Does this O/C look ok?

winky666 · Nov 23, 2011

hi guys this is my first post
iv just had my first attempt at overclocking my cpu and it seems stable,im just wanting a heads up if i have gone the right way about it?
heres the link to my cpuz http://valid.canardpc.com/show_oc.php?id=2112770

my temps are 46c under load, if anybody has any tips for tweaking or am i just killing my cpu? pleease let me know regards

Janus67 · Nov 23, 2011

If your temps are only 46C under 100% load that is quite amazing. What are you using for cooling and what are you using to put the CPU under load?

winky666 · Nov 23, 2011

Janus67 said:
If your temps are only 46C under 100% load that is quite amazing. What are you using for cooling and what are you using to put the CPU under load?

Ek supreme water block, 240mm rad, can't remember the pump off the top of my head, it's all in a antec 1200 case, I'm using prime 95 and core temp for the test mate

notJUSTguitar · Nov 23, 2011

That looks very nice. Looks like you could go higher too if you wanted.

winky666 · Nov 23, 2011

I was just a little bit unsure about only using the multipyer to o/c??

notJUSTguitar · Nov 23, 2011

The multiplier is the easiest way to OC Black edition CPUs (or intel K series).

winky666 · Nov 24, 2011

Cheers, I had prime 95 going for 4 hrs no hick ups and temps peaked at 50 c, I don't want to touch the fsb because it under clocks my memory, what voltage will this chip handle?

Aquaman78 · Nov 24, 2011

winky666 said:
Cheers, I had prime 95 going for 4 hrs no hick ups and temps peaked at 50 c, I don't want to touch the fsb because it under clocks my memory, what voltage will this chip handle?

You can get that chip to 1.5v (maybe 1.55v) as long as your cooler keeps the temps in check...which it looks like it is doing.

manu2b · Nov 24, 2011

Aquaman78 said:
You can get that chip to 1.5v (maybe 1.55v) as long as your cooler keeps the temps in check...which it looks like it is doing.

Doesn't depend on the cpu temp only.
Do you have heatsink on the mosfet? If not, above 1.5v, they can go... Plop!

marjamar · Nov 24, 2011

If your memory timings will allow it, I'd work on getting buss speed up to, or over 300 MHz. Even if your CPU timings are lower, you will see better benchmarks with the higher buss speed.

On the same MB, my 1090T was stable at 4.17GHz with a buss speed of 305 MHz. Got this using air cooling and ran really stable for extended Prime95 runs (See below).

-Rodger

trents · Nov 24, 2011

You should be able to get the CPUNB ("NB" in CPU-z) up to 2600 mhz with a little extra CPUNB voltage 1.25-1.3 is usually sufficient. On the Denebs, leave the HT Link speed and voltage at stock. A higher CPUNB frequency improves memory performance.

There is no advantage to increasing the bus speed with an unlocked multiplier CPU as all the things that increase with jacking up the bus speed can be independently manipulated with a multiplier unlocked CPU. Doing it with bus speed just adds to the potential for instability because any given motherboard has a ceiling for fsb speed increase, fsb being sort of the metronome for the system. Don't speed up the metronome unless you have to as it carries no performance force of its own.

marjamar · Nov 24, 2011

trents said:
You should be able to get the CPUNB ("NB" in CPU-z) up to 2600 mhz with a little extra CPUNB voltage 1.25-1.3 is usually sufficient. On the Denebs, leave the HT Link speed and voltage at stock. A higher CPUNB frequency improves memory performance.

There is no advantage to increasing the bus speed with an unlocked multiplier CPU as all the things that increase with jacking up the bus speed can be independently manipulated with a multiplier unlocked CPU. Doing it with bus speed just adds to the potential for instability because any given motherboard has a ceiling for fsb speed increase, fsb being sort of the metronome for the system. Don't speed up the metronome unless you have to as it carries no performance force of its own.

Sorry, but I can't see how this is true. FSB is the pipeline which moves data to and from CPU and memory controller. How this isn't fundamental to overall throughput would be a mystery to me.

Also, even if it were true that adjustments outside of increasing buss speeds do the same thing, why would you wish to attempt this when simply adjusting the buss already does this.

I'd suggest you provide some examples here, as obviously I could be wrong in my thinking, and it would help me to understand how this is happening and why.

Thanks.

-Rodger

winky666 · Nov 24, 2011

manu2b said:
Doesn't depend on the cpu temp only.
Do you have heatsink on the mosfet? If not, above 1.5v, they can go... Plop!

my boards a totally standard Asus crosshair iv formula mate, im new to this o/cing, where would i get heatsinks from? im in the uk

trents said:
You should be able to get the CPUNB ("NB" in CPU-z) up to 2600 mhz with a little extra CPUNB voltage 1.25-1.3 is usually sufficient. On the Denebs, leave the HT Link speed and voltage at stock. A higher CPUNB frequency improves memory performance..

one a go anything above 1.25v the number turns red

Thanks for the reply's guys much appreciated

trents · Nov 25, 2011

marjamar said:
Sorry, but I can't see how this is true. FSB is the pipeline which moves data to and from CPU and memory controller. How this isn't fundamental to overall throughput would be a mystery to me.

Also, even if it were true that adjustments outside of increasing buss speeds do the same thing, why would you wish to attempt this when simply adjusting the buss already does this.

I'd suggest you provide some examples here, as obviously I could be wrong in my thinking, and it would help me to understand how this is happening and why.

Thanks.

-Rodger

Examples?

I think that looking at the fsb as a data pipeline is a misconception. No data flows down the fsb. It is a logical entity, not a physical one.

The reason for not using the fsb when you have an unlocked multiplier CPU is that it allows you to 1. overclock one component at a time rather than all at once. This eliminates the need to have to go back and lower the multipliers of the various components such as CPU, ram, HT Link and CPUNB while trying to find the limit of anyone of them. And 2. It avoids the problem of exceeding the capacity of the motherboard to keep up with the ever-quickening pace of the "metronome", to refer to my earlier analogy, and, therefore eliminates another variable that can cause instability in overclocking. To prove this I would point to the fact that you can down-clock all the other overclocking components by lowering their multipliers below stock and if you just keep increasing the fsb you will encounter instability. Now obviously in this situation you cannot point to the fact that the fsb bandwidth can't keep up with the increasing data load because the load is actually less than it was at stock and the pipeline (if there was one) is bigger. Yet, instability still occurs because the metronome can only go so fast without getting out of sync. With higher quality motherboards the better components permit the system metronome to beat faster without getting out of time.

So, let me put the ball in your court. I would challenge you to overclock a computer without changing the fsb (using only the component multipliers) and then overclock the same computer by increasing the fsb instead of the individual component multipliers.Then do some benching to see if there is actually a difference.

manu2b · Nov 25, 2011

winky666 said:
my boards a totally standard Asus crosshair iv formula mate, im new to this o/cing, where would i get heatsinks from? im in the uk

one a go anything above 1.25v the number turns red

Thanks for the reply's guys much appreciated

No problem with your board then! If your temps are in line, 1.55V is the max that you should push your CPU.
Don't bother about the red! 1.3v is still OK.

But as said earlier, all cpu's are differents. I need 1.2v for a 2640MHz CPU-NB, some need 1.175 and other 1.275...

marjamar · Nov 25, 2011

trents said:
Examples?

I think that looking at the fsb as a data pipeline is a misconception. No data flows down the fsb. It is a logical entity, not a physical one.

The reason for not using the fsb when you have an unlocked multiplier CPU is that it allows you to 1. overclock one component at a time rather than all at once. This eliminates the need to have to go back and lower the multipliers of the various components such as CPU, ram, HT Link and CPUNB while trying to find the limit of anyone of them. And 2. It avoids the problem of exceeding the capacity of the motherboard to keep up with the ever-quickening pace of the "metronome", to refer to my earlier analogy, and, therefore eliminates another variable that can cause instability in overclocking. To prove this I would point to the fact that you can down-clock all the other overclocking components by lowering their multipliers below stock and if you just keep increasing the fsb you will encounter instability. Now obviously in this situation you cannot point to the fact that the fsb bandwidth can't keep up with the increasing data load because the load is actually less than it was at stock and the pipeline (if there was one) is bigger. Yet, instability still occurs because the metronome can only go so fast without getting out of sync. With higher quality motherboards the better components permit the system metronome to beat faster without getting out of time.

So, let me put the ball in your court. I would challenge you to overclock a computer without changing the fsb (using only the component multipliers) and then overclock the same computer by increasing the fsb instead of the individual component multipliers.Then do some benching to see if there is actually a difference.

Well, it seems we are talking semantics quite a lot here. Since I'm "old school", I'm guilty of continuing to envision CPU architecture in ways that are easier to fathom and less to learn. But, the fact still remains that without using every clock cycle to it's fullest, speeds will be less, throughput will be less. Your approach to "fine tune" components isn't wrong, but to say there isn't any actual FSB (prior to your editing your initial post regarding this, which is what I started my reply to BTW) without mentioning HT or external bus isn't exactly "fair" either. Just because architecture has evolved to allow for higher capacities and speed, referencing older terms for the facility, although not exactly accurate, still makes sense when adjustments to them provide a similar outcome we're after -- Higher clocks -- Higher throughput.

As to your challenge, perhaps we could turn bus speeds to 1 and up clocks/multipliers on other components and still have a fast processor? I don't think so. The most fair way to accept these different types of ideologies is to recognize the usefulness in both. Without question, the better the fine tuning is done at it's most fundamental levels, obviously the greater the rewards will be. But, just as with anything complex, the more involved the complexity, the greater the chance of error when making changes. It would therefore seem wise to factor in the expertise of the one making these changes, would it not.

-Rodger

winky666 · Nov 25, 2011

well my pc has been fine for 2 days played bf3 for a few hrs last night! turned it on today ram p95 just to check everythings ok and my O/C failed give it a little more V and still no joy!

trents · Nov 26, 2011

marjamar said:

Well, it seems we are talking semantics quite a lot here. Since I'm "old school", I'm guilty of continuing to envision CPU architecture in ways that are easier to fathom and less to learn. But, the fact still remains that without using every clock cycle to it's fullest, speeds will be less, throughput will be less. Your approach to "fine tune" components isn't wrong, but to say there isn't any actual FSB (prior to your editing your initial post regarding this, which is what I started my reply to BTW) without mentioning HT or external bus isn't exactly "fair" either. Just because architecture has evolved to allow for higher capacities and speed, referencing older terms for the facility, although not exactly accurate, still makes sense when adjustments to them provide a similar outcome we're after -- Higher clocks -- Higher throughput . . .

Click to expand...

FSB has become HT Reference since the memory controller has been moved onto the CPU die. But that's not the key to my argument so its not just a matter of symantics. Actually, I use the two synonymously (even though its technically incorrect) as do a lot of veteran overclockers. But HT Reference, aka, fsb, is not to be confused with HT Link frequency. What I am saying is that the only purpose of manipulating the fsb is to speed up the the frequency of the ram, the HT Link, the CPUNB and the CPU when there is no other way of doing it, i.e. when you don't have a multiplier unlocked CPU and when you don't have bios options to change those other frequencies individually. Using the fsb just does it in concert. If you like the pipeline analogy then I would say speeding up all those individual component frequencies using their multipliers without touching the "CPU frequency" (the term used by most bioses to refer to the HT Reference or the fsb) speeds up the pipeline without putting extra stress on the system metronome and also results in the same performance increase.

neo668 · Nov 26, 2011

Hi trents,

Do you mean that if it is an unlocked CPU, focusing on the multiplier alone is sufficient? The FSB (HT Ref) can be left alone?

Thank you.

trents · Nov 26, 2011

neo668 said:
Hi trents,

Do you mean that if it is an unlocked CPU, focusing on the multiplier alone is sufficient? The FSB (HT Ref) can be left alone?

Thank you.

I am saying that if you increase the CPU multiplier, the CPUNB, the HT Link and set the ram frequency as individual components you will achieve the same performance effect as if you overclock with the HT Reference and will avoid the issue of running out of head room with the HT Reference. I am not saying that the only performance parameter you need to give attention to is the CPU multiplier. When using the HT Reference to overclock all these other parameters increase in lockstep with it and may be carried to levels that cause them to become sources of instability so that you would need to go back and deal with them individually anyway. Therefore, IMO the only value in using the HT Reference to overclock when you have an unlocked multiplier CPU and a detailed bios is to fine tune the ram speed so that you could set it to a value that fell in between two standard steps; like somewhere between 1333 and 1600.

Does this O/C look ok?

New Member

Benching Team Leader

New Member

Member

New Member

Member

New Member

Member

Member

Member

Attachments

Senior Member

Member

New Member

Senior Member

Member

Member

New Member

Senior Member

Member

Senior Member

Similar threads