Compatible ram for Mac Pro 2010

[Philocetes]

Based on some great threads here, I decided to give my mac pro 2010 a mid-life upgrade. Got a hex core on order and looking to up the ram. So, its currently a single quad core with 4 memory slots. I currently have 12 gig with 3 4 gig sticks. Thinking I'd like to go to 24 or 32 gig. Or go nuts and get 48 gig?

In any case, to keep a reliable machine, just how compatible does the ram have to be? Anybody have favorite sources of good ram at good prices? I have purchased from one outfit who claims their ram is better because it has the metal heatsinks and a mac proprietary temperature sensor. I see on amazon, ram that is quite a bit less expensive, but appears to lack a heat sink and do doubt no temp sensor. I think this stuff is simply your bread and butter 1333 ECC ram harvested from retired servers, or NOS stuff laying around.

I did a few forum searches and was not able to find any discussion of the relative merits of brands of ecc 1333 ram, so a new thread. Apologies if I missed the same.

Another quick one: I am reading threads here and elsewhere regarding pcie ssd. If I install an ssd pcie card, format with guid fs and so forth, can it be bootable? Are there special apple requirements for such a card. I see OWC has some stuff along these lines, but wondering if there are less expensive options that work just as well. Don't want to pay the apple tax unless its required.

I have an ssd boot drive now that is plugged into the regular sata drive bays using a physical adapter--maybe going pcie would not be a signifiant speed boost...

Thanks.

 

[h9826790]

The 5,1 almost accept any DDR 3 RAM, ECC vs no ECC (or mix with them), UDIMM vs RDIMM (CANNOT mix), mix use different size / speed sticks. So, no need to worry compatibility.

The 1333MHz ECC server RAM is so cheap now, 32G only cost something like $75. 48G sure will cost more, if you don't need that, there is no need to go for the more expensive 16GB stick.

Yes, the PCIe SSD should be bootable. And yes, you are also right, you should not able to feel any significant improvement over the SATA SSD for normal use.

 

[Philocetes]

Thank you so much for the advice, I thought might be the case. Reading things here keeps me from spending money that noobs might spend, not knowing what the good deals are--noob that I am myself.

 

[orph]

pci SSD is relay good for a scratch drive if your doing something that requires it & 32GB of ram is more than most people need :E for normal use i hardly ever go past 10GB use it's only once i do something like AE work that iv seen my ram full.

got to ask do you need more than 12GB of ram ? have you seen much 'swap used' in activity monitor, doing 1080p video editing i saw an improvement going from 10-18GBram but past that for most things it's unusual to use more.
unless your running some pro apps that eat ram

but 32gb of ram is super cheep on ebay so >_> not to bad

 

[Philocetes]

Orph, you make a good point, and I could have been more disciplined about ensuring that I need all of 32 gig, but it is pretty cheap. I do have plans on using final cut pro and editing video and I have read that uses quite a bit of ram, so I felt like getting the ram all in one shot rather than going back or when I need more.

 

[MarkC426]

The 5,1 almost accept any DDR 3 RAM, ECC vs no ECC (or mix with them), UDIMM vs RDIMM (CANNOT mix), mix use different size / speed sticks. So, no need to worry compatibility.

The 1333MHz ECC server RAM is so cheap now, 32G only cost something like $75. 48G sure will cost more, if you don't need that, there is no need to go for the more expensive 16GB stick.

Yes, the PCIe SSD should be bootable. And yes, you are also right, you should not able to feel any significant improvement over the SATA SSD for normal use.

I don't know what type of ram that is for $75 (about £60), but from Crucial UK 32gb is £195 gbp.
I recently got 32gb Hynix from mrmemory, because I wanted the same make that my mac originally came with (but that's just me, there are probably cheaper alternatives).

 

[orph]

cmp ram is relay cheep on ebay MarkC426, it's the same ram they used to use in servers so pulled server ram is cheep http://www.ebay.co.uk/itm/Hynix-4x8...306819?hash=item54270b5483:g:jMMAAOSwIwhWTZm0
^^ snap hynix (thats the ram i got i think)
when i say cheep i still remember spending £50+ on 8GB of ram for my 3.1 >.< pain

there are good things that you do need lots of ram for, last time i used AE it ate it all, but for most normal users it's not needed id gess 12GB is still quiet a bit, open activity monitor and check in the memory section if swap is used (you may just want to leave activity monitor open and check now and agen).

 

[ITguy2016]

The use of memory with heat sinks was applicable to the 1,1 - 3,1 Mac Pro's. The 4,1 and 5,1 do not require memory with heat sinks. While there's no harm in using memory with heat sinks I wouldn't pay extra for them. My 5,1 has 32GB of RAM that does not have a heat sink. It run 24 / 7 and I've never had a problem with it. The best advice I can offer is to buy a decent quality RAM. Not bargain basement nor the most expensive.

 

[Kolvir]

I'll second the registered memory pulled from servers on ebay. I put in 24GB, 6 x 4GB RDIMMs very cheap. Just remember you can't mix registered and non registered ram.

You might not even be able to mix different registered lots, so get it all from the same guy if you can. Can anyone verify or clarify this point?


If you can get it really cheap, I upgraded about 9 months ago so I'm sure prices are different, I'd max it out.

 

[orph]

yep the heat sinks are like "go faster stripes" for 4.1/5.1 ram :E it wont do anything.
the two sticks that came with this 5.1 had no heat sinks and my 32Gb have no heat sinks.

if you want you can run http://www.kelleycomputing.net/rember/ to test ram for errors

 

[Philocetes]

...
if you want you can run http://www.kelleycomputing.net/rember/ to test ram for errors

Thx, I will use this. Ram should show up in the next couple of days. I have a vague recollection from the olden days of database server support that sometimes a memory error would not be revealed until the server got busy enough to use that ram. A test ought to be able to exercise the whole lot within the return time window.

 

[shaunp]

Have a look on the crucial website - just put in the model of Mac you have and it will show you compatible RAM. I've been buying RAM from Crucial since around 2009 and it's always been good and their prices are good too.

 

[flowrider]

A post I made in another thread:

My RAM supplier since the 80s has been:

http://www.datamemorysystems.com/apple-mac-pro-memory-upgrades/

They are an excellent source and I recommend them highly. They are Mac specialists and their stuff is guaranteed for life.

BTW, I just got a check for $310. from the Attorney General of the state of Arizona because of the alleged price fixing by DRAM Manufacturer's between 1998 and 2002

Lou

 

[Philocetes]

Thanks for all the ram vendor suggestions. I purchased the ram before you made them, so I will just keep them in mind next time. I normally work with amazon with ebay as a second choice, and creating a new login for some other site only when I can't get my product from the first two. On amazon, I found 32 gig (4 x 8 gig) of 'a-tech' ECC ram for $88. For what its worth they had 'made for mac' or some such stickers on the box. The Ram went in fine and I ran the rember memory check program against it and it reported no errors--a single round of tests ran for several hours.

The ram chips were too hot to touch for more than a few seconds, I presume that is normal. Gonna run a tad hotter once I get the new hex chip installed and the ram goes from 1066 to 1333. But all is within spec. The Hex chip and special hex wrench are incoming in the us mail--hopefully today or tomorrow.

 

[orph]

:O dont go touching hot things it hurts,

that ram test pushes the ram way past any normal use a bit like
furmark (but less likely to make you pc melt).

moving up to 1333 will not bring any heat change (not noticeable at least).

(even if it's hot it still runs about half the temp of macpro 3.1 ram )

 

[Philocetes]

My understanding of logic devices is that they mostly generate heat during the transition between conducting and non-conducting--number of transitions direct correlation with clock speed. So, when they are conducting, no heat because no resistance, when non-conducting, no heat because no current. The transition period is tiny, but it exists. Therefore: faster cycling generates more heat.

So, my thinking is that 1333 is 22.5 % faster and hotter to the same degree--but glad to hear it is of no concern regardless. I appreciate the practical experience people are sharing here--saved me a lot of money by not going with that premium vendor where the ram had the heat sink and temp sensor on it, but no need for either.

@orph, that memory tester ran for a long time--I had it set for 10 cycles, but after an hour and it was still on first cycle, I set it to 1 cycle, which still looks like it ran for 4 hours--overnite. It was a great reassurance and good to stress the ram while its within the 30 day return policy, or whatever.

I have a 3,1 at home, my wife uses it to remote into her office to work from home--so its income producing property! IIRC, it has a few cornfield type rows of 1 or 2 gig sticks with their heat sinks, just roasting away. The beauty of the MP is sleep mode--not warming the house unless we are using them.

 

[orph]

maybe it will be a tad hotter :E but relay nothing you will notice compared to the 3.1 the rams frosty cold, to get the ram hot you need to stress it hard.
iv been playing games & using firefox etc over the last while & HW monitor shows ram slot 3 peeked at 60c (thats the hottest one i gess slot 1 peeked at 50c), there all about 44-52c at the mo
(i think 1333 ram is only vary vary slightly faster in the macpro might be 25% higher in number but i think the actual speedup was more in the 3% kind of zone in tasks that actually stress ram, and not much dose stress ram)

ps yep that ram test is super slow i did it over night on 1 pass think it dose a bunch of different things

pps did you see any speedup with having way to much ram ^^

 

[Philocetes]

@orph--yes, its lots faster opening mail and browsing the interweb....

Actually, its Logic pro that would use more ram, and I am thinking about getting Final Cut Pro and playing around with video, so...

 

[h9826790]

@orph--yes, its lots faster opening mail and browsing the interweb....

UPDATE 1: Sorry, I miss read your post. You mean the system work much faster now because you have more RAM, but not the RAM speed from 1066 to 1333. Please ignore my original comment.


---------------------------------------------------------

That speed improvement has nothing to do from 1066 to 1333. Especailly browsing the web or opening mail is not memory bandwidth intensive task, but more network speed dependent.

The RAM run at 1333 can process more data per second. However, it's more like capable to do more work at the same time, but not able to finish the work faster. Because the RAM either run at 1066MHz CL7, or 1333MHz CL9. Therefore, if you put some data in, they actually come out more or less at the same time.

7x(1/1066) = 0.0065666041s
9x(1/1333) = 0.0067516879s

even though this is not the actual time required to finish the process, but a simple and good enough way to compare the time required to finish the same job.

as you can see, the difference is about 2.8%, nothing close to 25%.

My understand is that 25% is more like the capacity (bandwidth) increment, but not really speed increment.

So, unless you are doing something that really memory bandwidth intensive. Then you may see some difference. However, for opening mail / browsing the web. I am sure the bottleneck is not there, and you should not able to see any difference.

If there is actually a significant improvement, it's from something else (e.g. something was reset due to hardware changed), but not the memory clock difference.

 

[Philocetes]

Sorry, my dry sense of humor is often lost in forums... I did put the wink after my comment--I was deliberately saying something ridiculous.

I am very familiar with computer technology including hardware--and performance tuning. The most dramatic speed boost from ram is when you have memory-bound* applications in a virtual memory environment. When a program wants to access memory that has been paged out to disk, a page fault occurs and the virtual memory manager reads the data from disk and puts it in ram. This takes maybe 10,000 times longer than reading data already in ram. (for spinning platter. SSD is obviously a lot closer to ram speed than a spinning platter disk--but still a long way off. I haven't worked the numbers to compare it to ram)

There are other performance bottlenecks/improvements like the bus speed between the cpu and ram and other devices, but these differences are minor compared to an app that is memory bound and is frequently generating page faults. One thing we look at in windows database servers is page life expectancy which is a very good indicator of whether an app would benefit from additional ram.

So... Adding ram makes a machine more performant to the extent that the app needs more memory than is physically available on the machine. Mail and web browsing have small memory requirements, so they would not benefit from the large amount of memory I added to my 2010 MP.

* Memory bound is referring to an app who's performance is constrained by the amount of ram present. CPU bound, etc. In performance optimization, if the cpu is sitting at 90% idle most of the time, more cpu will not improve performance. If cpu is pinned to 100%, there is potential improvement by increasing cpu resources. CPU bound can be an indication that ram is not a constraint, otherwise the cpu would spend a lot of idle time waiting for disk i/o to complete. These are general thoughts, not to be rigidly applied to every situation. Sorry, I just enjoy talking about performance tuning.

 

[h9826790]

Sorry, my dry sense of humor is often lost in forums... I did put the wink after my comment--I was deliberately saying something ridiculous.

I am very familiar with computer technology including hardware--and performance tuning. The most dramatic speed boost from ram is when you have memory-bound* applications in a virtual memory environment. When a program wants to access memory that has been paged out to disk, a page fault occurs and the virtual memory manager reads the data from disk and puts it in ram. This takes maybe 10,000 times longer than reading data already in ram.

There are other performance bottlenecks/improvements like the bus speed between the cpu and ram and other devices, but these differences are minor compared to an app that is memory bound and is frequently generating page faults. One thing we look at in windows database servers is page life expectancy which is a very good indicator of whether an app would benefit from additional ram.

So... Adding ram makes a machine more performant to the extent that the app needs more memory than is physically available on the machine. Mail and web browsing have small memory requirements, so they would not benefit from the large amount of memory I added to my 2010 MP.

* Memory bound is referring to an app who's performance is constrained by the amount of ram present. CPU bound, etc. In performance optimization, if the cpu is sitting at 90% idle most of the time, more cpu will not improve performance. If cpu is pinned to 100%, there is potential improvement by increasing cpu resources. CPU bound can be an indication that ram is not a constraint, otherwise the cpu would spend a lot of idle time waiting for disk i/o to complete. These are general thoughts, not to be rigidly applied to every situation. Sorry, I just enjoy talking about performance tuning.

I should learn how to read the real meaning of emoji

 

[Philocetes]

I should learn how to read the real meaning of emoji

Its all good, fun to hang out and chat. I have a perceptual blind spot--I forget that people in a forum have no way of knowing my knowledge or my use of ridiculous statements as a form of humor.

 

[ITguy2016]

UPDATE 1: Sorry, I miss read your post. You mean the system work much faster now because you have more RAM, but not the RAM speed from 1066 to 1333. Please ignore my original comment.


---------------------------------------------------------

That speed improvement has nothing to do from 1066 to 1333. Especailly browsing the web or opening mail is not memory bandwidth intensive task, but more network speed dependent.

The RAM run at 1333 can process more data per second. However, it's more like capable to do more work at the same time, but not able to finish the work faster. Because the RAM either run at 1066MHz CL7, or 1333MHz CL9. Therefore, if you put some data in, they actually come out more or less at the same time.

To a point. I used to chuckle at the Intel TV commercial where they claimed their processors led to a faster Internet experience. I think this claim was in the days of dial up / early broadband. Like you I was like "It's the network, not the end user system".

Today that's not entirely the case. While network speed is important I've found the end system is important too. Due to all of the active content on today's web one needs a more capability on the end system than days gone by. For instance I have a 1.8GHz single processor G5 (the second least capable of the G5 systems). Browsing the web with it is noticeably slow. Sometimes the spinning disc appears for tens of seconds at a time. Looking at Top or Activity Monitor shows the CPU is almost completely consumed 100% of the time by TFF. If I disable Java Script then the browsing speed improves significantly with the CPU usage decreasing dramatically.

As for the memory speed in the vast majority of cases one will not see a noticeable increase in performance from 1066 to 1333 (except in synthetic benchmarks which measure memory bandwidth). I will always choose the faster memory (usually when buying an older system the faster memory is the only choice as manufacturers cease production of the slower memory).
[doublepost=1475172261][/doublepost]

So... Adding ram makes a machine more performant to the extent that the app needs more memory than is physically available on the machine. Mail and web browsing have small memory requirements, so they would not benefit from the large amount of memory I added to my 2010 MP.

I am surprised by the amount of memory a web browser can consume. While it may not be in the multi gigabyte size it's certainly more than I would have thought it would be. I've read people stating a single tab in FireFox can consume 100MB of memory. If you have 10 tabs open that's 1GB of memory right there (though much of it might be paged out to disk).

 

[orph]

you know how to tell if something is Memory bound so few people know how to tell, there is hope.
i shude have just skipped my 'how much ram are you using bit at the start' sorry not used to people who know how to read ram usage.

id be interested in how logic scales & uses ram.

 

[h9826790]

To a point. I used to chuckle at the Intel TV commercial where they claimed their processors led to a faster Internet experience. I think this claim was in the days of dial up / early broadband. Like you I was like "It's the network, not the end user system".

Today that's not entirely the case. While network speed is important I've found the end system is important too. Due to all of the active content on today's web one needs a more capability on the end system than days gone by. For instance I have a 1.8GHz single processor G5 (the second least capable of the G5 systems). Browsing the web with it is noticeably slow. Sometimes the spinning disc appears for tens of seconds at a time. Looking at Top or Activity Monitor shows the CPU is almost completely consumed 100% of the time by TFF. If I disable Java Script then the browsing speed improves significantly with the CPU usage decreasing dramatically.

As for the memory speed in the vast majority of cases one will not see a noticeable increase in performance from 1066 to 1333 (except in synthetic benchmarks which measure memory bandwidth). I will always choose the faster memory (usually when buying an older system the faster memory is the only choice as manufacturers cease production of the slower memory).
[doublepost=1475172261][/doublepost]
I am surprised by the amount of memory a web browser can consume. While it may not be in the multi gigabyte size it's certainly more than I would have thought it would be. I've read people stating a single tab in FireFox can consume 100MB of memory. If you have 10 tabs open that's 1GB of memory right there (though much of it might be paged out to disk).

I totally agree that a newer system can be significantly faster on browsing the web. Even just a different browser can make 100% speed improvement. However, what I want to point out was the significant browsing speed increment should not be coming from the RAM running at high clock speed.

TBH, I also quite surprise that the software company keep comparing their browser's performance. For me, the internet content should be very "light weight" for modern computer. So, hardly can be a benchmark, but it seems I am totally wrong. At least some internet content like Flash can be very heavy weight.

I am no where near any expert in this area. I doubt if a G5 really cannot browse the web with proper speed (in terms of hardware). However, due to lack of software update, the old browser (software) actually cause more problem than the hardware. Because the old browser may not able to decode the new stuff correctly and effectively (even though hardware resource is not an issue).

And it's new to me that Intel actually put something in the CPU to improve internet experience. Thanks for teach me something today. Do you have any idea what's that? A hardware decoder?