Haha... maybe that would be true for you. We each have different uses for our computers.
And I need my computer to work consistently and reliably.
Reason why I asked is that what I've read makes it seem as if ~3.2 GHz is about as high as you can go (given memory limitations and such). Having faster processors in my Mac Pro isn't going to speed up my memory's capabilities. I probably wouldn't be using this sort of program anyway, but it will be interesting to follow.
I think most of those saying it goes up to 3.1 are owners of 2.8s. You already have 3.2 without increasing the memory clock from stock. I'd think that you should easily be able to get to 3.4. This is because a 3.2 has a higher CPU to memory speed multiplier than a 2.8.
I'm usually pretty knowledgeable with technical stuff, but I'm having difficulty understanding why the Mac Pro can't be pushed a lot higher? On the PC side, ever since the release of the original 65nm quad-core parts and especially now with 45nm, people have been pushing 2.4-2.6ghz parts up to 3.6-3.8ghz on air cooling! Is it the Xeon platform's requirement of FB-DIMMs the problem? Or something else that is Mac specific?
Quote:
Originally Posted by JeffDM
I think most of those saying it goes up to 3.1 are owners of 2.8s. You already have 3.2 without increasing the memory clock from stock. I'd think that you should easily be able to get to 3.4. This is because a 3.2 has a higher CPU to memory speed multiplier than a 2.8.
From what I briefly read, this is because the CPU is being overclocked without adjusting the memory speed multiplier, as JeffDM says.
It's actually pretty bloody impressive that the RAM can be overclocked so much, being FBDIMM makes a big difference.
Like right now on my PC, I overclocked my Core2Duo from 2.0ghz to 2.5ghz. The memory is set to run at 667mhz. Overclocking enthusiasts will tell you the Intel 45 and 65 nm's overclock like crazy on air cooling alone. I have a Zalman CPU fan, so the CPU is overclocked 25%. However, the RAM is only rated at 667mhz. If the memory multiplier was kept "the same", I would need RAM that can handle 667+25% = 833 -- that means 800mhz to 1066mhz ddr3 rated RAM, etc. etc.
Thus in the Mac Pro, say a 2.8ghz to 3.2ghz overclock is a 15% overclock of CPU *and RAM that is not meant to be overclocked*.
You *have* to be able to independently change the "RAM divider" in the BIOS somehow. If so, we could see 3.5ghz Mac Pros. Alternatively, you could go from a 2.8ghz to 3.5ghz on "stock memory divider settings" if you had 800mhz+25%= 1066mhz FB-DIMMs that are Mac Pro compatible.
Yessss, the dark arts of teh PC 0verclocking. I have dabbled in it.
Most recently I took the components from my old PC, tried a Dremel for the 1st time ever, and cut up an old PowerMac G5 case [heresy, yes, this is why I call it a dark art] ...and put it in there. CPU still going 2.0ghz to 2.5ghz, RAM is on 667mhz 2gb, and I've got SLI'ed 8500GTs that are factory overclocked and then manually overclocked. 8500GTs run at about 450mhz clock speed, I'm running both at about 668mhz clock, 668mhz vram. Got myself a Dell 22" 1680x1050 from eBay. It is SWEET.
I wish I had a Mac Pro...
I have also started to "document" the results, particularly using nHancer.com profiles which is a bit tricky in and of itself.
... (it was tricky fixing the iWeb2.0~Firefox3.0 bug)
...
...
Note: the "memory divider" settings is another big weird thing to learn of, since you have to choose the setting that brings you the highest RAM speed possible for a particular bus speed overclock, etc. Unless you had some wicked motherboard that can handle independent 1mhz memory +/- increments. I still don't quite understand this part fully, except that the memory speed is always a multiplier of the bus speed. for example 250mhz bus speed (quad pumped) = 1ghz "FSB", 10x CPU multiplier gives you 2.5ghz, RAM to run at around 667mhz means a ram divider ratio of ...etc. etc...
Note2: the other thing is actually, you shouldn't be changing the bus speed of the motherboard and hence RAM, if you can access the "CPU Multiplier" part. That way is the best, so that you overclock the CPU independently. But for Conroes and Penryn desktops, usually only the "Conroe Extreme" and "Penryn Extreme" etc. have "unlocked" multipliers. In my CPU, for example, 10x is the most I can push the CPU before I have to increase the bus speed to increase CPU speed.
I think most of those saying it goes up to 3.1 are owners of 2.8s. You already have 3.2 without increasing the memory clock from stock. I'd think that you should easily be able to get to 3.4. This is because a 3.2 has a higher CPU to memory speed multiplier than a 2.8.
Thanks for the feedback on that. It will be interesting to see how this develops. Hopefully they iron out the program and people get an idea of how it works and what it is capable of. It would be nice to milk some extra performance out of the system without any noteworthy drawbacks, but with how fast it already is I don't see any sense in doing something that could reduce its lifespan.
How much RAM do you have and if more than stock what is the make of the extra? I have 4 GIGs of a so called 'good' make extra, 6 GIGs total, and I cannot get more than 6% increase without a crash. I also have two video cards, a stock card + 8800 so that could be an issue. This on an 8 Core 2.8 very new Mac Pro.
From what I briefly read, this is because the CPU is being overclocked without adjusting the memory speed multiplier, as JeffDM says.
I have a Zalman CPU fan, so the CPU is overclocked 25%. However, the RAM is only rated at 667mhz. If the memory multiplier was kept "the same", I would need RAM that can handle 667+25% = 833 -- that means 800mhz to 1066mhz ddr3 rated RAM, etc. etc.
*snip *
Thus in the Mac Pro, say a 2.8ghz to 3.2ghz overclock is a 15% overclock of CPU *and RAM that is not meant to be overclocked*.
*snip *
You *have* to be able to independently change the "RAM divider" in the BIOS somehow. If so, we could see 3.5ghz Mac Pros. Alternatively, you could go from a 2.8ghz to 3.5ghz on "stock memory divider settings" if you had 800mhz+25%= 1066mhz FB-DIMMs that are Mac Pro compatible.
Thank you very much for the info. So, getting this straight, the main clock you set is the front-side bus clock. The CPU frequency is then set as a multiple (2x ... 6x) of that base FSB clock. The SDRAM also has it's own multiple of the FSB clock as well.
If you don't have one of the Core 2 "extreme" chips, then the CPU multiplier is locked, and thus if you wish to overclock the CPU, you actually have to adjust the base front-side-bus clock, which has the effect of increasing the frequency for the FSB, the CPU, and the RAM. In the case of the Mac Pro, the CPU multiplier is locked, so the reason that the people cannot overclock it much is because of the secondary effect of increasing the memory frequency, and memory can only be overclocked a little bit before failing. And in this case, the FB-DIMMs are a lot less flexible than DDR2/DDR3 SDRAM that is used with all Non-Xeon chips...
If the above is all correct, then tell me what the consequences of the Mac Pro moving to Nehalem will be. Obviously the front-side-bus is gone, and the memory controller sits on the CPU. What acts as the base clock for the CPU multiplier if there is no FSB? Are all nehalems going to be able to increase CPU speed without having to increase memory speed, or will it again only be the "extreme" additions? Also, the dual-processor Nehalem Xeon will NOT be using FB-DIMM any more, and is going back to DDR3. In the case that the future CPUs can only be overclocked by also overclocking the RAM, wouldn't this be really good since DDR3 can run really fast?
How much RAM do you have and if more than stock what is the make of the extra? I have 4 GIGs of a so called 'good' make extra, 6 GIGs total, and I cannot get more than 6% increase without a crash. I also have two video cards, a stock card + 8800 so that could be an issue. This on an 8 Core 2.8 very new Mac Pro.
have exactly the same computer model, video card etc except I only have 4 GB RAM, wich was put in by Apple when they configured it
oh wait, I also only have one video card, the 8800 GT
How much RAM do you have and if more than stock what is the make of the extra? I have 4 GIGs of a so called 'good' make extra, 6 GIGs total, and I cannot get more than 6% increase without a crash. I also have two video cards, a stock card + 8800 so that could be an issue. This on an 8 Core 2.8 very new Mac Pro.
Usually the GPUs don't matter so much since they have their default clocks set which is quite independent of the FSB. The PCI Express bus clock usually runs at 100mhz and is not affected by the Mac Pro overclock utility, as far as I can tell. As such, I think the issue you are facing is the mixed RAM environment. Usually, if you mix Apple and non-Apple RAM, or even different brands, this can affect RAM overclocks. You can try one set of sticks at a time and see if your overclocks go further.
Quote:
Originally Posted by winterspan
Thank you very much for the info. So, getting this straight, the main clock you set is the front-side bus clock. The CPU frequency is then set as a multiple (2x ... 6x) of that base FSB clock. The SDRAM also has it's own multiple of the FSB clock as well...
Yes, that's my understanding of things... No worries, I'm happy to share my knowledge in the non-hardcore-overclockers scene.
Quote:
Originally Posted by winterspan
If you don't have one of the Core 2 "extreme" chips, then the CPU multiplier is locked...
It is "locked" in the sense that there is a maximum multiplier, yes...
Quote:
Originally Posted by winterspan
...if you wish to overclock the CPU, you actually have to adjust the base front-side-bus clock, which has the effect of increasing the frequency for the FSB, the CPU, and the RAM...
Yeah, the ZDNet screenshot seems to show this.
Quote:
Originally Posted by winterspan
...And in this case, the FB-DIMMs are a lot less flexible than DDR2/DDR3 SDRAM that is used with all Non-Xeon chips...
Actually, FB-DIMMS I would say are *more* flexible. We're looking at RAM being overclocked by up to 15% or so, well over the specification. The error correction of FB-DIMMs come in very useful here. Most PC RAM can't handle more than 5% over spec.
DDR2/DDR3 is useful in PC Overclocking because you can increase your bus/CPU speed without having to mess with the dividers too much. 1066mhz RAM, for example, gives a lot of "overhead".
It seems that the *memory divider* is the most important part, which is in the BIOS, which the Mac Pro utility can't seem to access yet. So the memory divider is the one that will determine what RAM speed the motherboard/memory controller(?) is driving the RAM at.
As suggested above, the easiest way to drive the Mac Pros to 3.5ghz with the utility is to try out 1066mhz FB-DIMMs. Given stock memory dividers, this should give a lot of "overhead" for the motherboard driving the RAM above 800mhz up a few hundred mhz.
Quote:
Originally Posted by winterspan
...If the above is all correct, then tell me what the consequences of the Mac Pro moving to Nehalem will be. Obviously the front-side-bus is gone, and the memory controller sits on the CPU. What acts as the base clock for the CPU multiplier if there is no FSB? Are all nehalems going to be able to increase CPU speed without having to increase memory speed, or will it again only be the "extreme" additions? Also, the dual-processor Nehalem Xeon will NOT be using FB-DIMM any more, and is going back to DDR3. In the case that the future CPUs can only be overclocked by also overclocking the RAM, wouldn't this be really good since DDR3 can run really fast?...Thanks in advance...
That's a really good question. I think even with Nehalem, there will have to be some sort of BIOS controls in place. That is, CPU clock, bus clock, memory clock controllers. In this case it will depend a lot on what Intel does, and what motherboard and BIOS manufacturers do. It would probably be similar to what we have now, though if you follow TomsHardware.com for example one can get down to the nitty-gritty.
As mentioned, the CPU multiplier limit is annoying, but workable if there is memory divider settings. That's why boards are marketed as "overclocking" motherboards. There is fine control over many esoteric settings like special memory divider setting control.
Again, it's not so much an issue of FB-DIMM vs DDR3. It's the speed at which they are rated. And right now, if you had a FB-DIMM that's rated at 1333mhz, that would wipe the floor of a 1066mhz DDR3 RAM, particularly with error correction.
It is interesting that Nehalem Xeons are dropping FB-DIMM, this is perhaps a cost issue, marketing issue, making Xeons more accessible to "commodity" RAM rather than "server" RAM... I think it is because Intel is confident that if the memory controller goes into the CPU itself, it can handle a huge amount of error correction just using non-ECC(?) DDR3 RAM and still be suitable for server-grade applications.
The big question I have is the next section of overclocking, which I haven't delved into yet: voltages. Maybe more on this if y'all are interested.
Mainly if a lot of the FSB is in the CPU itself, than rather than FSB and CPU voltages which are usually "separate", we'd be dealing with the CPU voltage mainly. In this case with so much focus on the CPU heat, enthusiasts may be interested in water cooling this, since the only other thing on the motherboard to cool is the southbridge(?).
"...The CPU core, for example, can be running at its own frequency and voltage while the memory system is running on its own and I/O is running on its own and each of them can be tuned for a different segment," Kumar said..."
How Mac OS X can access these fine controls as opposed to PC BIOSes, that is of course a big question mark.
"Intel looks set to follow the release of its 'Bloomfield' 45nm processors - all based on the 'Nehalem' architecture - with a gamer-oriented chipset, the X58.
The chipset's specs aren't known, but it's not hard to guess. The northbridge will be 'Tylersburg', the chip that links to the host processor over Intel's new HyperTransport-like QuickPath Interconnect (QPI) bus.
Tylserburg provides the system with a PCI Express 2.0 bus - memory is managed by the 1366-pin CPU's own, on-board DDR 3 memory controller. Expect X58-based boards to offer four x8 PCIe slots and support AMD's CrossFire X multi-GPU technology.
Intel is believed to be keen to support SLI too - whether it does so will hinge on the outcome of negotiations with Nvidia.
In turn, Tylersburg connects to the ICH10 southbridge, which handles the system I/O - all the customary HD audio, USB, Gigabit Ethernet and SATA ports are provided.
The X58 should ship in Q4, alongside Bloomfield, which is expected to debut in three version clocked from 2.66GHz up to 3.2GHz. All three CPUs contain 8MB of L3 cache shared across all four cores - each core has its own complement of L1 and L2 cache; 256KB of the latter - and support all the usual Intel extension technologies except TXT (Trusted eXecution Technology). They're all said to have a power and thermal envelope of 130W."
Key points for me: "HyperTransport-like" -- this was the killer advantage AMD64 had a few years ago that made it such a hit in gaming circles while the Pentium 4 floundered badly. Also, 8MB L3 cache across four cores, 130W is not bad given the integrated northbridge.
So whats the point? a well specced machine that works HAS to be better than a slightly faster box .. that wont actually turn on!
It's the "thrill" that overclockers thrive on. For pro apps, Mac Pro usage, obviously this is all not recommended.
For enthusiasts, it's like driving your car through the city streets fast, seeing how fast you can go without crashing and killing someone or yourself.
Although, this is much safer, because the only thing you would destroy is your PC. Perhaps it keeps the kids of the streets too.
For gamers, it's the challenge of "can I play CounterStrike Source for 3 hours at a 25% overclock and +10fps without it crashing and making me look bad to my friends?".
More formally, it relates to overall system stability. This is where you do "burn-in" tests. So in PC land, you would run Prime95 and SuperPI, looped constant Futuremark 3DMark06 or 3DMark Vantage... Basically for a given overclock setting you would peg your CPU and GPU at 100% and see if it lasts 24 hours. If it does, then you overclock is pretty darn stable, and you've got a significant speed increase, greater value of your investment, large satisfaction, a good feeling, and without compromise to stability.
There is the issue of component degradation due to running over spec and at higher voltages. However, given a component is only useful for about 6-12 months (in enthusiasts' eyes) before it is obsolete, if you can push it for 6-12 months, and it craps out in the 2nd year, well, you've got your money's worth.
It is really a very different mindset. This is why I use a MacBook for all my professional work life, and personal life. My PC is for experimenting, gaming, watching videos and other non-essential stuff. If I had a Mac Pro right now, well, I'd be a different person, I think... I'm not in that situation right now that owning a Mac Pro is important. MacBook Pro 512mb vram if I get a certain Apple-related job, perhaps...
At the end of the day, it's all got to do with why people have a beat-up '70s Mustang or something like that in their garage, or a pimped-out Evo. I just have never been good with cars and stuff, and I epic failed the Mecanno set that my parents got for me when I was a kid, can't solder to save my life, and have no clue what you are supposed to do with Lego except for this kind of stuff: http://www.youtube.com/watch?v=sxLvnLI6GlU
Usually the GPUs don't matter so much since they have their default clocks set which is quite independent of the FSB. The PCI Express bus clock usually runs at 100mhz and is not affected by the Mac Pro overclock utility, as far as I can tell. As such, I think the issue you are facing is the mixed RAM environment. Usually, if you mix Apple and non-Apple RAM, or even different brands, this can affect RAM overclocks. You can try one set of sticks at a time and see if your overclocks go further.
Thanks for the input. When you say 'one set of sticks', are you saying I can take out just one 2 GIG stick? I thought they had to remain as pair.
Wait a minute ... Now I think about it the set up was odd to me when I installed, both Apple DIMMS are on one riser and the two 2 GIG 3rd party DIMMS are on the other not alternating as in all Macs I have owned before. So, I can try just one at a time thus having a 4 GIG Mac in both tests. Is this correct?
Even if I never run ZDNet Clock again after this, if it showed up my 3rd party RAM as so poor it was useful. Am I totally incorrect here or does this RAM, if it the cause, that cannot clock more than a 6% increase, have potentially other downsides for me running a high end system like this?
Thanks for the input. When you say 'one set of sticks', are you saying I can take out just one 2 GIG stick? I thought they had to remain as pair.
Wait a minute ... Now I think about it the set up was odd to me when I installed, both Apple DIMMS are on one riser and the two 2 GIG 3rd party DIMMS are on the other not alternating as in all Macs I have owned before. So, I can try just one at a time thus having a 4 GIG Mac in both tests. Is this correct?
Even if I never run ZDNet Clock again after this, if it showed up my 3rd party RAM as so poor it was useful. Am I totally incorrect here or does this RAM, if it the cause, that cannot clock more than a 6% increase, have potentially other downsides for me running a high end system like this?
RAM set up: 2x 1 GIG Apple + 2 x 2 GIG 3rd party
Yes, so 1 set would be 1 pair. For example, take out the 2 3rd party sticks of RAM. Try the overclock. Put it back, take out the Apple sticks. Try to overclock.
Let us know how it goes.
I wouldn't say that this-or-that RAM is better for overclocking, because they've never been spec'ed or marketed as such. The other thing about overclocking, is... sometimes, it's all a matter of luck. The 3rd party RAM may well overclock better than the Apple RAM...
The RAM should be running well without any overclocking in your high-end system, because it should all be within spec. If your concerned about "bad" RAM, you would need to run your system at stock speeds with very heavy CPU & RAM usage for about 24 hours, if it is still stable, you're fine there.
Best thing is to just get 5-10 DVDs and HandBrake (handbrake.fr) it all... Start to really put your Mac Pro to full work. Handbrake is not a great example for such a high-end system, but hey, you'll have converted your favourite DVDs at the end of this "test"...!
Yes, so 1 set would be 1 pair. For example, take out the 2 3rd party sticks of RAM. Try the overclock. Put it back, take out the Apple sticks. Try to overclock.
Let us know how it goes.
I wouldn't say that this-or-that RAM is better for overclocking, because they've never been spec'ed or marketed as such. The other thing about overclocking, is... sometimes, it's all a matter of luck. The 3rd party RAM may well overclock better than the Apple RAM...
The RAM should be running well without any overclocking in your high-end system, because it should all be within spec. If your concerned about "bad" RAM, you would need to run your system at stock speeds with very heavy CPU & RAM usage for about 24 hours, if it is still stable, you're fine there.
Best thing is to just get 5-10 DVDs and HandBrake (handbrake.fr) it all... Start to really put your Mac Pro to full work. Handbrake is not a great example for such a high-end system, but hey, you'll have converted your favourite DVDs at the end of this "test"...!
OK will do later, thanks much for advice. I bet the 3rd party RAM is barely up to spec and is the problem.
OK will do later, thanks much for advice. I bet the 3rd party RAM is barely up to spec and is the problem.
Maybe it's a matter of viewpoint, but because it's not as overclockable doesn't mean it's not good. I've not seen any errors, correctable or otherwise, with my OWC memory at stock clock. I can't say anything about overclocking as I'm only using Tiger on a first gen Mac Pro, neither of which are compatible yet, if ever.
I guess the aftermarket memory I bought doesn't do well with over clocking..I have a macpro 2.8 quad 2008 model and have 8gb of Mushkin memory in it. Though on the PC world Mushkin has some of better overclocking modules out there....perhaps they didn't incorporate that into their FB-DIMMS.
Yeah, it is sometimes annoying when you can't overclock. However as JeffDM says, within specified normal speeds, if everything works well then the RAM should be alright overall. The ZDNet article does hint that "cheaper" memory doesn't overclock that well... It's hard to say for sure, because FB-DIMMs have never really been marketed as "overclocking" memory...
Try with just the stock Apple RAM that came with the machine and see if your overclock works better, to see if you can isolate the 3rd party RAM as being problematic.
Waiting for new memory to arrive, but with 2x1gb Apple memory, I can run the zdnet clock utility (conservatively) - typically at 3.1 with no errors. I suppose I can try higher bus freq, but thought a few simple tests at this speed would be interesting.
However, running either cinebench 10, or Xbench 1.3, I see absolutely no performance increase? This compared with the tool set back to 400 mhz and the application off.
And yes, I am careful to hit 'apply' to the zdnet clock tool.
The system log does seem to suggest that the bus freq has been increased....
As the author wrote, its dynamically loaded, yes?, and not necessary to reboot.
Waiting for new memory to arrive, but with 2x1gb Apple memory, I can run the zdnet clock utility (conservatively) - typically at 3.1 with no errors. I suppose I can try higher bus freq, but thought a few simple tests at this speed would be interesting.
However, running either cinebench 10, or Xbench 1.3, I see absolutely no performance increase? This compared with the tool set back to 400 mhz and the application off.
And yes, I am careful to hit 'apply' to the zdnet clock tool.
The system log does seem to suggest that the bus freq has been increased....
As the author wrote, its dynamically loaded, yes?, and not necessary to reboot.
As a side effect overclocking Mac OS X causes the real time clock to run faster. On a MacPro3,1 a warm reboot solves this problem.
Please read our FAQ, especially the following questions:
Is there a chance that my computer will not boot after running ZDNet clock?
Benchmarks, for example XBench, Cinebench and Geekbench, do not run faster after overclocking.
Why does the system clock run faster? Is there anything I can do?
Comments
Haha... maybe that would be true for you. We each have different uses for our computers.
And I need my computer to work consistently and reliably.
Reason why I asked is that what I've read makes it seem as if ~3.2 GHz is about as high as you can go (given memory limitations and such). Having faster processors in my Mac Pro isn't going to speed up my memory's capabilities. I probably wouldn't be using this sort of program anyway, but it will be interesting to follow.
I think most of those saying it goes up to 3.1 are owners of 2.8s. You already have 3.2 without increasing the memory clock from stock. I'd think that you should easily be able to get to 3.4. This is because a 3.2 has a higher CPU to memory speed multiplier than a 2.8.
I'm usually pretty knowledgeable with technical stuff, but I'm having difficulty understanding why the Mac Pro can't be pushed a lot higher? On the PC side, ever since the release of the original 65nm quad-core parts and especially now with 45nm, people have been pushing 2.4-2.6ghz parts up to 3.6-3.8ghz on air cooling! Is it the Xeon platform's requirement of FB-DIMMs the problem? Or something else that is Mac specific?
I think most of those saying it goes up to 3.1 are owners of 2.8s. You already have 3.2 without increasing the memory clock from stock. I'd think that you should easily be able to get to 3.4. This is because a 3.2 has a higher CPU to memory speed multiplier than a 2.8.
From what I briefly read, this is because the CPU is being overclocked without adjusting the memory speed multiplier, as JeffDM says.
It's actually pretty bloody impressive that the RAM can be overclocked so much, being FBDIMM makes a big difference.
Like right now on my PC, I overclocked my Core2Duo from 2.0ghz to 2.5ghz. The memory is set to run at 667mhz. Overclocking enthusiasts will tell you the Intel 45 and 65 nm's overclock like crazy on air cooling alone. I have a Zalman CPU fan, so the CPU is overclocked 25%. However, the RAM is only rated at 667mhz. If the memory multiplier was kept "the same", I would need RAM that can handle 667+25% = 833 -- that means 800mhz to 1066mhz ddr3 rated RAM, etc. etc.
Thus in the Mac Pro, say a 2.8ghz to 3.2ghz overclock is a 15% overclock of CPU *and RAM that is not meant to be overclocked*.
You *have* to be able to independently change the "RAM divider" in the BIOS somehow. If so, we could see 3.5ghz Mac Pros. Alternatively, you could go from a 2.8ghz to 3.5ghz on "stock memory divider settings" if you had 800mhz+25%= 1066mhz FB-DIMMs that are Mac Pro compatible.
Yessss, the dark arts of teh PC 0verclocking. I have dabbled in it.
Most recently I took the components from my old PC, tried a Dremel for the 1st time ever, and cut up an old PowerMac G5 case [heresy, yes, this is why I call it a dark art] ...and put it in there. CPU still going 2.0ghz to 2.5ghz, RAM is on 667mhz 2gb, and I've got SLI'ed 8500GTs that are factory overclocked and then manually overclocked. 8500GTs run at about 450mhz clock speed, I'm running both at about 668mhz clock, 668mhz vram. Got myself a Dell 22" 1680x1050 from eBay. It is SWEET.
I wish I had a Mac Pro...
I have also started to "document" the results, particularly using nHancer.com profiles which is a bit tricky in and of itself.
http://web.mac.com/sr2008
... (it was tricky fixing the iWeb2.0~Firefox3.0 bug)
...
...
Note: the "memory divider" settings is another big weird thing to learn of, since you have to choose the setting that brings you the highest RAM speed possible for a particular bus speed overclock, etc. Unless you had some wicked motherboard that can handle independent 1mhz memory +/- increments. I still don't quite understand this part fully, except that the memory speed is always a multiplier of the bus speed. for example 250mhz bus speed (quad pumped) = 1ghz "FSB", 10x CPU multiplier gives you 2.5ghz, RAM to run at around 667mhz means a ram divider ratio of ...etc. etc...
Note2: the other thing is actually, you shouldn't be changing the bus speed of the motherboard and hence RAM, if you can access the "CPU Multiplier" part. That way is the best, so that you overclock the CPU independently. But for Conroes and Penryn desktops, usually only the "Conroe Extreme" and "Penryn Extreme" etc. have "unlocked" multipliers. In my CPU, for example, 10x is the most I can push the CPU before I have to increase the bus speed to increase CPU speed.
I think most of those saying it goes up to 3.1 are owners of 2.8s. You already have 3.2 without increasing the memory clock from stock. I'd think that you should easily be able to get to 3.4. This is because a 3.2 has a higher CPU to memory speed multiplier than a 2.8.
Thanks for the feedback on that. It will be interesting to see how this develops. Hopefully they iron out the program and people get an idea of how it works and what it is capable of. It would be nice to milk some extra performance out of the system without any noteworthy drawbacks, but with how fast it already is I don't see any sense in doing something that could reduce its lifespan.
Thanks a lot! My 2,8GHz Mac Pro now runs 3,0 GHz
How much RAM do you have and if more than stock what is the make of the extra? I have 4 GIGs of a so called 'good' make extra, 6 GIGs total, and I cannot get more than 6% increase without a crash. I also have two video cards, a stock card + 8800 so that could be an issue. This on an 8 Core 2.8 very new Mac Pro.
From what I briefly read, this is because the CPU is being overclocked without adjusting the memory speed multiplier, as JeffDM says.
I have a Zalman CPU fan, so the CPU is overclocked 25%. However, the RAM is only rated at 667mhz. If the memory multiplier was kept "the same", I would need RAM that can handle 667+25% = 833 -- that means 800mhz to 1066mhz ddr3 rated RAM, etc. etc.
*snip *
Thus in the Mac Pro, say a 2.8ghz to 3.2ghz overclock is a 15% overclock of CPU *and RAM that is not meant to be overclocked*.
*snip *
You *have* to be able to independently change the "RAM divider" in the BIOS somehow. If so, we could see 3.5ghz Mac Pros. Alternatively, you could go from a 2.8ghz to 3.5ghz on "stock memory divider settings" if you had 800mhz+25%= 1066mhz FB-DIMMs that are Mac Pro compatible.
Thank you very much for the info. So, getting this straight, the main clock you set is the front-side bus clock. The CPU frequency is then set as a multiple (2x ... 6x) of that base FSB clock. The SDRAM also has it's own multiple of the FSB clock as well.
If you don't have one of the Core 2 "extreme" chips, then the CPU multiplier is locked, and thus if you wish to overclock the CPU, you actually have to adjust the base front-side-bus clock, which has the effect of increasing the frequency for the FSB, the CPU, and the RAM. In the case of the Mac Pro, the CPU multiplier is locked, so the reason that the people cannot overclock it much is because of the secondary effect of increasing the memory frequency, and memory can only be overclocked a little bit before failing. And in this case, the FB-DIMMs are a lot less flexible than DDR2/DDR3 SDRAM that is used with all Non-Xeon chips...
If the above is all correct, then tell me what the consequences of the Mac Pro moving to Nehalem will be. Obviously the front-side-bus is gone, and the memory controller sits on the CPU. What acts as the base clock for the CPU multiplier if there is no FSB? Are all nehalems going to be able to increase CPU speed without having to increase memory speed, or will it again only be the "extreme" additions? Also, the dual-processor Nehalem Xeon will NOT be using FB-DIMM any more, and is going back to DDR3. In the case that the future CPUs can only be overclocked by also overclocking the RAM, wouldn't this be really good since DDR3 can run really fast?
Thanks in advance...
How much RAM do you have and if more than stock what is the make of the extra? I have 4 GIGs of a so called 'good' make extra, 6 GIGs total, and I cannot get more than 6% increase without a crash. I also have two video cards, a stock card + 8800 so that could be an issue. This on an 8 Core 2.8 very new Mac Pro.
have exactly the same computer model, video card etc except I only have 4 GB RAM, wich was put in by Apple when they configured it
oh wait, I also only have one video card, the 8800 GT
the possibility exists that the system won't start up correctly --
So whats the point?
a well specced machine that works HAS to be better than a slightly faster box .. that wont actually turn on!
How much RAM do you have and if more than stock what is the make of the extra? I have 4 GIGs of a so called 'good' make extra, 6 GIGs total, and I cannot get more than 6% increase without a crash. I also have two video cards, a stock card + 8800 so that could be an issue. This on an 8 Core 2.8 very new Mac Pro.
Usually the GPUs don't matter so much since they have their default clocks set which is quite independent of the FSB. The PCI Express bus clock usually runs at 100mhz and is not affected by the Mac Pro overclock utility, as far as I can tell. As such, I think the issue you are facing is the mixed RAM environment. Usually, if you mix Apple and non-Apple RAM, or even different brands, this can affect RAM overclocks. You can try one set of sticks at a time and see if your overclocks go further.
Thank you very much for the info. So, getting this straight, the main clock you set is the front-side bus clock. The CPU frequency is then set as a multiple (2x ... 6x) of that base FSB clock. The SDRAM also has it's own multiple of the FSB clock as well...
Yes, that's my understanding of things...
If you don't have one of the Core 2 "extreme" chips, then the CPU multiplier is locked...
It is "locked" in the sense that there is a maximum multiplier, yes...
...if you wish to overclock the CPU, you actually have to adjust the base front-side-bus clock, which has the effect of increasing the frequency for the FSB, the CPU, and the RAM...
Yeah, the ZDNet screenshot seems to show this.
...And in this case, the FB-DIMMs are a lot less flexible than DDR2/DDR3 SDRAM that is used with all Non-Xeon chips...
Actually, FB-DIMMS I would say are *more* flexible. We're looking at RAM being overclocked by up to 15% or so, well over the specification. The error correction of FB-DIMMs come in very useful here. Most PC RAM can't handle more than 5% over spec.
DDR2/DDR3 is useful in PC Overclocking because you can increase your bus/CPU speed without having to mess with the dividers too much. 1066mhz RAM, for example, gives a lot of "overhead".
It seems that the *memory divider* is the most important part, which is in the BIOS, which the Mac Pro utility can't seem to access yet. So the memory divider is the one that will determine what RAM speed the motherboard/memory controller(?) is driving the RAM at.
As suggested above, the easiest way to drive the Mac Pros to 3.5ghz with the utility is to try out 1066mhz FB-DIMMs. Given stock memory dividers, this should give a lot of "overhead" for the motherboard driving the RAM above 800mhz up a few hundred mhz.
...If the above is all correct, then tell me what the consequences of the Mac Pro moving to Nehalem will be. Obviously the front-side-bus is gone, and the memory controller sits on the CPU. What acts as the base clock for the CPU multiplier if there is no FSB? Are all nehalems going to be able to increase CPU speed without having to increase memory speed, or will it again only be the "extreme" additions? Also, the dual-processor Nehalem Xeon will NOT be using FB-DIMM any more, and is going back to DDR3. In the case that the future CPUs can only be overclocked by also overclocking the RAM, wouldn't this be really good since DDR3 can run really fast?...Thanks in advance...
That's a really good question. I think even with Nehalem, there will have to be some sort of BIOS controls in place. That is, CPU clock, bus clock, memory clock controllers. In this case it will depend a lot on what Intel does, and what motherboard and BIOS manufacturers do. It would probably be similar to what we have now, though if you follow TomsHardware.com for example one can get down to the nitty-gritty.
As mentioned, the CPU multiplier limit is annoying, but workable if there is memory divider settings. That's why boards are marketed as "overclocking" motherboards. There is fine control over many esoteric settings like special memory divider setting control.
Again, it's not so much an issue of FB-DIMM vs DDR3. It's the speed at which they are rated. And right now, if you had a FB-DIMM that's rated at 1333mhz, that would wipe the floor of a 1066mhz DDR3 RAM, particularly with error correction.
It is interesting that Nehalem Xeons are dropping FB-DIMM, this is perhaps a cost issue, marketing issue, making Xeons more accessible to "commodity" RAM rather than "server" RAM... I think it is because Intel is confident that if the memory controller goes into the CPU itself, it can handle a huge amount of error correction just using non-ECC(?) DDR3 RAM and still be suitable for server-grade applications.
The big question I have is the next section of overclocking, which I haven't delved into yet: voltages. Maybe more on this if y'all are interested.
Mainly if a lot of the FSB is in the CPU itself, than rather than FSB and CPU voltages which are usually "separate", we'd be dealing with the CPU voltage mainly. In this case with so much focus on the CPU heat, enthusiasts may be interested in water cooling this, since the only other thing on the motherboard to cool is the southbridge(?).
http://www.eweek.com/c/a/Infrastruct...-into-Nehalem/
"...The CPU core, for example, can be running at its own frequency and voltage while the memory system is running on its own and I/O is running on its own and each of them can be tuned for a different segment," Kumar said..."
How Mac OS X can access these fine controls as opposed to PC BIOSes, that is of course a big question mark.
As a gamer, this is where the beef is for me:
http://www.channelregister.co.uk/200...l_x58_details/
"Intel looks set to follow the release of its 'Bloomfield' 45nm processors - all based on the 'Nehalem' architecture - with a gamer-oriented chipset, the X58.
The chipset's specs aren't known, but it's not hard to guess. The northbridge will be 'Tylersburg', the chip that links to the host processor over Intel's new HyperTransport-like QuickPath Interconnect (QPI) bus.
Tylserburg provides the system with a PCI Express 2.0 bus - memory is managed by the 1366-pin CPU's own, on-board DDR 3 memory controller. Expect X58-based boards to offer four x8 PCIe slots and support AMD's CrossFire X multi-GPU technology.
Intel is believed to be keen to support SLI too - whether it does so will hinge on the outcome of negotiations with Nvidia.
In turn, Tylersburg connects to the ICH10 southbridge, which handles the system I/O - all the customary HD audio, USB, Gigabit Ethernet and SATA ports are provided.
The X58 should ship in Q4, alongside Bloomfield, which is expected to debut in three version clocked from 2.66GHz up to 3.2GHz. All three CPUs contain 8MB of L3 cache shared across all four cores - each core has its own complement of L1 and L2 cache; 256KB of the latter - and support all the usual Intel extension technologies except TXT (Trusted eXecution Technology). They're all said to have a power and thermal envelope of 130W."
Key points for me: "HyperTransport-like" -- this was the killer advantage AMD64 had a few years ago that made it such a hit in gaming circles while the Pentium 4 floundered badly. Also, 8MB L3 cache across four cores, 130W is not bad given the integrated northbridge.
So whats the point? a well specced machine that works HAS to be better than a slightly faster box .. that wont actually turn on!
It's the "thrill" that overclockers thrive on. For pro apps, Mac Pro usage, obviously this is all not recommended.
For enthusiasts, it's like driving your car through the city streets fast, seeing how fast you can go without crashing and killing someone or yourself.
Although, this is much safer, because the only thing you would destroy is your PC. Perhaps it keeps the kids of the streets too.
For gamers, it's the challenge of "can I play CounterStrike Source for 3 hours at a 25% overclock and +10fps without it crashing and making me look bad to my friends?".
More formally, it relates to overall system stability. This is where you do "burn-in" tests. So in PC land, you would run Prime95 and SuperPI, looped constant Futuremark 3DMark06 or 3DMark Vantage... Basically for a given overclock setting you would peg your CPU and GPU at 100% and see if it lasts 24 hours. If it does, then you overclock is pretty darn stable, and you've got a significant speed increase, greater value of your investment, large satisfaction, a good feeling, and without compromise to stability.
There is the issue of component degradation due to running over spec and at higher voltages. However, given a component is only useful for about 6-12 months (in enthusiasts' eyes) before it is obsolete, if you can push it for 6-12 months, and it craps out in the 2nd year, well, you've got your money's worth.
It is really a very different mindset. This is why I use a MacBook for all my professional work life, and personal life. My PC is for experimenting, gaming, watching videos and other non-essential stuff. If I had a Mac Pro right now, well, I'd be a different person, I think... I'm not in that situation right now that owning a Mac Pro is important. MacBook Pro 512mb vram if I get a certain Apple-related job, perhaps...
At the end of the day, it's all got to do with why people have a beat-up '70s Mustang or something like that in their garage, or a pimped-out Evo. I just have never been good with cars and stuff, and I epic failed the Mecanno set that my parents got for me when I was a kid, can't solder to save my life, and have no clue what you are supposed to do with Lego except for this kind of stuff: http://www.youtube.com/watch?v=sxLvnLI6GlU
Usually the GPUs don't matter so much since they have their default clocks set which is quite independent of the FSB. The PCI Express bus clock usually runs at 100mhz and is not affected by the Mac Pro overclock utility, as far as I can tell. As such, I think the issue you are facing is the mixed RAM environment. Usually, if you mix Apple and non-Apple RAM, or even different brands, this can affect RAM overclocks. You can try one set of sticks at a time and see if your overclocks go further.
Thanks for the input. When you say 'one set of sticks', are you saying I can take out just one 2 GIG stick? I thought they had to remain as pair.
Wait a minute ... Now I think about it the set up was odd to me when I installed, both Apple DIMMS are on one riser and the two 2 GIG 3rd party DIMMS are on the other not alternating as in all Macs I have owned before. So, I can try just one at a time thus having a 4 GIG Mac in both tests. Is this correct?
Even if I never run ZDNet Clock again after this, if it showed up my 3rd party RAM as so poor it was useful. Am I totally incorrect here or does this RAM, if it the cause, that cannot clock more than a 6% increase, have potentially other downsides for me running a high end system like this?
RAM set up: 2x 1 GIG Apple + 2 x 2 GIG 3rd party
Thanks for the input. When you say 'one set of sticks', are you saying I can take out just one 2 GIG stick? I thought they had to remain as pair.
Wait a minute ... Now I think about it the set up was odd to me when I installed, both Apple DIMMS are on one riser and the two 2 GIG 3rd party DIMMS are on the other not alternating as in all Macs I have owned before. So, I can try just one at a time thus having a 4 GIG Mac in both tests. Is this correct?
Even if I never run ZDNet Clock again after this, if it showed up my 3rd party RAM as so poor it was useful. Am I totally incorrect here or does this RAM, if it the cause, that cannot clock more than a 6% increase, have potentially other downsides for me running a high end system like this?
RAM set up: 2x 1 GIG Apple + 2 x 2 GIG 3rd party
Yes, so 1 set would be 1 pair. For example, take out the 2 3rd party sticks of RAM. Try the overclock. Put it back, take out the Apple sticks. Try to overclock.
Let us know how it goes.
I wouldn't say that this-or-that RAM is better for overclocking, because they've never been spec'ed or marketed as such. The other thing about overclocking, is... sometimes, it's all a matter of luck. The 3rd party RAM may well overclock better than the Apple RAM...
The RAM should be running well without any overclocking in your high-end system, because it should all be within spec. If your concerned about "bad" RAM, you would need to run your system at stock speeds with very heavy CPU & RAM usage for about 24 hours, if it is still stable, you're fine there.
Best thing is to just get 5-10 DVDs and HandBrake (handbrake.fr) it all... Start to really put your Mac Pro to full work. Handbrake is not a great example for such a high-end system, but hey, you'll have converted your favourite DVDs at the end of this "test"...!
Yes, so 1 set would be 1 pair. For example, take out the 2 3rd party sticks of RAM. Try the overclock. Put it back, take out the Apple sticks. Try to overclock.
Let us know how it goes.
I wouldn't say that this-or-that RAM is better for overclocking, because they've never been spec'ed or marketed as such. The other thing about overclocking, is... sometimes, it's all a matter of luck. The 3rd party RAM may well overclock better than the Apple RAM...
The RAM should be running well without any overclocking in your high-end system, because it should all be within spec. If your concerned about "bad" RAM, you would need to run your system at stock speeds with very heavy CPU & RAM usage for about 24 hours, if it is still stable, you're fine there.
Best thing is to just get 5-10 DVDs and HandBrake (handbrake.fr) it all... Start to really put your Mac Pro to full work. Handbrake is not a great example for such a high-end system, but hey, you'll have converted your favourite DVDs at the end of this "test"...!
OK will do later, thanks much for advice. I bet the 3rd party RAM is barely up to spec and is the problem.
OK will do later, thanks much for advice. I bet the 3rd party RAM is barely up to spec and is the problem.
Maybe it's a matter of viewpoint, but because it's not as overclockable doesn't mean it's not good. I've not seen any errors, correctable or otherwise, with my OWC memory at stock clock. I can't say anything about overclocking as I'm only using Tiger on a first gen Mac Pro, neither of which are compatible yet, if ever.
Try with just the stock Apple RAM that came with the machine and see if your overclock works better, to see if you can isolate the 3rd party RAM as being problematic.
So this doesn't really mean much for someone that already has 3.2GHz?
Users reporting running the 3.2 GHz at more than 3.7 GHz.
-Christoph
Waiting for new memory to arrive, but with 2x1gb Apple memory, I can run the zdnet clock utility (conservatively) - typically at 3.1 with no errors. I suppose I can try higher bus freq, but thought a few simple tests at this speed would be interesting.
However, running either cinebench 10, or Xbench 1.3, I see absolutely no performance increase? This compared with the tool set back to 400 mhz and the application off.
And yes, I am careful to hit 'apply' to the zdnet clock tool.
The system log does seem to suggest that the bus freq has been increased....
As the author wrote, its dynamically loaded, yes?, and not necessary to reboot.
I have a new Mac Pro 3.1 system, [email protected].
Waiting for new memory to arrive, but with 2x1gb Apple memory, I can run the zdnet clock utility (conservatively) - typically at 3.1 with no errors. I suppose I can try higher bus freq, but thought a few simple tests at this speed would be interesting.
However, running either cinebench 10, or Xbench 1.3, I see absolutely no performance increase? This compared with the tool set back to 400 mhz and the application off.
And yes, I am careful to hit 'apply' to the zdnet clock tool.
The system log does seem to suggest that the bus freq has been increased....
As the author wrote, its dynamically loaded, yes?, and not necessary to reboot.
As a side effect overclocking Mac OS X causes the real time clock to run faster. On a MacPro3,1 a warm reboot solves this problem.
Please read our FAQ, especially the following questions:
- Is there a chance that my computer will not boot after running ZDNet clock?
- Benchmarks, for example XBench, Cinebench and Geekbench, do not run faster after overclocking.
- Why does the system clock run faster? Is there anything I can do?
-ChristophUsers reporting running the 3.2 GHz at more than 3.7 GHz.
-Christoph
Nice to see you here on AI Forums to give us the "official" take on things.