Why does Vista cache so much ram?

masslac · 21 Mar 2007 at 11:16

garyh said:
For all those who have trouble understanding how Windows Vista's memeory management is of benefit;

I suggest you read this and this and then come back once you've finished trolling around spouting utter rubbish

The amount of self-confessed Vista gurus (who actually know very little about it other than the front end) on this forum makes me giggle.

Best post I've seen in some time.

afraser2k · 21 Mar 2007 at 11:33

masslac said:
Best post I've seen in some time.

I'm reading through them now, very informative.

NathanE · 21 Mar 2007 at 12:19

squiffy said:
So clearing RAM of crap that you're not even going to load doesn't take up CPU cycles? Come off it.

Come off what? Memory is not "cleared", it is simply overwritten. There is no additional cost whatsoever in doing this.

tripitaka · 21 Mar 2007 at 12:22

Also googling 'Superfetch' should provide some explanations in laymans terms.

Perhaps people would also like to be able to stop using the cache on their cpu, on the offchance that it may end up caching something that is not actually needed.
I'm sure that if there was a way to disable to prefetch, on the offchance that it might occasionally prefetch data you don't actually need you would see a massive performance increase

Or perhaps people should just familiarise themselves with the concept of a 'cache' a little more...

dirtydog · 21 Mar 2007 at 12:27

tripitaka said:
I'm sure that if there was a way to disable to prefetch, on the offchance that it might occasionally prefetch data you don't actually need you would see a massive performance increase

I know you are joking, but in fact you are right - Vista can prefetch programs or files which it shouldn't, which does result in a performance loss. I posted about it a while back.

Here it is: http://forums.overclockers.co.uk/showpost.php?p=8773680&postcount=33

NathanE · 21 Mar 2007 at 12:42

dirtydog said:
I know you are joking, but in fact you are right - Vista can prefetch programs or files which it shouldn't, which does result in a performance loss. I posted about it a while back.

Here it is: http://forums.overclockers.co.uk/showpost.php?p=8773680&postcount=33

Yes and I seem to recall rubbishing those claims that it hurts performance with the reason that the kernel supports I/O prioritisation.

tripitaka · 21 Mar 2007 at 12:50

I was talking about diabling CPU cache. You definitely wouldn't see performance gains there!

No doubt there may be the odd occasion where it causes problems, but 99% of the time you would see an increase or no noticeable decrease in performance.

And it can always be disabled.

dirtydog · 21 Mar 2007 at 12:56

NathanE said:
Yes and I seem to recall rubbishing those claims that it hurts performance with the reason that the kernel supports I/O prioritisation.

How can you rubbish it, unless you think the person who made the post was lying about his experience? You seem to think that this I/O prioritisation is a magic bullet which makes everything all right. It isn't. It doesn't work as wonderfully as you say it does.

dirtydog · 21 Mar 2007 at 12:57

tripitaka said:
I was talking about diabling CPU cache. You definitely wouldn't see performance gains there!

No doubt there may be the odd occasion where it causes problems, but 99% of the time you would see an increase or no noticeable decrease in performance.

And it can always be disabled.

99%, is that scientific?

NathanE · 21 Mar 2007 at 13:04

dirtydog said:
How can you rubbish it, unless you think the person who made the post was lying about his experience? You seem to think that this I/O prioritisation is a magic bullet which makes everything all right. It isn't. It doesn't work as wonderfully as you say it does.

Because it is a magic bullet?

Quite simply the prefetcher will not run when other processes are using the hard disk. It's as simple as that.

Who knows what setup that guy was running - for all we know it could have been some antiquated VIA chipset with the hard disks running in PIO mode... I certainly aren't saying he's a liar - just advising that one mans experience should not rule out the hundreds of thousands of people that now love Superfetch.

dirtydog · 21 Mar 2007 at 13:08

NathanE said:
Because it is a magic bullet? Quite simply the prefetcher will not run when other processes are using the hard disk. It's as simple as that.

Who knows what setup that guy was running - for all we know it could have been some antiquated VIA chipset with the hard disks running in PIO mode... I certainly aren't saying he's a liar - just advising that one mans experience should not rule out the hundreds of thousands of people that now love Superfetch.

He isn't the only person to have had such an experience. And if you read his post, I think we can infer that he knows what he's doing where computers are concerned, and hasn't overlooked such elementary issues as running in PIO mode.

NathanE · 21 Mar 2007 at 13:25

Bet'cha if he purchased a silent HD he would start singing the praises of Superfetch like everybody else. Since the dawn of personal computers we've had it drummed into us that "hard disk activity = slow down" but that just isn't the case in this scenario on an OS like Vista.

dirtydog · 21 Mar 2007 at 13:40

NathanE said:
Bet'cha if he purchased a silent HD he would start singing the praises of Superfetch like everybody else. Since the dawn of personal computers we've had it drummed into us that "hard disk activity = slow down" but that just isn't the case in this scenario on an OS like Vista.

He said his computer slowed down, not just that it was noisier. Other people have said the same thing. But in any event, excessive HDD activity does not do the HDD any favours in terms of its lifespan.

NathanE · 21 Mar 2007 at 13:46

dirtydog said:
He said his computer slowed down, not just that it was noisier. Other people have said the same thing. But in any event, excessive HDD activity does not do the HDD any favours in terms of its lifespan.

I know that

I was indicating there is a psychological link between HD noise and system slow downs that is mostly redundant on an OS like Vista with its Superfetch feature.

HDD lifespan... apparently this too is now firmly myth as well. Google did a massive research project not long ago and basically said hard drives will fail in the first few months due to manufacturing defects but once they get past this threshold should have a long life - with the only reductions coming from power cycling. They also said heat has little impact.

Burnsy2023 · 21 Mar 2007 at 13:50

NathanE said:
HDD lifespan... apparently this too is now firmly myth as well. Google did a massive research project not long ago and basically said hard drives will fail in the first few months due to manufacturing defects but once they get past this threshold should have a long life - with the only reductions coming from power cycling. They also said heat has little impact.

I also read this research, seems pretty conclusive.

Burnsy

ShakenNstirred · 21 Mar 2007 at 14:04

dirtydog said:
He said his computer slowed down, not just that it was noisier. Other people have said the same thing. But in any event, excessive HDD activity does not do the HDD any favours in terms of its lifespan.

I'm not going to be worrying about the hdd`s, in 12 years of PC`s and before that 2 years of Amiga 1200 with hard drive fitted, i haven't had a hdd fail.
so im not going to lose sleep on that point

dirtydog · 21 Mar 2007 at 14:08

I'm not an expert on HDDs but surely the fact that it is a mechanical device means that it will wear out and fail, it's just a question of when. And also surely the more it is used, the quicker that day will come. Don't they say that defragging puts a lot of strain on a drive so it shouldn't be done too frequently? If that is true then Vista's penchant for HDD thrashing will surely have a similar effect.

Oh and I don't see how heat can't be an issue either? That makes no sense. HDDs have a maximum safe temperature, above which damage can result. A thrashed HDD in hot weather is going to get hotter than one getting an easier ride under say XP. (And it will contribute to a higher case temperature also.)

squiffy · 21 Mar 2007 at 14:14

Constant (or more ) HD access isn't really a issue, you've probably got servers which have had constant accessing which are still working with no bad sectors. And even by the time that happens the HD is probably so useless it really doesn't matter.

Ie I have a 20GB Samsung which still works, but it wouldn't fuss me if it stops working.

NathanE · 21 Mar 2007 at 14:18

That's what most people thought but Google say otherwise. To be honest HD failure always has been quite a rarity. They are a sealed product afterall.

The introduction of fluid-based bearings several years back basically relieved any and all stress on the electric motor. The I/O arm actuators are also stupidly high quality components these days - mostly done for quietness but with reliability as a side affect.

I'd guess that the average P2P user gives their HD a thrashing probably several thousand times larger than Superfetch ever does. Superfetch afterall is read-only.

Burnsy2023 · 21 Mar 2007 at 14:19

dirtydog said:
Oh and I don't see how heat can't be an issue either? That makes no sense. HDDs have a maximum safe temperature, above which damage can result. A thrashed HDD in hot weather is going to get hotter than one getting an easier ride under say XP. (And it will contribute to a higher case temperature also.)

Its the fact that the drive heats up during use and cools down when off. This causes contraction and expansion of the metals and other components in the drive which causes stress and will eventually cause faliure.

Burnsy