SSD drives in education

[CyberNerd]

I.T. is ALL about shifting bottlenecks around. It always has been.

The majority of bottlenecks are created by badly written or bloated software.

For everything else there is a hardware upgrade.

[SYNACK]

What - you mean a physical machine like a Hard Disc?? So replacing a physical machine like an HD with a solid state ..

 

oh ok - I give up.

 

A physical machine like a drill and three stepper motors that's job is to move slowly through a simple (in digital terms) queue of movements. The CPU could probably simulate the outputs required almost instantly but physics and energyconsumption prevents you from moving objects at near light speed while machining them.

 

It depends on how the hardware is designed (onboard processor and memory to look at sensors and work through the job queue) as to how CPU heavy it is. For all we know it could be offloading most of the 'realtime' events to its onboard electronincs - which is likely.

 

A hard drive works at the microscopic level for many tasks and have buffers etc. it is also specificly designed to minimise friction. A mill uses all sorts of friction and must work slowly - comparitivly - so as not to simply melt or destroy the material being machined.

[pcstru]

The challenge for CNC is generally the real time demands placed on the driving machine. It may be slow comparatively to a CPU, but the sequence of operations have to be performed to reasonably precise timings. It's no use if the PC is busy dealing with your browsing request or rendering your CAD drawing, while the last machining sequence on a part which you have already spent 10 hours machining, blackens and burns because the bit is not moving through the material at the required speed. To be clear, as an application it is not necessarily demanding of a processor, but it is hugely intolerant of machines where the processor is a major bottleneck.

 

The point is the PC is adequate for the tasks and as an upgrade, an SSD would make a noticeable difference to the user in ways that are important (boot time, application load, file load) whereas the "weedy" processor bottlenecks seem to remain reasonably well hidden.

Edited May 10, 2012 by pcstru

[gshaw]

The problem I'm finding with SSD drives is PC manufacturers (or their suppliers)

 

Crucial M4 64GB... about £50 on any website e.g. Scan / eBuyer

Crucial M4 64GB instead of HDD in a new PC... £75

 

That's from a well known UK custom build manufacturer, apparently due to supply chain costs.

 

Put that extra over an average order of 50 machines that's more than £1000 extra for no good reason. Can't buy the units with no HDD as installing our own invalidates the warranty... at this point trying to make a case for the upgrade becomes a lot more difficult.

 

As for Dell / HP / Lenovo I dread to wonder the premium they'll add on. Question is why?

 

On a technical level unless you need the local storage space SSDs are the clear choice for desktops but something is keeping HDDs in until the bitter end.

Edited May 10, 2012 by gshaw

[SYNACK]

The challenge for CNC is generally the real time demands placed on the driving machine. It may be slow comparatively to a CPU, but the sequence of operations have to be performed to reasonably precise timings. It's no use if the PC is busy dealing with your browsing request or rendering your CAD drawing, while the last machining sequence on a part which you have already spent 10 hours machining, blackens and burns because the bit is not moving through the material at the required speed. To be clear, as an application it is not necessarily demanding of a processor, but it is hugely intolerant of machines where the processor is a major bottleneck.

 

The point is the PC is adequate for the tasks and as an upgrade, an SSD would make a noticeable difference to the user in ways that are important (boot time, application load, file load) whereas the "weedy" processor bottlenecks seem to remain reasonably well hidden.

 

As above, how much is being offloaded to onboard hardware on the CNC, with the buffer that provides it should never be slowing down enough to be noticable if the PC is up to it. Besides, you are kind of proving my point that even with only two apps open a browser and dodgy web stuff can bind the CPU. Try that with a dozen tabs and word, excel and SIMS open you will notice.

[gshaw]

As above, how much is being offloaded to onboard hardware on the CNC, with the buffer that provides it should never be slowing down enough to be noticable if the PC is up to it. Besides, you are kind of proving my point that even with only two apps open a browser and dodgy web stuff can bind the CPU. Try that with a dozen tabs and word, excel and SIMS open you will notice.

 

I used to run Denford CNC machines on a 233MHz Pentium with 64MB RAM so the software must've gone horribly wrong somewhere in the last few years for it not to work on a newish machine with Dual Core CPU

[pcstru]

As above, how much is being offloaded to onboard hardware on the CNC, with the buffer that provides it should never be slowing down enough to be noticable if the PC is up to it.

You are making an assumption that there is a buffer. On my old machine there was absolutely no buffer, it was a straight feed from the parallell port to a xylotec stepper driver. On the current machine I use a USB to Parallel device, which has a small buffer (simply because USB is serial, it couldn't do the job without at least an 8 bit buffer) but it is not actually doing anything other than that conversion. The main signal controller is the PC.

Besides, you are kind of proving my point that even with only two apps open a browser and dodgy web stuff can bind the CPU. Try that with a dozen tabs and word, excel and SIMS open you will notice.

 

More likely you will first notice a slow down as the system runs out of memory and pages to disc - another aspect that would only benefit from running with an SSD. (now now, fairs fair, if you can make a load of assumptions, so can I).

[SYNACK]

Serial or parallel, it is such a slow interface that the cpu would have to be completly busy for some time to cause a delay of any kind as there would be amount of time for the CPU to deal with stuff inbetween each character transmition given the realitive speeds. This is kind of talked out though, whatever you say I am still going to consider a Pentium D an old slow cpu and even if the speed is not an issue the age of the hardware makes it less reliable thanks to wear and tear. What you decide is of the best value to you or your school is of no concern to me, I don't care. My maths for my own schools though comes out with Pentium Ds not being worth that kind of investement when compared to a new machine.

[pcstru]

You seem to be very confused. I'm not investing in them. I was just responding to a question with an implied statement - that a Pentium D would see no benefit from an SSD because the processor is the bottleneck. HD's are so slow compared to any processor in the last 15 years that that statement does not stand up to any serious scrutiny. You really wanted a different question.

[Flatpackhamster]

You seem to be very confused. I'm not investing in them. I was just responding to a question with an implied statement - that a Pentium D would see no benefit from an SSD because the processor is the bottleneck. HD's are so slow compared to any processor in the last 15 years that that statement does not stand up to any serious scrutiny. You really wanted a different question.

 

You also seem to be confused so allow me to remind you that I never said that a Pentium D would see no benefit from an SSD. I stated that the fundamental problem with those machines was the processor, not the hard disk, and it is.

[pcstru]

You also seem to be confused so allow me to remind you that I never said that a Pentium D would see no benefit from an SSD.

 

Really?

 

"If you're running Pentium Ds then the bottleneck will be the processor, won't it? Changing the hard drive to an SSD will not solve the fundamental problem with those machines which is that their rather weedy processor can't cut the mustard.

 

I can definitely see the benefit in this in some machines, I just can't see it with a Pentium D."

Edited May 10, 2012 by pcstru
Edited for clarity

[synaesthesia]

With the current pricing of 120gb drives, I am now absolutely certain you'd be fairly daft to buy anything lesser, capacity-wise unless you're running a fat VDI client then it'd just have no benefit.

[pcstru]

Serial or parallel, it is such a slow interface that the cpu would have to be completly busy for some time

Whoa there - is 'some time' like an SI unit? USB is running at 60MBs, 480Mbs. Parallel could be 32Mbs and the machine running 1/16 1.8 deg step at 130mm/s on a 1mm turn ball screw on three axis is hungry for steps one and a quarter million times every second. If it doesn't get that data in reasonable time, it will start to, at best, skip steps (for 1/100th mm accuracy - a not unreasonable request on some jobs, you can miss 32 steps over the life of the reference). At worst it will crash probably wrecking the bit and if your unlucky, the job which wasn't even looking for a part to 1mm let alone .02mm .

 

Now, I'm not saying it is anything special for a computer to deal with, but you can look at it a reasonably demanding *user*. Hey, I'm happy to wait a second for stuff and then I get grumpy, but I don't just trash expensive tooling if the PC doesn't serve me up the goods a couple of million times per second. Or trash the job you have just worked on for hours, days or perhaps weeks. If the processor is so prone to being the bottleneck, then I'd expect it to show up in that application in a way that really punishes ignorance paraded beside the rhetoric of "I'm entitled to my 'opinions'/we must agree to differ".

 

to cause a delay of any kind as there would be amount of time for the CPU to deal with stuff inbetween each character transmition given the realitive speeds.

As long as you don't actually want to do any numbers, think much, or know much about the actual issues of working with CNC machines, that might just pass as a 'reasonable opinion'.

This is kind of talked out though,

What? We haven't even started! I've bought numbers and everything! Damn - and no refund on the rental of those, I got all one to ten of them for the whole weekend too. Damn you sir!

whatever you say I am still going to consider a Pentium D an old slow cpu and even if the speed is not an issue the age of the hardware makes it less reliable thanks to wear and tear.

Wear and tear! on ... err ... a Pentium D! Essentially the very essence of an entirely solid state device!! What? Do the electrons wear out the walls of the wires? Do they gouge channels out of solid copper as they rush up the conductors and then spookily, tunnel through solid crystalline rock in a peculiarly uncertain quantum way?

What you decide is of the best value to you or your school is of no concern to me, I don't care. My maths for my own schools though comes out with Pentium Ds not being worth that kind of investement when compared to a new machine.

 

That seems to be a maths without numbers. Very abstract. Nice. I so wish my boss was ignorant enough that I could get away with that kind of shit. "Nullus in verbia" or something, seems to be his catchphrase.

 

Personally, even the passing implication that any 'average' user of a Pentium D spec system is going to see the processor [rather than/before they/or really in any 'general' sense, experience] as a bottleneck *BEFORE* they notice the limitations of a 7ms (merely 7 thousands of times a second) seek time HD or maybe even a sustained ***100 IOPS*** disc substem (merely a hundred or so times a second) ... well that's just totally ******g nuts. It's sort of on the outer boundary of possible but your users must be doing some crazy shit to actually be data points in that particular part of the ... curve.

 

But OK, do actually pity my users. I'll put it as I see it here.

 

To me the maths says, 1x classroom, 30x D spec pentium - students not running SIMS and generally being asked by the teacher to do one thing at once. PC quite capable of that. Biggest single improvement to the user experience - faster login, quicker load of apps on image, any pageing issues (random IO) improved by two or three orders of magnitude. Longevity of *investment*, well, hey, I can probably use the SSD's in the PC's I eventually buy to replace the D's. Meanwhile I can get another year out of the D's AND offer an improved user experience and spend the difference on something else that really ... really matters ...

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

... like my bonus....

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

... what

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

... you don't get one of those?

 

 

 

 

 

 

 

 

 

 

 

 

... you surprise me.

 

 

 

 

 

 

 

(now - that last bit - that really was trolling)

 

[Observant readers will notice I answered the same post twice, which means I is actually *been* totally trolled - so I is bin de dubble dumpin dummbed - or sumin. Ho Humm]

[SYNACK]

You're taking very great pains to argue something that was not a clearly defined statement rather an 'implication' that you have found yourself. At any rate; the wear and tear I am talking about it more to do with motherboards becoming unreliable thanks to the long term effects of dust, thermal expantion and contraction, crystalising solder, leaking caps, power surge related damage, vibration damage, pitting of heat exchangers, wearing out of cooling fans etc. Damage to these can also easily cause damage to the CPU, especially random power spikes over a long period of time.

 

Do you service them regularly as in clean, replace fans where needed etc. even that only gets them to last so long as they are not designed to last anymore.

 

As to you're CNC machine, my argument was that we did not know exactly how it was configured from a hardware point of view or how the software and drivers actually dealt with it, we still don't so we only have your idea of the kind of throughput that it actually requires, as someone else was running one on a 233MHz machine you have to admit there is some varience.

 

Now that you are actually providing numbers of your own, 1 year, 30 pcs and proposing reuse (in new self built machines?) it sounds more feasible but again, for us we prefer to get rid of the 6-7 year old PCs before they start causing massive issues.

 

Anyhow, as I have said before this is kind of talked out as neither of us is likely to change our minds on the issue so give the under bridge dwelling a rest.

[pcstru]

You're taking very great pains to argue something that was not a clearly defined statement rather an 'implication' that you have found yourself.

I think I have been clear and consistent, both about exactly what I am replying to and how I have interpreted it.

At any rate; the wear and tear I am talking about it more to do with motherboards becoming unreliable thanks to the long term effects of dust, thermal expantion and contraction, crystalising solder, leaking caps, power surge related damage, vibration damage, pitting of heat exchangers, wearing out of cooling fans etc. Damage to these can also easily cause damage to the CPU, especially random power spikes over a long period of time.

 

Do you service them regularly as in clean, replace fans where needed etc. even that only gets them to last so long as they are not designed to last anymore.

'eh? It's digital - it is dead or alive. As as an operational demand, 5 year old machines don't make demands, beyond new machines. Statistically I'd say new machines are more likely to be ... time consuming. The risk tends to be skewed towards problems with large numbers of new machines as a function of age and you are inclined to spend time sorting them out rather than just make a quick 'scrap unit' decision.

 

As to you're CNC machine, my argument was that we did not know exactly how it was configured from a hardware point of view or how the software and drivers actually dealt with it, we still don't so we only have your idea of the kind of throughput that it actually requires, as someone else was running one on a 233MHz machine you have to admit there is some varience.

If they used one in anger rather than than merely passing one by in the corridor, they might appreciate the actual machinery and the issues particularly the working limitations. But perhaps they can chip in with tooling, materials, bits, spindle, feed speeds and attained accuracy? Not that it matters much - solid state drives yield several orders of magnitude of possible performance increase over processor and those improvements are smack bang in the zone where users actually likely to notice them.

 

Now that you are actually providing numbers of your own,

You cheeky git! My first post in thread had numbers. Processor speeds, memory speed, bus speed. Examples of you supplying hard data are even now ... non existent. FYLS!

 

1 year, 30 pcs and proposing reuse (in new self built machines?) it sounds more feasible but again, for us we prefer to get rid of the 6-7 year old PCs before they start causing massive issues.

 

Anyhow, as I have said before this is kind of talked out as neither of us is likely to change our minds on the issue so give the under bridge dwelling a rest.

 

I'm not likely to change my mind because frankly I'm right and have bought the evidence to back that up. Why you won't change your mind is beyond me - but since you admit even good evidence won't shift you, I guess you have to live with that and bleat out the "I'm entitled to my opinion" fallacy.

 

As to who is the real troll ...? It must be you... surely??!!

 

:-Q.

[sonofsanta]

With the current pricing of 120gb drives, I am now absolutely certain you'd be fairly daft to buy anything lesser, capacity-wise unless you're running a fat VDI client then it'd just have no benefit.

 

I don't know - XP certainly could fit Windows, Office, CS5 and others into 60gb, Windows 7, well... I'd be more comfortable with 80, but it should fit. After all, there's no reason not to fill an SSD - they don't slow down as they near capacity, unlike mechanical.

 

a 120gb drive will perform faster than a 60gb anyway, of course, but not to the same difference as SSD v HDD in the first place.

[synaesthesia]

I entirely discount XP on SSD's on purpose - not for TRIM or anything of the sort as it won't be a problem in the machines lifetime however with XP's grave being dug, it's not worth factoring it in.

 

From experience, 7 on a 60GB drive is a squeeze unless you're really careful with profile sizes and what-not. At home even with all apps installed on a HDD, 60GB gets filled very quickly indeed thanks to much of the cack Windows soon builds up which ended up in me making a lot of symlinks for folders like winSXS, Installer folders etc. In a network environment you'd need to narrow down a lot via GPO to stop or limit caching of profiles, windows updates and what-not. It can certainly be done but the price difference is less hassle than the hassle

[zag]

With the current pricing of 120gb drives, I am now absolutely certain you'd be fairly daft to buy anything lesser, capacity-wise unless you're running a fat VDI client then it'd just have no benefit.

 

We've bought 40gb drives for all our workstations and its been fine.

 

Our main windows 7 image with all software is 12gb and we don't store anything locally so whats the point?

 

Laptops we are putting 80gb in now and have never had any complaints.

 

And to those who continue to insist that CPU's are the bottleneck, go have a try of the new Raspberry PI to see what that means. Its CPU is the bottleneck but thats not suprising as its 300mhz!!

Edited May 11, 2012 by zag

[zag]

a 120gb drive will perform faster than a 60gb anyway, of course, but not to the same difference as SSD v HDD in the first place.

 

Not really. The random access time is exactly the same on all SSD's. That's the thing that matters in a real world environment.

 

Those benchmarks you see on review sites are for sustained transfer speeds. Unless you are copying DVDs from one machine over USB 3 to another with equivalent SSD drive its a pointless statistic.

[sonofsanta]

Not really. The random access time is exactly the same on all SSD's. That's the thing that matters in a real world environment.

 

Those benchmarks you see on review sites are for sustained transfer speeds. Unless you are copying DVDs from one machine over USB 3 to another with equivalent SSD drive its a pointless statistic.

 

I thought the random would be higher as well? As there are more chips used so when pulling in all the various DLLs etc. to load Photoshop, it can grab them from 8 baskets instead of 4, like some kind of mad silicon octopus.

 

I wouldn't expect most people to particularly notice the difference between capacities of SSDs, though. I'm more just concerned about fitting everything on (with a little bit of space left over for some additional over-provisioning, ideally)

[synaesthesia]

No idea on speed side of things - I do have to ask @zag though if the size of those drives remains useable even with windows updates etc? Update cache on a 6 month old PC (looking at it currently) is nigh on 10gb on it's own!

[sonofsanta]

No idea on speed side of things - I do have to ask @zag though if the size of those drives remains useable even with windows updates etc? Update cache on a 6 month old PC (looking at it currently) is nigh on 10gb on it's own!

 

Is there any way of clearing out that update cache? Always frustrated me how much it bloats, as it's never clear exactly which files you can rid yourself of and cleanmgr doesn't touch it.

[Arthur]

Is there any way of clearing out that update cache?

Not sure about the update cache, but there is a workaround for the Windows Installer cache.

 

What Changes in Windows Installer 5 (Windows 7 and later)

Windows Installer version 5 (Windows 7 release) caches the entire .MSI package to its cache location - including any internal application source files. So if you have 1.5 GB .MSI, you will find the entire 1.5 GB file in the cache with a random name. So you may start finding some very large .MSI files in c:\Windows\Installer on Windows 7.

 

Why does it do this? It turns out that a 1.5 GB package that is signed will give a nice blue UAC prompt indicating that everything is cool - the package is signed! If, however, the files are stripped out of it during caching, the signature is broken and an uninstall from Programs and Features gives the nasty alarming Yellow UAC prompt indicating that software from an unknown source wants to elevate on your system.

 

This is, in fact, exactly how it works on Vista. Microsoft struggled with how to fix this situation all during Vista's lifetime. The solution chosen was caching of the entire package.

 

There is no way to disable this behavior and simply take the yellow UAC prompt hit. Even if you are a corporation who never let's users uninstall software from Programs and Features!

 

This behavior also happens even if the package is not signed.

 

How Do I Prevent Full Caching?

If your company made the packages, then change your package format to have external .CAB files. The external cabs are not copied into the cache. (Source)

[gshaw]

If they used one in anger rather than than merely passing one by in the corridor, they might appreciate the actual machinery and the issues particularly the working limitations. But perhaps they can chip in with tooling, materials, bits, spindle, feed speeds and attained accuracy? Not that it matters much - solid state drives yield several orders of magnitude of possible performance increase over processor and those improvements are smack bang in the zone where users actually likely to notice them.

 

Yup used to do all of that, setting the start of material (zero axis iirc?) where you had to have the drill bit *just* touching the surface of the material in 0.1mm steps, choosing correct speeds so the material \ drill bit didn't shatter and so on. Had hours of fun milling aluminium for injection moulding n 6th form where I had to keep squirting the drill bit with white spirit to stop it melting into a gloopy mess during the mill

 

Thinking about it the machines weren't even 233Mhz, they were old Apricot PCs either 120Mhz or 133Mhz but never had a problem with the milling itself. Granted I wouldn't run any other program while it was going but I'd still do the same today even if an i7 was running it as it was more of a distrust of Windows \ other app not crashing than the speed of the box doing the milling.

[zag]

No idea on speed side of things - I do have to ask @zag though if the size of those drives remains useable even with windows updates etc? Update cache on a 6 month old PC (looking at it currently) is nigh on 10gb on it's own!

 

Yes, We've used 40gb SSD's site wide for 2 years now. Never run out of hard disk space on any of them. Infact we've never been above 20gb that I know of on our Windows 7 machines fully loaded with software.

 

I never quite got why people buy computers in a school with 300gb drives in, doesnt make sense to me for client machines.

 

EDIT: Obviously you need to be sensible and remove old profiles ect. once a year or something.

Edited May 11, 2012 by zag