I work with business applications, many of which are database intensive. I can think of many examples which would greatly improve performance and application reliability, if I could run processes in parallel, rather than in series.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
8-Core (Clovertown) Mac Pro Benchmarks
- Thread starter MacRumors
- Start date
- Sort by reaction score
I work with business applications, many of which are database intensive. I can think of many examples which would greatly improve performance and application reliability, if I could run processes in parallel, rather than in series.
No. The number of core in the system does not imply you need more memory however it does imply you may need more memory bandwidth (depends on what you plan to use the cores for).
Running many applications, especially memory hungry ones, concurrently is what could require additional memory to run efficiently (depends on working set of the applications you plan to run).
With that said ... to get the most memory bandwidth potential in a Mac Pro you need fully populate peer banks with DIMMs. For example in the following graphic you would want to populate slots A1, A2, B1 and B2 before any other slots to get the widest data path to memory.
For more details on Mac Pro memory review Mac Pro RAM Expansion Details or the simpler to read Mac Pro Memory Guide (pdf)
Difference Between Two 512 + Two 2GB Mods Vs. Only Two 2GB Mods Question
One question about 512MB vs 1 & 2 GB mod bandwidth. If the Mac comes with two 512MB mods installed in Riser A and I buy two 2GB modules to put into Riser B - total 5GB, will the slower bandwidth (or whatever it's called) of the two 512MB mods slowing down the faster 2GB mods be offset by having RAM installed in both risers rather than pulling the 512's and only having two 2GB mods installed in Riser A only - total 4GB? The difference of 1GB of RAM doesn't seem like that part would matter if the 512 mods slow everything down.No. The number of core in the system does not imply you need more memory however it does imply you may need more memory bandwidth (depends on what you plan to use the cores for).
Running many applications, especially memory hungry ones, concurrently is what could require additional memory to run efficiently (depends on working set of the applications you plan to run).
With that said ... to get the most memory bandwidth potential in a Mac Pro you need fully populate peer banks with DIMMs. For example in the following graphic you would want to populate slots A1, A2, B1 and B2 before any other slots to get the widest data path to memory.
For more details on Mac Pro memory review Mac Pro RAM Expansion Details or the simpler to read Mac Pro Memory Guide (pdf)
Notwithstandign such long-standing facts, there are still some MS fanboys here who think Windows is better for multicore usage (not to mention multitasking, which has been ALWAYS better in OS X)...
Zune is dead, Windows is dead...face it.
More Cores
Well, with the Intel Core roadmap for 2016 possibly getting up to close to 300 heterogeneous cores per motherboard/PC, certainly Windows is out of the race but OS XX 20.x will have to have evolved to probably DAML/OWL Semantic Web Ontology based, with System Strategy and System Policy Reasoners, i.e. a Cognitive OS with a flexible Operating Envelope. I think that would definitely make Safari snappier!
Notwithstandign such long-standing facts, there are still some MS fanboys here who think Windows is better for multicore usage (not to mention multitasking, which has been ALWAYS better in OS X)...
Zune is dead, Windows is dead...face it.
Well, with the Intel Core roadmap for 2016 possibly getting up to close to 300 heterogeneous cores per motherboard/PC, certainly Windows is out of the race but OS XX 20.x will have to have evolved to probably DAML/OWL Semantic Web Ontology based, with System Strategy and System Policy Reasoners, i.e. a Cognitive OS with a flexible Operating Envelope. I think that would definitely make Safari snappier!
Well, this reminds me of that presentation given by Steve Jobs (on the new G4s, I think), where he said, for a certain specific technology ("latent semantic" or whatever) that:
"I have no idea what that is; but trust me, it works..!"
But you're missing the fact that the 8 cores are at a slower clock speed. If you compare 4 versus 8 at the same clock, you're looking at a respectable 47% improvement.
I almost NEVER use handbrake from an optical DVD. That makes no sense to me. Why would you do that?
To rip DVD's. Why add additional, unnecessary steps?
Apple REALLY needs to get apps like quicktime and iTunes to run on any number of cores. Even if they don't use multiple cores on a single file, it should be a piece of cake to get them to process multiple files at once. If I want to convert eight files, it should just run each conversion at once on a separate core - it's the equivalent of running eight copies of the app (which shouldn't be necessary).
I'd love to see them run Logic Pro - it supports four cores finally, and I'd like to know if they just upped it to four or if it goes beyond that.
The bandwidth of DIMMs doesn't really change with their capacity (assuming their timings are the same). It is the number of active channels that gives you the increase in memory bandwidth.
I believe the memory controller will interleave the lower 512MB (since you have a branch populated with 512MB) of each DIMM resulting in full bandwidth access (4 channels) to that range of memory (512 x 4 = 2 GB) with the remainder of the 2 GB DIMMs (the remaining 3 GB) only getting the benefit of a single branch (two channels).
Personally I wouldn't purchase 2 GB DIMMs given their extra cost in relation to two 1 GB DIMMs... in other words use more of the memory slots you have ... unless you really want to leave room to add more RAM later.
yup, and my webpages will load in the blink of an eye... definitely worth whatever apple will charge.
seriously though, how hard is it to get a program to multi-thread? (if thats the right term; being a complete programming novice, i've no idea)
It depends on what the program does. Some programs don't lend themselves to multi-threading at all and others practically require it. It can be quite a chore to go back and multi-thread an existing program.
I would be happy if existing programs couldn't hog all the resources of the OS whether they were multi-threaded or not. Windows is the worst offender by far, but OS X has allowed an ill behaved application to dominate my computer now and then also.
Usually true, but with FB-DIMMs the 256MB and 512MB only use one of the onboard buffer channels (as in buffer channels on the module itself). Whiel the timings and I/O speeds are the same as other larger capacity modules, they are only capable of sustaining 1/2 the full bandwidth load. OTOH, the latency is a bit lower with 256MB and 512MB modules as the modules' onboard memory controller only has to deal with a single channel buffer.
Anyway, If you can live with 2GB in a Mac Pro, then 4x512MB wouldn't be a bad deal. I'm not sure if you really could get by with only 2GB in one of these... If so, you probably don't need a Mac Pro and that's a whole different discussion.
But If you're going with 4GB or more, you definitely will want 1GB or 2GB modules to capitalize on the increased bandwidth abilities and I wouldn't recommend mixing half gate buffer and full gate buffer modules within the same system.
Since the RAM all runs at the same speed and timings, no. The 512MB half-gate modules won't slow down the other RAM. However, any data going through the 512MB modules will see a very slight improvement in latency, but the upper ceiling for bandwidth out of those modules will be about half of what you would get out of a 1 or 2 GB module.
Personally, I would want all my RAM to be consistant...
An unlikely scenario. Don't expect any price drops on mac pros for a long time after clovertown chips are in them.
Humm ... never ran across any documentation to that effect. You have an online reference? (I believe you, just like having reference information)
Agreed.
Personally my Mac Pro has 1 GB DIMMs in A1, A2, B1, and B2, and 512 MB DIMMs in A3, A4, B3, B4 (since the model I picked up from Apple had the four 512 MB DIMMs in it)... yields a total of 6 GB of RAM.
This result in each channel connecting to a matching DIMM mix in a matching progression... ideally allowing the memory controller to have an easier time of muxing access to the RAM (in terms of a more optimal interleaving configuration).
Of course if 512 modules only can feed the AMB at half bandwidth relative to 1 GB (or greater modules) then removing them from the system could improve overall throughput (assuming they get hit) ... however latency difference may mitigate that ... hard to answer without real-world profiling with the tasks you most often do.
...off to find docs on Intel memory controller for information on how it handles things...
Thats the beauty of going Intel, you dont drop the price ? DELL, HP and the other competitors WILL.
Taking advantage of the multi-cores.
Previous question: How hard could it be to take advangate of the multi-cores.
The first thing is that it depends on what you are starting with. If you have zero code out there, you can come up with a nice design for your program that takes advantage of as many cores as you throw at it. If on the other hand you have large chunks of legacy code that was written in the time of single cores, it may be close to a re-write to fully take advantage of the hardware. In some cases it will be easier in some cases to throw the old code away.
But some of it is imagination, if you can look at a problem and the solution you orginaly came up with, and using your imagination look at the problem at hand in inovative ways, parts of the programs could be re-written to take advantage of the hardware and other parts can be left alone (for the short term). This is an incremental step, you gain X% in one area and little to nothing in another area. The key is to determine what your program spends most of it time doing and re-write/re-design that section of the code for the biggest short-term gains.
I remeber working in assembler and selecting the correct combination of instructions based on their function and the number of CPU cycles it took to execute each instruction. Sometimes a set of 12 instructions was faster than a different set of 8 instructions in accomplishing the same result. Use your imagination and look at the problem from a different angle. If your brain only sees a number of serialized steps, you won't be able to come up with anything that takes advange of the hardware.
What you start with (old code) and your imagination can get you there quicker or slower.
Short answer: It depends.
Previous question: How hard could it be to take advangate of the multi-cores.
The first thing is that it depends on what you are starting with. If you have zero code out there, you can come up with a nice design for your program that takes advantage of as many cores as you throw at it. If on the other hand you have large chunks of legacy code that was written in the time of single cores, it may be close to a re-write to fully take advantage of the hardware. In some cases it will be easier in some cases to throw the old code away.
But some of it is imagination, if you can look at a problem and the solution you orginaly came up with, and using your imagination look at the problem at hand in inovative ways, parts of the programs could be re-written to take advantage of the hardware and other parts can be left alone (for the short term). This is an incremental step, you gain X% in one area and little to nothing in another area. The key is to determine what your program spends most of it time doing and re-write/re-design that section of the code for the biggest short-term gains.
I remeber working in assembler and selecting the correct combination of instructions based on their function and the number of CPU cycles it took to execute each instruction. Sometimes a set of 12 instructions was faster than a different set of 8 instructions in accomplishing the same result. Use your imagination and look at the problem from a different angle. If your brain only sees a number of serialized steps, you won't be able to come up with anything that takes advange of the hardware.
What you start with (old code) and your imagination can get you there quicker or slower.
Short answer: It depends.
In this class of RAM the 2GB sticks are now less than two 1GB sticks. So I don't see why buying only 1GB sticks would be advisable any more.
AV was saying that the 512 sticks run half as fast as 1 and 2GB sticks. Is that not correct or did I misunderstand what he meant?
It already is the lowest.
I agree with you 100%. Apple Mac Pro is already the lowest priced Dual Woodcrest product on the market. Dell costs a lot more. So why would you expect Clovertown to impact the price of the existing line that is already at a rock bottom price Alpinism?
Apple started lowest and none of the other vendors followed so your premise is not turning out to be true in the real world.
I don't rip DVDs. I rip DVD Images created with Toast from EyeTV broadcast recordings. So it's not an extra step. If you mean why not export direct to mp4 from EyeTV the answer is because they look like C**p. You want a really good looking mp4 file, Handbrake is the only way to go IMHO. And the maximum quality Toast encode to DVD image is also the best way to provide Handbrake with a superior master to rip from.
Anyway that's why what I use is ready for an 8-core Mac Pro now.
I'm thinking about my future 8 core Macpro:
2 questions for you:
- Do you think the 8 core proc will produce a lot more heat than the current core duo 2 ? I'm asking because I need a very quiet computer ...
-As always: shall we expect this one in the Macpro before 2007 ?
thx !
2 questions for you:
- Do you think the 8 core proc will produce a lot more heat than the current core duo 2 ? I'm asking because I need a very quiet computer ...
-As always: shall we expect this one in the Macpro before 2007 ?
thx !
I don't know if I'd expect that either. Has intel dropped the prices on dual core version yet? Or just introduced the quad core at higher prices?
You asked why anyone would use handbrake to rip from optical disk. I answered your question. People do it all the time, it's very common to rip DVDs.