable machine.

Let me know what you think.
Thanks in advance!

Click to expand...

I think you should buy him a new 1075t and tell him his is broke (then keep it). Lucky guy!

To answer your question it depends on the temps. If it was my machine, sure, do whatever...but for a regular guy I would let the turbo work it's magic. The beauty of the turbo is it only overclocks when the overclock is needed....saving electricity and producing less heat. I doubt your friend would know the difference either way. I say better safe than sorry on someone else's rig.

The thing that I read about Turbo is that it only turns up the speed on 3 cores and drops the speed on the other 3 cores. This way it would be running at 3.5GHz on all cores.

He's fine with my OCing as long as I stay away from the voltage.

I think the only reason I'm able to do this is because of a "bug" in the Gigabyte bios, because when I update the bios the feature was gone.

Bondo said:

Well the thing with Turbo is that it only turns up the speed on 3 cores and drops the speed on 3. This was it would be running at 3.5GHz on all cores.

He's fine with my OCing as long as I stay away from the voltage.

I think the only reason I'm able to do this is because of a "bug" in the Gigabyte bios, because when I update the bios the feature was gone.

Click to expand...

Well if he is cool with it heck yes!

(It shouldn't drop three cores FYI, my 1090t will randomly bounce a few cores up to 3.6, but the rest (if under load) stay at 3.2. They will drop down to 800mhz, but that is a function of CE1, not turbo, and thats only if those cores are not in use at all.)

Bondo said:

The thing that I read about Turbo is that it only turns up the speed on 3 cores and drops the speed on the other 3 cores. This way it would be running at 3.5GHz on all cores.

He's fine with my OCing as long as I stay away from the voltage.

I think the only reason I'm able to do this is because of a "bug" in the Gigabyte bios, because when I update the bios the feature was gone.

Click to expand...

The thing is you can probably "overclock" it to 3.5 GHz using the stock vCore regardless of the BIOS version. Most mild overclocks like that don't need a voltage bump. You can probably give the cpuNB a little nudge, too, without increasing it's voltage and that will help both L3 and RAM performance.

Those factory settings are meant to work for every CPU that comes out of the factory - including those that barely pass the testing. By shear odds the CPU you have is an "average" chip (better than those that barely passed) and will run faster with those factory voltage settings ...

Bad news. The reason this option was missing in later bios editions is because it doesn't actually do anything. I just got the copy of Win 7 Pro that I'm loading on the computer. Sure enough, even though it's saying that it's at a 17.5x multiplier in the bios. It's only at a 15x once I get into Windows.....I kind of figure this was too good to be true.

Any ideas of where I can go from here? I'd like to take it to around 3.5GHz w/o Turbo and then possibly 4GHz with Turbo on.

You want to keep the "stock" voltages, right? That's where I'd start. Go into BIOS and set the vCore and cpuNB VID manually to 1.35v and 1.15v, respectively. Make sure you don't confuse the cpuNB voltage (often labeled as CPU NB VID) with the NB (chipset) voltage, which never includes the "CPU" portion. I would also set the RAM voltage to 1.55v just to make sure it's not coming up short. You'll need to find the RAM speed setting and drop that down one notch to 533 (or 1066, however it's labeled). You can go back to the RAM later to optimize it.

From there it's just a matter of increasing the clock speed upward from 200 MHz and testing at each step to make sure it's (somewhat) stable. I'd use a 10 MHz step for the first couple of steps then start using a 5 MHz step. After each clock increase run OCCT or Prime95+CoreTemp for 5-10 minutes and check the core temp near the end of the test. With "stock" voltages you shouldn't have a temp problem but it's always good to check. The system shouldn't go above 55°C on the load core temp.

At some point P95 or OCCT will fail. When you get to that point, since you're not raising the voltage, you're done. Back the clock down to the last stable setting you had and test again for an hour or more. Give that a try to see how that works for you then let us know how it turned out.


PS
Be sure to turn off Turbo, CoolNQuiet, and C1E in BIOS. Also, make sure Windows is set to Performance instead of Power Saving ...

Be sure and check your vCore, Gigabyte is known for setting it higher than other mb manufactures; My Gigabyte is 1.41 vCore stock, my Asus is 1.31 vCore stock. I had my 1055T at 223 mhz fsb in my asus for months without ever turning up any voltages and never had a problem. That gave me 3.1 Ghz/ 3.4 Ghz Turbo (mild but noticable) on a 2.8 Ghz chip with no extra heat.

Bondo said:

Bad news. The reason this option was missing in later bios editions is because it doesn't actually do anything. I just got the copy of Win 7 Pro that I'm loading on the computer. Sure enough, even though it's saying that it's at a 17.5x multiplier in the bios. It's only at a 15x once I get into Windows.....I kind of figure this was too good to be true.

Click to expand...

The 17.5x is what it can do in Turbo mode, guess they forgot to edit it out of the Bios screen.

Alright it looks like we are stable at an FSB of 220. Which brings me to 3.3GHz. I was running both P95 and OCCT at the same time. Not sure if this was a way to stress it more or not. It appears that my 5th core is the weak link.

Should I have lowered my HT-Link and NB to stay near 2000MHz?

Also I did leave Turbo enabled, not sure if this is going to make it less stable or not. Is there some way I can test I can test this to see if it will be stable once it clocks 3 of the cores up to 3.8GHz?

You should probably run Prime95 or OCCT - but it doesn't hurt anything to run them both.

You can try that to see if it will help. The X6's seem to like the HT Link and cpuNB the same speed, though, so if you lower one lower them both.

Hmmmm, If you started P95 and closed three workers (and nothing else is running on your machine) that SHOULD make it switch to Turbo. If that doesn't work you might try shutting down the fourth worker leaving only two ...

Which of the P95 tests should I be running? I did the one that said maximum heat. It's In-place large FFT's. Currently it's been running P95 for about 19 hours no issues.

When I ran both OCCT and P95 OCCT said there was an error on Core One after 75 minutes.

I tried only running 2 or 3 workers in P95, but it kept switching them around so nothing ever got to 100%. I saw a couple of cores jump to around 3.5-3.6 GHz, but they mainly stayed at around 3.45GHz. I saw this changes by looking in the status monitor in the AMD OverDrive Utility. CoreTemp didn't show any change.

That's why I give people both of them. Some machines won't run P95, others won't run OCCT - and I have no clue why. I usually use small FFTs for the quick test but it's not that critical - large FFTs is OK too.

For a longer test like yours I'd use Blend because it also stresses the RAM ...

Good to know.
Ok so running P95 only three workers worked for almost 90 minutes. Then one of the workers failed. Two are still going. Should I clock it back a little still? Or would it be safe to up the voltage a little? I talked to my friend and he's ok pushing it a bit as long as it's nothing crazy. Currently I have it at 1.35 which is down .75 from where Gibabyte had it set for auto.

Currently my hypertransport is is set to 2.2GHz. My NB is also at 2.2GHz.

Let me know how to proceed when you get a chance. I appreciate the help.

Were they running at Turbo speed?

AMD says your safe voltage range is 1.0-1.475v so you've got some play.

You still need to watch those temps, though ...

They weren't exactly running at Turbo, but Turbo mode is enabled. They would clock past 3.3GHZ, but never all the way up to 3.8GHz. P95 is spreading the 3 test load over all 6 cores.

Are the only temps I need to worry about the core temps? Or is the another measure that I need to watch. If it's just the cores then I'm a high of around 30 Celsius.

I'm running P95 on Blend right now on 3 workers. It's still getting spread over all 6 cores, but I want to see if I can get it to fail again.

Should I be going anything with my RAM yet?

Running three workers isn't a good stability test. You asked how to get the cores running at Turbo speed and that was my suggestion. I've got an old copy of P95 I keep around that's not multi-threaded. Instead of having workers running under one program you have several copies of the program and you can set the Affinity (which core it runs on) for each copy. If you could dig one of those up on-line that might solve the problem.

Otherwise, for a full stress test and to check load core temp, you need to run all six workers ...

If you have workers failing, especially #5, bump your voltage up one notch and try prime95 again with all 6 workers running, repeat until you have no more failing workers. *If you want to see if turbo is working you can watch it with AMD Overdrive when no test is running or with CPU-Z.

Well things are looking pretty good over here. I was able to run P95 (on all 6 cores) for 72 hours straight with not even a hicup. The core temp reached a max of 30C with an average of 25C.
If I ran only 3 works P95 would have an error after around 6 hours. I dropped my NB and hypertransport speeds down to around 1980MHz and now the 3 works last from between 12 hours to 24 hours.

I still have turbo enabled, but I think with these results show that even with Turbo the system is sufficiently stable.

What do I need to do to tighten up my RAM timings?
Any other things that I should do to this system?

Bondo said:

What do I need to do to tighten up my RAM timings?
Any other things that I should do to this system?

Click to expand...

It's kind of hard to answer that question without knowing the RAM or it's timings/voltage/speed. If you can post/attach your CPU-Z Memory and SPD tabs it would help ...