I gather from reading multiple sources online that overclocking is relatively safe with modern hardware, if done properly (proper cooling, voltages in spec, good PSU, watching temps, going in small steps, stress-testing, etc.). If I were ever to overclock, is there some particular strategy that Dolphin responds to better?
For example, I know it's much more CPU intensive than GPU intensive, so overclocking the graphics processor would be a waste of time. But, would it perform better with a mild core clock bump and good memory overclocking or would Dolphin bench better with just pure clock increases? Are there any benchmarks that respond to changes similarly to Dolphin?
Quote:overclocking is relatively safe with modern hardware, if tested properly
Fixed that for you. This is the key. You can't f*ck anything up as long as you thoroughly verify the temperature and stability of the system under a heavy load......and stay away from high voltages.
Quote:If I were ever to overclock, is there some particular strategy that Dolphin responds to better?
No.
Quote:But, would it perform better with a mild core clock bump and good memory overclocking or would Dolphin bench better with just pure clock increases?
Your usage of the word clock in this sentence doesn't make sense. Both memory and the cpu have a clock rate. What exactly is a "pure clock increase"?
If you're asking whether to overclock just the cpu or cpu + memory I would go for just the cpu.
Quote:Are there any benchmarks that respond to changes similarly to Dolphin?
We have no way of knowing that.
(01-20-2012, 10:09 AM)NaturalViolence Wrote: [ -> ]Quote:overclocking is relatively safe with modern hardware, if tested properly
Fixed that for you. This is the key. You can't f*ck anything up as long as you thoroughly verify the temperature and stability of the system under a heavy load......and stay away from high voltages.
(01-20-2012, 02:18 AM)4evrplan Wrote: [ -> ]...if done properly (proper cooling, voltages in spec, good PSU, watching temps, going in small steps, stress-testing, etc.)
I sort of already said that. Besides, "done properly" implies proper testing and voltages. I'm not touching my BIOS without doing my research.
Quote:Your usage of the word clock in this sentence doesn't make sense. Both memory and the cpu have a clock rate. What exactly is a "pure clock increase"?
Obviously, I'd already the mentioned the memory clock as a counterpoint, so the context should make it clear I meant the CPU.
Quote:Quote:Are there any benchmarks that respond to changes similarly to Dolphin?
We have no way of knowing that.
Sure you do, if you've OCed and used Dolphin as a benchmarking tool along with other benchmarking tools as well, for comparison. I'm not suggesting you do it for me, but if someone on this forum already has, there's no point in me reinventing the wheel. And, if anyone has benchmarked with Dolphin and other tools at the same time, this forum is exactly where I'd expect to find them. A simple "I don't know" or just no comment at all is sufficient.
I don't mean to seem ungrateful NV, and truly, I am. After all, you're the only one who's tried to answer my questions in this thread so far. But with the tone of your replies, I felt the need to defend myself. Perhaps, I've misunderstood your intent, in which case, I am sorry. In any case, thank you for the information.
Do you have any other questions? Preferably more specific questions.
I'll let you know when I do.
If you mean like multiplier overclock vs. FSB overclock, it won't make a difference. Unless of course you let that FSB overclock increase the speeds of your memory and northbridge and stuff. In which case, the performance difference is minute if at all apparent in Dolphin. Multiplier overclock would be best (unless your processor doesn't have an unlocked multiplier). From personal experience, overclocking will help, but it won't make a huge difference. I have a Phenom quad at 3.9GHz and from 3.2 I've noticed maybe a few frames of improvement from stock speed. CPU architecture is probably the most important factor in running Dolphin effectively, and Sandy Bridge has it.
By the way, if you're already running a Sandy Bridge processor, with good cooling you could make a pretty big increase in performance going to like 4.5GHz. If not, I doubt overclocking will help you much.
Quote:If you mean like multiplier overclock vs. FSB overclock, it won't make a difference. Unless of course you let that FSB overclock increase the speeds of your memory and northbridge and stuff.
That depends on the architecture.
That is the sort of thing I mean. Reason I ask is I read somewhere a guy benchmarked the performance of his system (with architecture very similar to mine - AMD socket AM3) with different HyperTranport bus speeds from 2600 MHz up, and the higher speeds actually performed less well. But, I know different programs and even different benchmarks perform differently with different tweaks, hence the question. My processor is an Athlon II x3 450 (locked multiplier).
You still haven't asked a specific question though.
Core clock rate affects how fast the core can fetch/decode/execute instructions like arithmetic (integer, floating point, and simd), conditional evaluation, jump, and memory access (load/store/copy/move/etc.). CPU execution bound applications are the most likely to see a speedup from increasing this.
HT Link clock rate affects the data transfer speed between the cpu and NB - chipset. Increasing this will allow for faster data transfers between main memory and video memory. It MAY improve performance in some games when running a very high end SLI/crossfire setup.
Remember that as far as AMD is concerned the NB is broken into two parts. One exists in the chipset to facilitate bridged data transfer between the cpu, pci-e controller, and SB. The other part of the NB is on the cpu die, and includes the L3 cache, L3 cache access, and memory controllers.
The NB clock rate affects the speed of the on-die NB. Increasing it speeds up the L3 cache and memory controllers. Memory bound applications are the most likely to get a speedup from this.
In most applications the core clock rate should be the only clock rate out of those three that improves performance, so it is considered the most important, and therefore your top priority. The "FSB" on an amd platform (it really shouldn't be called that since that's not what it is but amd called it that so that users wouldn't have to learn a new word) is also called HTT, reference clock, base clock, and some other stuff if I recall. Increasing this increases all of the above (and more).....can't think of much else to say until you ask a more specific question.
damn...
my first gen i7 @3.8ghz
runned with 1.18v
... i down voltage him (stock is 1.2v)
maybe thats why he rebooted every 5-6h...
its over a year ago that i oc that thing -.-