Ad
Ad
Ad
Pages: [1] 2 »   Bottom of Page
Print
Author Topic: Help with memory upgrade and Bios settings  (Read 6745 times)
walter.sk
Sr. Member
****
Offline Offline

Posts: 1328


« on: January 19, 2011, 11:11:01 AM »
ReplyReply

I recently received a Dell/Alienware Area 51 computer, and am looking to replace the memory with the maximum possible.  Here is what I have:

Motherboard: X58
CPU: i7 930 at 2.8 Mhz
Video: nVidia GTX 460 with 1GB memory
RAM: 6GB DDR3 triple-channel memory, 1333 Mhz, non-ECC, unbuffered, unregistered.
There are 3 memory slots total, and the specs in the manual say either DDR3 or DDR3-XMP will work.  
(On the Dell website, this computer has an option of 12GB at either 1333Mhz or 1600Mhz.)

I looked at the Kingston website and found only one 12Gb kit at 1600Mhz in their HyperX Genesis line:
KHX1600C9D3K3/12GX.  Further specs are CAS9, 1.65 volts, Timing 9/9/9/27.
At one time I saw this particular kit available on the Dell website but no longer.  I was assured by a Dell technician that the kit would be compatible.

While Dell wants $1300 for an upgrade to this amount of memory at 1600Mhz, the Kingston kit can be had for less than $150 from a very reputable dealer, sent directly from Kingston  The difference between this and the price of the 1333Mhz kit is so small I'd rather get the faster kit, which also comes with better heat sinks.

My question is mainly this:  If I replace my current memory with this kit, will the BIOS automatically detect the settings needed, or would I have to set the values manually?

And if I have to supply the manual values, what settings do I use?

I know enough to enter the settings, but do not really understand what they are.  I also do not want to overclock the memory.
« Last Edit: January 19, 2011, 11:14:19 AM by walter.sk » Logged
Gemmtech
Sr. Member
****
Offline Offline

Posts: 526


« Reply #1 on: January 19, 2011, 12:00:40 PM »
ReplyReply

I always check my BIOS for the correct settings regardless.  Install the memory and the correct settings should be in the BIOS, however once you enter the BIOS just check to make sure it is per the manufacturers specifications.  You can usually overclock memory without many problems, but the performance difference IMHO isn't enough to do it, I keep all my built machines as stock.  The procedure is pretty straight forward.

Logged
walter.sk
Sr. Member
****
Offline Offline

Posts: 1328


« Reply #2 on: January 19, 2011, 01:43:09 PM »
ReplyReply

I always check my BIOS for the correct settings regardless.  Install the memory and the correct settings should be in the BIOS, however once you enter the BIOS just check to make sure it is per the manufacturers specifications.  You can usually overclock memory without many problems, but the performance difference IMHO isn't enough to do it, I keep all my built machines as stock.  The procedure is pretty straight forward.

Thanks. I've ordered the memory, so I've got my fingers crossed.
Logged
Steve Weldon
Sr. Member
****
Offline Offline

Posts: 1447



WWW
« Reply #3 on: January 19, 2011, 02:56:38 PM »
ReplyReply

Wow.. I never realized anyone made a x58 chipset motherboard with only 3 RAM slots.  One of the advantages of the x58/1366 socket MB's is the 6 slots for RAM.  Learn something new every day..

About memory settings.

It's true that if you install compatible memory the motherboard will automatically sense it and use the default settings.  You won't need to do anything else.

However, it's a misnomer to think faster memory is "overclocking" or hard to use.  While it's true that simply increasing your Bclk setting categorically increases all clock rates (including RAM) across the motherboard, its still possible to use faster memory providing your motherboard supports it.

Faster memory, such as Kingston's Hyper-X and others.. will have built in profiles a compatible motherboard will pick up.  Lets say its 2333 speed memory.  A compatible motherboard installs this as 1066 at default.  You can use it this way.   You can also go to your motherboard BIOS settings under "XMP MEMORY PROFILES" and choose a. Default  b.  Profile 1  c.  Profile 2.

If you choose Profile 1, the embedded profile of the compatible RAM will be read and the motherboard will then know, and set the RAM's settings to their maximum 2333 speed WITHOUT overclocking or doing anything else with the motherboard.  Profile 2 is usually some speed between default and the max..

Memory by itself will only make a 5-6% total system performance gain at best.. but it's something to consider when ordering your RAM provided you have a compatible motherboard.
Logged

----------------------------------------------
http://www.BangkokImages.com
Gemmtech
Sr. Member
****
Offline Offline

Posts: 526


« Reply #4 on: January 19, 2011, 03:48:08 PM »
ReplyReply

"However, it's a misnomer to think faster memory is "overclocking" or hard to use."

If it's out of spec it is overclocking.  And generally speaking it just depends on how far you want to push the settings.  I just have found so little gain from overclocking the CPU and memory unless I took the time to take care of the other parts of the system and especially the cooling of the CPU/System

Logged
Steve Weldon
Sr. Member
****
Offline Offline

Posts: 1447



WWW
« Reply #5 on: January 19, 2011, 09:41:06 PM »
ReplyReply

"However, it's a misnomer to think faster memory is "overclocking" or hard to use."

If it's out of spec it is overclocking.  And generally speaking it just depends on how far you want to push the settings.  I just have found so little gain from overclocking the CPU and memory unless I took the time to take care of the other parts of the system and especially the cooling of the CPU/System


That would be true, but the specifications when buying memory run from 1066 to over 2333.  The motherboard's default of 1066 is not the standard to be guided by.  If the motherboard supports faster spec'd RAM then it's not over clocking if you use faster spec'd RAM.  It's simply running it at it's designed speed.

And there is a huge gain, up to 30+%, to be gained with over clocking the CPU.  Many 'factory' computers come over clocked.  Intel's own 'turbo boost' is a form of over clocking.  Cooling the system isn't that difficult or expensive.  In most cases a $40 CPU cooler will be all one needs to reap significant increases.

The thing is, for most people the CPU/RAM is not their bottleneck.  So over clocking 'appears' to do less than it really does.  The bottleneck for most people is still their storage devices.  Slow hard drives.  Or even slow SSD's.  Fortunately we're moving in the right direction and faster SSD's like this one I recently installed in a new system have really impressive performance and shift the bottle neck away from the drives to other parts of the system such as a slow video card or CPU.

In an ideal system all components will perform to the same capacity.  Unfortunately technology hasn't yet taken us there.
Logged

----------------------------------------------
http://www.BangkokImages.com
Gemmtech
Sr. Member
****
Offline Offline

Posts: 526


« Reply #6 on: January 19, 2011, 09:58:59 PM »
ReplyReply

"That would be true, but the specifications when buying memory run from 1066 to over 2333.  The motherboard's default of 1066 is not the standard to be guided by.  If the motherboard supports faster spec'd RAM then it's not over clocking if you use faster spec'd RAM.  It's simply running it at it's designed speed."

I didn't say it was the motherboard's spec, it IS the memory spec that matters if the motherboard supports the speed there's not an issue and I agree it's NOT overclocking.

"And there is a huge gain, up to 30+%, to be gained with over clocking the CPU."  

The problem is this isn't applicable to what one sees in the "real" world, IOW, does everything else fall in line, software, as you state later the HD is a bottle neck as well as video card and does your software take advantage of the overclocking?  If you are a big time gamer it might, other factors involved as well.  
The best approach if you want to overclock, do it in steps and test the system i.e. hardware/software to see if there are improvements.

Though I'm using SSDs I'm not quite sold on them yet, I want to see what happens in the long term.

"In an ideal system all components will perform to the same capacity.  Unfortunately technology hasn't yet taken us there."

 and the software will take advantage of it.  Word doesn't take advantage of a 10 year old box.

"Fortunately we're moving in the right direction and faster SSD's like this one I recently installed in a new system have really impressive performance and shift the bottle neck away from the drives to other parts of the system such as a slow video card or CPU."

That would have to be one slooooooooooooow CPU
« Last Edit: January 19, 2011, 10:01:57 PM by Gemmtech » Logged
Steve Weldon
Sr. Member
****
Offline Offline

Posts: 1447



WWW
« Reply #7 on: January 19, 2011, 10:22:44 PM »
ReplyReply

"I didn't say it was the motherboard's spec, it IS the memory spec that matters if the motherboard supports the speed there's not an issue and I agree it's NOT overclocking."

Then we are in agreement.  I suppose I didn't understand your need to respond to this the first time.

"The problem is this isn't applicable to what one sees in the "real" world, IOW, does everything else fall in line, software, as you state later the HD is a bottle neck as well as video card and does your software take advantage of the overclocking?  If you are a big time gamer it might, other factors involved as well. "

Actually it is.  It depends on the task at hand.  You can make the exact same argument for a 10 year old CPU/RAM combination and it would be true with some tasks.. but not all.  And of course software takes advantage of over clocking.  Software instructions are executed at the speed of the system, over clocked or not.  Whether or not this makes a difference on a specific task depends on the task and how it loads the different systems of your workstation.

It would be better to just agree a faster PC is better than a slower PC.  We all know this.  If over clocking can be achieved with 100% reliability and the system components still have a life-span of more than you'll keep the machine, then there is no argument against over clocking.  And it can, easily.
 
"Though I'm using SSDs I'm not quite sold on them yet, I want to see what happens in the long term. "

Really?  What aren't you sold on?

"That would have to be one slooooooooooooow CPU"

Not at all.  It would merely be a system where my tasks tax the SSD less than the CPU.  I'm sure you understand that some tasks like noise reduction for instance.. are highly CPU intensive?  RAM is important here as well.  CPU/RAM is being taxed for far more of the required time to execute the command than is the SSD..  But if it was a mechanical HDD this might not be the case.  It depends on the tasks.  Not everyone is running a 10 year old Word program..

[/quote]
Logged

----------------------------------------------
http://www.BangkokImages.com
Gemmtech
Sr. Member
****
Offline Offline

Posts: 526


« Reply #8 on: January 20, 2011, 03:43:09 AM »
ReplyReply

"It would be better to just agree a faster PC is better than a slower PC.  We all know this.  If over clocking can be achieved with 100% reliability and the system components still have a life-span of more than you'll keep the machine, then there is no argument against over clocking.  And it can, easily."

I can't agree with something that isn't always true.  I've been building machines and testing them for a lot of years and while it's true that the gaming market has driven the speed in the consumer market and lots of different software in the pro market (what I use) drives the need for faster machines the fact is we are kind of at a new place NOW; it's no longer true that we see huge performance increases.  So, what do you mean by "better"?  If I am using a 5 year old machine for basic computing tasks I'll see no performance benefit with a new computer let alone an overclocked one.

"Actually it is.  It depends on the task at hand."  

LOL, that's what I said!  It depends on the task and other components as well as sometimes performance numbers don't translate through the software one is using.

"Software instructions are executed at the speed of the system, over clocked or not."

Not sure what you are saying here, but I can assure you that not all software will take advantage of a dual CPU system. 
Not all software will take advantage of a 12 core machine.  I guess I'll wait for a better explanation of what you are attempting to say, but software is NOT always written to take advantage of the hardware.  And why do you suppose it is that software will run differently under different OS using the same hardware?   

"Really?  What aren't you sold on?"

Right now SSDs have a couple of issues, #1 they still seem to have an issue with performance degradation over time and #2. their MTBF hasn't been established over the long term.  They don't have unlimited read/write.

"It would merely be a system where my tasks tax the SSD less than the CPU.  I'm sure you understand that some tasks like noise reduction for instance.. are highly CPU intensive?  RAM is important here as well.  CPU/RAM is being taxed for far more of the required time to execute the command than is the SSD.."

If you build your computer correctly you will NEVER tax the HD even an SSD over a CPU/RAM.  Your SSD doesn't need to be taxed at all while working in Photoshop.  Right now today, the hard drive is the bottleneck and HD performance hasn't kept up with CPU RAM performance.  An SSD can't keep up with the CPU or RAM.

As I have stated, one should approach overclocking in a very methodical way and take it one step at a time.  When you say 30% performance increase what exactly do you mean?  For me it would mean my CAD software or Photoshop would do a task in 1 minute that used to take 85 seconds and I just haven't seen those numbers produced.  If you are overclocking a machine because you use one filter in Photoshop per month that would take advantage or it, I'd say don't bother.  We used to overclock a lot of machines, today not so much.

I guess the best analogy I can give is if I have a car with 400 HP and it can hit 0-60 in 4 seconds if I up the HP to 800 my car wont hit 60 in 2 seconds.  Benchmark performance and what one sees in the "real" world don't run in parallel.




« Last Edit: January 20, 2011, 04:56:56 AM by Gemmtech » Logged
Steve Weldon
Sr. Member
****
Offline Offline

Posts: 1447



WWW
« Reply #9 on: January 20, 2011, 07:42:52 AM »
ReplyReply


"I can't agree with something that isn't always true."

Well.. if you don't agree that a faster machine is better than a slower machine.. I have some stuff in the closet I'd be glad to sell you..


"Not sure what you are saying here, but I can assure you that not all software will take advantage of a dual CPU system. 
Not all software will take advantage of a 12 core machine."

Gosh.. it feels like you're really looking for exceptions to everything.  Its common knowledge not all software takes advantage of multiple cores. Agreed.  But we were talking over clocking.  ANY software will execute it's instructions faster on a faster clocked machine, than a slower clocked machine.  This is all I'm trying to say.  The total task might not reach a 30+% increase because many tasks are disk intensive, but I started this by saying you'll see a 30+% increase in 'motherboard' functions.  And you will.  Software won't change this.  If you're software is using 4 cores now, it will use 4 cores over clocked.  If it only uses one core, it will use that one core up to 30+% faster if properly over clocked.

"Right now SSDs have a couple of issues, #1 they still seem to have an issue with performance degradation over time and #2. their MTBF hasn't been established over the long term.  They don't have unlimited read/write."

I'm not buying this.  #1, the degradation is well documented and even though TRIM doesn't eliminate all of it, the performance is still so much higher than the next alternative that it's really no contest.  #2, MTBF has been established and published using the exact same methods as mechanical hard drives.  If you're trying to say their long term reliability hasn't been established I'd still disagree.  #3, NO storage device has unlimited read/writes.  This is a non-issue for any practical application.  If you plan on keeping your SSD 10+ years the we might have some issues.. but like I said, practical.


"If you build your computer correctly you will NEVER tax the HD even an SSD over a CPU/RAM.  Your SSD doesn't need to be taxed at all while working in Photoshop.  Right now today, the hard drive is the bottleneck and HD performance hasn't kept up with CPU RAM performance.  An SSD can't keep up with the CPU or RAM."

Really.. And where do you pull and store files from if not a HDD/SSD?  However, if you're thinking of eliminating the scratch disk for CS5 where it doesn't create temp files at all.. show me how.  I'm interested in learning.  If you're talking about doing it 100% on the motherboard (RAM/CPU) then my point about increased performance is made all the more.  And no one has ever claimed a SSD can transfer data as fast as the CPU/RAM.  I merely said there is a balance when performing certain tasks where the required time for CPU/RAM functions will exceed that of SSD functions.. for instance in your example above of performing Photoshop tasks and not accessing the storage media in the process.

"As I have stated, one should approach overclocking in a very methodical way and take it one step at a time."

The methods of over clocking were never at issue.  But if they were, I might not agree with this is 'some' circumstances.  It depends.

 "When you say 30% performance increase what exactly do you mean?"

An 30+% increase in all motherboard devices, CPU, RAM, Bus, etc, etc.. this is what changing the system clock that runs these things on does..  Of course 30+% might not apply to all CPU's, but they do apply to the more popular i7-920/930/950 models.  And it applies to the more popular 1156 socket CPU's as well.

" For me it would mean my CAD software or Photoshop would do a task in 1 minute that used to take 85 seconds and I just haven't seen those numbers produced."

No it doesn't mean this.  It would depend on how much of your task is a motherboard function, and how much is a GPU function, and how much is a disk function.  You will NOT see this number across the board.  But still, a 30+% increase in all motherboard functions for the cost of a $40 CPU cooler and proper selection of components is a significant increase.  Especially if it can be done with 100% reliability.

" If you are overclocking a machine because you use one filter in Photoshop per month that would take advantage or it, I'd say don't bother.  We used to overclock a lot of machines, today not so much."

Does anyone over clock a machine for one filter used once a month?  Really?  We're talking overall motherboard performance for every task you do.  In some cases this equates to just milliseconds, in others full seconds, and in others it can equal several minutes per task.  It doesn't really matter, if it can be done with 100% reliability/longevity and very low cost.. there is no argument against it.  Now.. I know some people don't want to be bothered.. but that's a personal choice and I respect their choice.  But I'm talking from a technical standpoint about there being no argument against it.  Which really explains Intel creeping up their speeds (over clocking) their original i7-920 at 2.66ghz to the i7-930 at 2.8ghs, to the i7-950 at 3.06ghs.. exact same chip, but different clocking defaults.  Some CPU's can do this with no drawbacks.  Why don't they go to 3.6-3.8ghz which any of these three can easily do on air?  Probably because they wouldn't sell many of their higher end CPU's.

And understand that there are several ways of over clocking.  When you do a very simple dirty blck increase, ALL motherboard functions increase.  RAM, Bus, etc.  These three chips handle this fine all the way to 3.6-3.8ghz.  Increasing the bclk is the only way to overclock these three.  But when Intel does it by increasing the speeds of the 920 to the 930 to the 950.. they're doing it differently where only the CPU clock rate is being increased and not all MB functions.  This is the big advantage to the upper level i7 CPU's.. you can over clock via blck.. or just the CPU. 

"I guess the best analogy I can give is if I have a car with 400 HP and it can hit 0-60 in 4 seconds if I up the HP to 800 my car wont hit 60 in 2 seconds.  Benchmark performance and what one sees in the "real" world don't run in parallel."

This is a pretty loose analogy which really doesn't apply to my original or subsequent statements.. a better one that reflects motherboard functions would be a car with 400hp shows 400hp on the dyno.  If you up the hp to 800hp.. it shows 800hp on the dyno.  The power increases from 400-800hp.  You now have 800hp to work with where before you had 400hp.  How you elect to use this power increase would depend on your tasking.





Logged

----------------------------------------------
http://www.BangkokImages.com
Gemmtech
Sr. Member
****
Offline Offline

Posts: 526


« Reply #10 on: January 20, 2011, 08:38:02 AM »
ReplyReply

Please do me a favor, don't state that something is 100% reliability/longevity or whatever, 100% doesn't exist.  

I believe my car analogy stands, you may have 800 HP to work with, but how much "better" (faster) is it in the real world?

One certainly can use a ram drive to speed things up, though it seems I read somewhere Mac OS has certain issues with these.
Naturally one can also use 2-4 HD in a RAID configuration which really speeds things up using HDs.

Seagate Momentus XT is a very nice alternative to the SSDs, I am currently testing them now.

"I'm not buying this.  #1, the degradation is well documented and even though TRIM doesn't eliminate all of it, the performance is still so much higher than the next alternative that it's really no contest.  #2, MTBF has been established and published using the exact same methods as mechanical hard drives.  If you're trying to say their long term reliability hasn't been established I'd still disagree.  #3, NO storage device has unlimited read/writes.  This is a non-issue for any practical application.  If you plan on keeping your SSD 10+ years the we might have some issues.. but like I said, practical."

You're not buying what?  We don't know how much the performance will degrade over time, some of the original SSDs really took a performance hit over time and we just don't know.  I remember the CD manufacturers telling us that they would last 200 years, well even some of my Kodak Golds are dying.  At least with mechanical HD we have had a long run.  You are saying the SSD long term reliability has been established?  How? they haven't been around long enough I do have SCSI and IDE HD that are over 10 years old.  I'm still using a 15K SCSI  HD in a RAID 0 config. works great.  

All I'm saying is why overclock just to overclock?  Why not test what you are doing?  Why run something out of spec if it yields nothing?  Software isn't always coded to run as efficiently as some believe it is.  Overclocking a system and then using Lightroom, Word, Excel, Photoshop (90%) would be a waste.  Overclock gradually and see if there's a performance gain for what you are doing.  Getting better benchmarks does NOT necessarily translate into an application running faster.  
« Last Edit: January 20, 2011, 10:27:32 AM by Gemmtech » Logged
walter.sk
Sr. Member
****
Offline Offline

Posts: 1328


« Reply #11 on: January 20, 2011, 12:10:16 PM »
ReplyReply

Well, it's interesting reading the back-and-forth between you guys, and I think I can grasp some of it.  Anyway, I was surprised to fine that my Alienware Area51-labeled X58 motherboard only had 3 RAM slots.  But, for using Photoshop CS5, plugins and some HDR programs, the 12GB will probably be as much as I need, together with the i7-930 CPU that came with the computer.  The computer, incidentally, was an exchange with Dell for a older XPS 730 that had problems over the almost 2 years I owned it, despite days on the phone and another mobo as well as many component replacements.

I'm not a gamer, other than for the Microsoft flight Simulator X, which doesn't really take advantage of my current system.

Thanks again for the help.

Logged
John.Murray
Sr. Member
****
Offline Offline

Posts: 893



WWW
« Reply #12 on: January 20, 2011, 01:02:09 PM »
ReplyReply

Walter - what didn't get mentioned is that the X58 chipset uses interleaved memory access, which is why you populate memory in 3's.  Because the memory controller is actually on the CPU  die, you get a great increase in memory throughput, especially compared to memory access via the "north bridge" on other intel chipsets.  

This interleaving also makes memory speed less important, if you see a significant price difference between 1033 vs 1600, you may want to consider that - in other words you will *not* see a a 33% increase in memory access performance as the numbers would imply - more like 10%

« Last Edit: January 20, 2011, 01:04:05 PM by John.Murray » Logged

walter.sk
Sr. Member
****
Offline Offline

Posts: 1328


« Reply #13 on: January 20, 2011, 01:17:38 PM »
ReplyReply

This interleaving also makes memory speed less important, if you see a significant price difference between 1033 vs 1600, you may want to consider that - in other words you will *not* see a a 33% increase in memory access performance as the numbers would imply - more like 10%
Thanks for the info about the interleaving.  As far as the price difference between the 1333MHz and 1600MHz kits, it is only a few dollars, and what I get is also what seems to be heat sinking on the higher speed.  At any rate, the memory is already on the way...
Logged
Gemmtech
Sr. Member
****
Offline Offline

Posts: 526


« Reply #14 on: January 20, 2011, 02:57:04 PM »
ReplyReply

I thought it was obvious that the X58 utilizes 1333 (or other speeds) triple channel memory A B C, you populate 3 or 6 DIMMS, either max 24GB or 48GB.  I do use mostly Intel motherboards. 
Logged
Steve Weldon
Sr. Member
****
Offline Offline

Posts: 1447



WWW
« Reply #15 on: January 21, 2011, 10:15:45 AM »
ReplyReply

"Please do me a favor, don't state that something is 100% reliability/longevity or whatever, 100% doesn't exist. "

Granted.  Bad choice of words.  I should have said 'it will be just as reliable as a non-over clocked system'.
 

"I believe my car analogy stands, you may have 800 HP to work with, but how much "better" (faster) is it in the real world?"

I think we agree on almost everything.. but it's fun splitting hairs because you learn things.  We both know the power of the engine/cpu will double, and we know that how well that power is put to the ground/tasked determines the value of the power increase.

"One certainly can use a ram drive to speed things up, though it seems I read somewhere Mac OS has certain issues with these."

AFAIK CS5 PS will still create a temp file on a regular drive.  I hope I'm wrong about this, I'd love to take all my PS tasks off the storage drives other than for pulling and storing files the files.  Anyone?

"Naturally one can also use 2-4 HD in a RAID configuration which really speeds things up using HDs."

Sure, but the performance still won't equal even a mid-range SSD.. and you can RAID SSD's for huge gains as well.

"Seagate Momentus XT is a very nice alternative to the SSDs, I am currently testing them now."

Funny you should mention these.  I reviewed a couple of these here, and I'm currently testing them in another configuration on a new system I'm working up.  On this system the System drive is a standard OCZRevoDrive which is screaming fast.  I was considering buying a Vertex 2 as a 'work drive' for my current image files I'm processing before off-loading them to a WD Black for my current archiving.  And then I remembered I had two 500g Momentus XT Hybird drives.  I put one in this new system (i7-950@3.8ghz/12g RAM2002/Revodrive, x58), put 20g's of RAW image files on the drive.. and another 20g on a current WD Black.  With the WD Black I was pulling roughly 70-80mbps burst and 65-70mbps sustained.  On the Momentus XT Hybrids I was getting 200-250mbps for the first 2-3g's of files, and then 50-55mbps sustained after that.  Basically, as I work, the first 2-3g's of files I pull are accessed at 200-250mbps which is great.  Most of my 'everyday' work while in LR/CS5/C1pro/etc falls in that 2-3g window.   Excellent performance for the the money.


"You're not buying what?  We don't know how much the performance will degrade over time, some of the original SSDs really took a performance hit over time and we just don't know."

If we don't know.. how do you know?  I've used SSD's extensively for over a year now.  I did benchmarks before and at intervals after.  There has been zero change since the drives settled in over the first month.  What does change, loading a SSD past 70% capacity will result in a roughly 3-5% performance hit.  You'll get an additional 2-3% performance hit from active garbage being processed through TRIM.. but I suspect this varies (a bit) from SSD to SSD because of different controllers and firmware versions.  I'd guess future TRIM enhancements will cut this down as well.  So overall, only 2-5% degradation from a brand new OS only SSD, to a fully loaded one month old SSD with zero degradation after.

Now.. if you're saying 20 years down the road this will decrease even further.. maybe.  I won't keep it that long.  And there is such a difference in performance between a SSD and HDD any decrease is irrelevant anyway.

"I remember the CD manufacturers telling us that they would last 200 years, well even some of my Kodak Golds are dying.  At least with mechanical HD we have had a long run.  You are saying the SSD long term reliability has been established?  How? they haven't been around long enough I do have SCSI and IDE HD that are over 10 years old.  I'm still using a 15K SCSI  HD in a RAID 0 config. works great. "

Read up on how manufacturers do extended testing to determine long term reliability.  It's not complicated and it's basically the same method they've used with HDD's and they've been right.  I personally won't be using any storage device 10 years.  Well.. that might change with SSD's.  We all know an adequately designed solid state device far outlasts even the best mechanical devices.. and the reason I don't keep mechanical hard drives past 4-5 years is due to 'mechanical' failure.. so perhaps SSD's will prove reliable far beyond.  What we already know is they're at least as good as mechanical HDD's..

"All I'm saying is why overclock just to overclock?"

This would depend on your needs no?  Some people are just equipment junkies and they get pleasure from this.  It's not for me to judge them.  For me personally, I over clock to gain up to a 30+% increase in overall motherboard functions.  It costs me nothing more than a $40 CPU cooler and the time to properly select my other components.  Like most others who do image processing as part of their livelihood, performance gains over the years due to more powerful CPU's, faster drives, faster and more RAM, have saved us hours each week.  And that's conservative.  So.. if I can get a 30+% increase in my CPU/RAM/BUS functions for $40.. that's cheap performance which I'll use.

"Why not test what you are doing?"

Who said anything about not testing?  We've never entered the part of conversation about 'how' to over clock.

"Why run something out of spec if it yields nothing?"

I'd agree, except it doesn't yield "nothing."  It yields a 30+% performance increase in motherboard functions.

"Software isn't always coded to run as efficiently as some believe it is.  Overclocking a system and then using Lightroom, Word, Excel, Photoshop (90%) would be a waste."

This just isn't true.  a.  All software runs it's instructions based on the clock rate of the CPU.  Faster clock rates means faster instruction execution.   This isn't debatable.  b.  Over clocking is a separate issue from core utilization. 

"Getting better benchmarks does NOT necessarily translate into an application running faster."

It does if your benchmark is replicating how you task your computer.. which is exactly why reviewers run multiple (many) benchmarks when they test/review a product.  So the reader can get a good indication on how that product performs for their use running their task.  Personally, when I review say a SSD.. I only use two benchmarks.  One that shows a theoretical maximum under the best of circumstances, and another which closely replicates how I personally use my SSD for imaging.  I don't run other benchmarks more relative to gaming or other tasks.  You need to learn which benchmarks are applicable to your needs/tasks.  If there isn't one, make one.  It's not difficult.

 
Logged

----------------------------------------------
http://www.BangkokImages.com
Steve Weldon
Sr. Member
****
Offline Offline

Posts: 1447



WWW
« Reply #16 on: January 21, 2011, 10:17:33 AM »
ReplyReply

the 12GB will probably be as much as I need, together with the i7-930 CPU that came with the computer. 

For still image processing I'd agree.  Video rendering will benefit from 24-48gb's and the fastest video card you can afford..
Logged

----------------------------------------------
http://www.BangkokImages.com
Steve Weldon
Sr. Member
****
Offline Offline

Posts: 1447



WWW
« Reply #17 on: January 21, 2011, 10:25:31 AM »
ReplyReply

in other words you will *not* see a a 33% increase in memory access performance as the numbers would imply - more like 10%

True.  From higher memory speeds along the figure is roughly 4-5%.  However, if you over clock the CPU via the blck method, more than the memory speed increases.  The CPU speed, the bus speeds, and the memory speeds.  Depending on how much you over clock you WILL see up to a 30+% increase depending on how far you go.  Not everyone will be able to go that far because CPU lots vary, case cooling varies, not everyone's power supply is up to it, they might not have used a great heatsink compound or CPU cooler.. so many will be limited.. but it's not hard at all to take a i7-920 at 2.66ghz from the proper lot (the lots are available for reference via google) and overclock to 3.8ghz with rock stability for a 33% increase.  This particular CPU is very popular for this reason.  The new i7-950 is already at 3.06.. so taking it to 3.8 obviously won't yield 33%.. but less.  Heck, the new 2600k Sandy Bridge CPU's are already being over clocked at 4.4 with full stability.  So yes, the percentage varies with application.. and simply increasing memory speed only yields a modest 4-5% increase because to take full advantage of this increase you'll need to increase CPU and Bus speeds which are both involved with memory channel transfers.
Logged

----------------------------------------------
http://www.BangkokImages.com
Gemmtech
Sr. Member
****
Offline Offline

Posts: 526


« Reply #18 on: January 21, 2011, 12:27:35 PM »
ReplyReply

"Granted.  Bad choice of words.  I should have said 'it will be just as reliable as a non-over clocked system'."

Actually no, we received 23% more calls regarding overclocked systems than non overclocked.  Lots of reasons, some of which you mention; not all motherboards like overclocking, CPUs (as mentioned the lots vary), memory etc.  And yes even software issues crop up.  I have found taking OC in gradual steps yields the best results, but again, only if you need it.  

"Naturally one can also use 2-4 HD in a RAID configuration which really speeds things up using HDs."

Sure, but the performance still won't equal even a mid-range SSD.. and you can RAID SSD's for huge gains as well.

I wasn't limiting the HD to just IDE, SATA , SCSI etc.  I was including SSDs in the mix.  

"If we don't know.. how do you know?"

I don't know, SSDs could be the greatest thing since cotton panty liners, but what if they aren't?  What if they turn out to be a lot like CD-Rs?  

"Now.. if you're saying 20 years down the road this will decrease even further.. maybe.  I won't keep it that long.  And there is such a difference in performance between a SSD and HDD any decrease is irrelevant anyway."

I'm saying within 5 years there could be a much larger hit in performance and what about reliability?  Data corruption?  There is a huge performance difference between mechanical HD and SSD sometimes, it depends on what you are doing with the computer.  Yes, the SSD boots up very quickly and loads the programs in a blink of an eye, but other times it doesn't make much of a difference.  I mean LR launches within 5 seconds with just a regular HD, who cares if it can be done in 2 seconds?  Windows boot time about 20 seconds with an SSD and 38-42 seconds with 7200 HD, it's significant for impatient people like myself.

"Read up on how manufacturers do extended testing to determine long term reliability.  It's not complicated and it's basically the same method they've used with HDD's and they've been right.  I personally won't be using any storage device 10 years.  Well.. that might change with SSD's.  We all know an adequately designed solid state device far outlasts even the best mechanical devices.. and the reason I don't keep mechanical hard drives past 4-5 years is due to 'mechanical' failure.. so perhaps SSD's will prove reliable far beyond.  What we already know is they're at least as good as mechanical HDD's.."

They've been right about what?  CD-R longevity?  Aren't USB drives SS?  I've had a few of these fail, data corruption and I couldn't recover it.  True, SSDs will never have a mechanical failure, but data corruption is another issue.  


"performance gains over the years due to more powerful CPU's, faster drives, faster and more RAM, have saved us hours each week."

I'll agree that over the years faster computers have save us lots of time, even now I'll sit and wait for a 3D image to be created by my CAD programs and I use the Dual Xeons, 48GB RAM and Nvidia Quadro 5000, however it's rare that I wait for an image to be processed today.  I mostly use LR and I suppose that's the reason, some filters in Photoshop take some time, especially panos, but for the most part does your image processing allow you to sit wait and have a drink?  

"Who said anything about not testing?  We've never entered the part of conversation about 'how' to over clock."

I've mentioned it several times.  

"I'd agree, except it doesn't yield "nothing."  It yields a 30+% performance increase in motherboard functions."

I when state yields nothing, I'm talking about real world use.  With today's computers if the benchmarks are a 30% performance increase that wont translate into a 30% increase in productivity, actually it may decrease productivity if you need to troubleshoot the computer.  IOW, when I say yields nothing I mean if you are using LR and processing an image you wont get done any faster with an overclocked system.  Want to stitch a few images together with Photoshop, it still wont be anywhere near 30%.  

Software isn't always coded to run as efficiently as some believe it is.  Overclocking a system and then using Lightroom, Word, Excel, Photoshop (90%) would be a waste.

"This just isn't true.  a.  All software runs it's instructions based on the clock rate of the CPU.  Faster clock rates means faster instruction execution.   This isn't debatable.  b.  Over clocking is a separate issue from core utilization."

What isn't true?  Try it, experiment as I have for the past 12 years.  Again, I'm talking about what you get in the real world.  All software runs differently depending on many factors one of which is the OS, hardware for hardware a program can run faster under different operating systems.  Today's computers are very fast and now only certain programs gain from higher performance hardware.  Even a supercomputer wont allow me to type a letter any faster; even voice recognition is getting very good.  I'm amazed how quickly Dragon on my IPhone works.

"It does if your benchmark is replicating how you task your computer.. which is exactly why reviewers run multiple (many) benchmarks when they test/review a product.  So the reader can get a good indication on how that product performs for their use running their task.  Personally, when I review say a SSD.. I only use two benchmarks.  One that shows a theoretical maximum under the best of circumstances, and another which closely replicates how I personally use my SSD for imaging.  I don't run other benchmarks more relative to gaming or other tasks.  You need to learn which benchmarks are applicable to your needs/tasks.  If there isn't one, make one.  It's not difficult."

Why not just test the programs you are using?  I'm more concerned with how much quicker my programs run and that's what I test most.  IOW, I'll use my CAD programs to determine how much specific hardware adds to productivity.  If I can render a 3D image in 30 seconds with one video card and the other knocks that time down to 20 seconds, that's huge and that's what concerns me.  My Dual Xeon machine doesn't process a LR image any quicker than my single CPU systems, I can't type a word doc any quicker and excel doesn't run any quicker.  I guarantee my dual Xeons have much nicer benchmarks.  If somebody tells me they want a computer for basic tasks, why install an SSD?  Obviously I wouldn't, but an SSD benchmark will look better, it just wont run the software any quicker.

Bottom line (And I'll use the car analogy again) how efficient is it at putting the extra power to use?  You seem to believe that all software will run faster with a faster computer?  I can take two cars, both with 400 HP, one can be (usually will be) faster than the other.  

Let me ask you this question and then I will know if you know what you are talking about.  Do you believe it is possible for a software program to run faster under a slower machine?  IOW if I have faster hardware in computer A can it run software faster than Computer B?
« Last Edit: January 21, 2011, 09:03:34 PM by Gemmtech » Logged
Steve Weldon
Sr. Member
****
Offline Offline

Posts: 1447



WWW
« Reply #19 on: January 22, 2011, 01:05:03 AM »
ReplyReply

"Actually no, we received 23% more calls regarding overclocked systems than non overclocked."

I'd fire the guy designing and over clocking the systems.  Clearly it wasn't done correctly.

"I don't know, SSDs could be the greatest thing since cotton panty liners, but what if they aren't?  What if they turn out to be a lot like CD-Rs? "

They won't.  SSD's are merely using existing technology which we've been using for years in other ways, and using it to create SSD devices.  You keep mentioning CD-R's, but there's really no valid comparison between silicone chip technology and the manufacture of optical discs.  And btw.. I've never had a bad experience with archival optical discs.  They pretty much work as advertised.  I've got a variety of cheap and high end cd/dvd's going back over 20 years with customer files on them which still work as well as the day I made them.
 

"I'm saying within 5 years there could be a much larger hit in performance and what about reliability?  Data corruption?"

The hit in performance isn't going to materialize in five years, if it hasn't happened in one.  Data corruption won't either as long as the SSD was designed with quality components spec'd within their design parameters.  Sure, it's possible a bad batch of chips could be put in SSD's resulting in failures.. but this would be a manufacturing defect and could happen with any device.  I just don't share your fears.

"They've been right about what?  CD-R longevity?  Aren't USB drives SS?  I've had a few of these fail, data corruption and I couldn't recover it.  True, SSDs will never have a mechanical failure, but data corruption is another issue."

Stop with the CD-R analogies.. they just don't hold.  The USB thumbdrive is much better and applicable.  Do you know why yours failed?  I've never had one fail, but I buy high quality thumbdrives in the same way I buy high quality flash cards for my camera.  We used to have people saying the same thing about flash cards.. telling us to buy 10 small ones instead of 1 large capacity card just in case.. Hogwash.. The only flash drives I've had fail is the infamous Lexar/Canon fiasco from some years back (firmware issue which they fixed), and a couple ultra-cheap SD cards where the physical case separated..

High-end flash memory cards would be a much better analogy.  We've been using them for over a decade.. they still work.  I've experienced no failures related to the actual components going bad.  I'm sure they do fail, but much less than a mechanical device would.  Mechanical hard drives fail like clockwork depending on the brand/model.. ask anyone who maintains RAID arrays.  If one goes, they'll all soon go.  And it's very rarely the solid state components which btw use much of the same technology as SSD's.. it's the mechanical components.  Usually stuck actuators, bad bearings, worn out heads..


"does your image processing allow you to sit wait and have a drink? "

Some do yes.  Time lapse rendering, focus bracketing, HDR, all regular parts of my processing.  Noise reduction would be a more common task which requires waiting.  Often 3-10 seconds per image almost all CPU/MB intensive resource tasking.  Even a simple action like adding a shadow border to an image can take 3-5 seconds.  In the realm of processing a wedding or something similar, a faster machine can save me hours.

"I've mentioned it several times. "

Right.  And I've ignored it several times as an unnecessary part of the discussion.  We've never talked about the 'how-to' part of over clocking.  Instead we've been discussing the 'why should I' part..


"I when state yields nothing, I'm talking about real world use.  With today's computers if the benchmarks are a 30% performance increase that wont translate into a 30% increase in productivity, actually it may decrease productivity if you need to troubleshoot the computer.  IOW, when I say yields nothing I mean if you are using LR and processing an image you wont get done any faster with an overclocked system.  Want to stitch a few images together with Photoshop, it still wont be anywhere near 30%."

You're having a problem with this concept so I'm going to address it just once more.  A 30+% increase in motherboard functions (RAM/CPU/BUS) is an undeniable fact.  I have NEVER said this results in an "overall" 30+% increase in productivity or image processing of any kind.  You just have a 30+% increase in motherboard functions.  When you process/render an image you have three main components in a computer that determine task completion time.  Motherboard functions (CPU/RAM/BUS), GPU, and storage devices.   Any given task utilizes these three areas in different amounts.  It depends on the task but image processing applies without doubt.  Do the math, a 30+% increase in one of these three areas MUST increase performance.  How much overall performance will be changed depends on tasking AND the level of your other components (GPU/Storage devices)

Now.. Add a fast SSD which can be several hundred percent faster than a HDD and we know we won't see a 200% increase in the overall task.  We'll only see a 200% increase in the speed on this one of three areas.  An SSD can really change the dynamics of a system, where before your bottleneck was the HDD which is very common, now your bottleneck can be one of the other TWO areas.  CPU or GPU.  Increase the speed of the CPU and now the GPU might be the weak link.

The ideal system for an individual is a system that won't bottleneck an unacceptably long enough period in any of these three areas.  However, what was 'acceptable' ten years ago is no longer acceptable today.  The bar has been lifted.  CPU/RAM/BUS performance has been and will continue to be perhaps the most vital 'overall' part of the equation.  Increasing these speeds 30+% can only result in faster performance of ANY task.  How much it affects the overall task, depends on how much the task draws on each of these three main areas.  CPU/RAM/BUS, GPU, Storage.

And btw.. fire whoever has been doing your over clocking.  If you're experiencing a higher failure rate or lost productivity of any kind, the person doing it needs to go..


"Why not just test the programs you are using?"

You'd think this would be obvious.  Reviewers don't know what programs you're using.  They assume so much, create benchmarks to reflect their assumptions, and we apply this to our choices.  It's not productive nor efficient for end users to first buy a load of equipment and then run a testing program focusing on their individual needs.  This is why we have benchmarks and why proper utilization of benchmarks is a necessary skill of a companies IT professional.

"Let me ask you this question and then I will know if you know what you are talking about."

Gosh you can be insulting.  Who's determined if you know what you're talking about?  I've been more than patient.

[/quote]
« Last Edit: January 22, 2011, 01:09:31 AM by Steve Weldon » Logged

----------------------------------------------
http://www.BangkokImages.com
Pages: [1] 2 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad