Originally posted by: PhyuckHughe
Originally posted by: Fresh Daemon
It isn't solid at all. If you think about the theory behind it, you will see this.
DDR-SDRAM has a typical access time of 2.5 nanoseconds. A hard drive has a typical access time of 9 milliseconds. That's 0.0000000025 seconds for RAM and 0.009 seconds for the hard drive. The hard drive is 3.6 million times slower than the RAM!
So how is a very small increase in hard drive performance going to have any noticeable impact on this incredible performance gulf? Moreover, if this huge performance gulf results in a quite small change in real-world performance, how are you going to notice a tiny change in it?
Wow, that is so freaking retarded.
First of all, you ignored how he mentioned your testing failed. You didn't thoroughly test it in the right way. It's easy to see this.
And since you say that the hard drive is
millions of times slower, wouldn't that be considered a bottleneck? Well, duh, doofus, making the part faster where the bottleneck is, is usually the best thing to do for performance. If you have to wait for the hard drive, because it's so much slower, than any increase in HD performance would just lessen the time you would have to wait for the hard drive.
Why are you still trying to act as if you know what you're talking about? The only thing I can think of for why you would have such a huge mental fallacy is if you are saying that this wouldn't make the functions of the program work faster, but that isn't even the point that's trying to be made. This is exactly why your tests fail. We're talking REAL WORLD performance. Switching between programs, or sub-parts to programs, is also part of that.
Yes, back in Windows 3.1 days when 1MB of RAM was common and virtual memory was a primitive science!
If you reach this point where overuse of the pagefile is causing disk grinding and stuttering performance, let me say this very clearly: No amount of pagefile tweaking is going to alleviate this situation. Performance will be terrible. Maybe it can be very slightly less terrible, but very slightly less than terrible is still pretty terrible!
Jesus, even more blatant stupidity. Do you even understand how hard drives or paging files work? It sure doesn't look like it.
Like mentioned already by several people (even including "BV"), having the paging file on the sytsem disk (or partition, even) is the worst option. Considering how it's the disk that the most used, it's clearly self evident that having the pagefile there would be the worst place. And also since it's the one that most used, the pagefile would also be more likely to become fragmented.
Do you know anything about Linux? Well from what I've seen, the default option is always to put the swapfile on its own partition. Separating it like this, even though it may still be on the same disk, already is helping performance. Windows doesn't do this by default because it's not the simplest solution (from the end user's perspective).
The only thing that helps is more RAM. Either run more lightweight programs, alter your computing habits, or buy some more.
Haha, again "just upgrade to get better performance, instead of getting the most out of what you have." Don't you see how you're sounding? The only solution you are giving is to spend more and upgrade. That's absolutely the worst kind of advice, considering what the topic is about. Upgrading isn't "performance tweaking".
And you say that there's basically nothing to do to help this, but if you have a 3 drive RAID 0 array with the pagefile, it's obvious that the hit of using so-called "virtual memory" would be greatly reduced.
Tweaking the pagefile to "help" this situation is like getting shot in the chest instead of the head. You may find some improvement, but it will be overshadowed by the catastrophic results of the situation as a whole anyway.
That just makes no sense. Since you're such an advocate of using a paging file, it naturally would make sense to adjust its setting for it to get better performance. If you're acting like it's
so much damn slower, then wouldn't the best thing be
disabling the paging file? Then that would eliminate what you say is several "million times slower".
That's assuming there's some improvement to be had. I've yet to see a benchmark where someone has proven that pagefile tweaking does actually produce a measurable performance increase.
Again, this falls back on your bad testing. The tests should be done in an environment where the tested thing is being used, not some arbitrary time of your choosing. Testing Quake 3 or some other nonsense surely isn't going to test the paging file and how it's related to real world performance.