Thats because it was loading data and processing. It takes awhile to process the write data. DDR3 SDRAM can go up to 10066 MB/s. CPU has a max transfer rate of 32GB/s. Are you done yet?
Last edited:
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.
SATA Express meets the '09 Mac Pro - Bootable NGFF PCIE SSD
- Thread starter handheldgames
- Start date
- Sort by reaction score
Just a good stable bootable combo. So many variables and the thread's opening post hasnt been updated in awhile..
The most stable way indeed is just plug the SATA SSD into one of the native SATA 2 port. That's 100% sure both OSX and Bootcamp bootable, able to update firmware, able to use boot manager, no extra boot time for initialising the card, etc.
And this basically won't affect your OS operating speed, but just very limiting sequential speed.
Tested it.
The cMP's memory hits around 2GB/s.
Forget about your apps ever writing scratch files at anything close to the numbers you cite. Your own tests had the apps writing to cache and scratch at 25MB/s under maximum load.
Creating a ram drive isn't the proper way to test memory bandwidth. Here is my results from novabench. 5.31GB/s
That was on my cMBP, i just got my 2012 cMP, installed cpu's and ect. Still waiting on my m.2 drives to come in. I'm only able to afford two 512GB m.2's right now. Will need to wait until next month to get two more to get it up to it's maximum speed with a 4x raid 0. I'll create a 2x raid 0 when i get them just to test. Should still get over 2GB/s.
Also i know you know this but the read speed was at 790 something MB/s and 25MB/s write. I already explained that AE has to process post processing and ect so it doesn't write at full speed. It grabs the footage and then applies the filters to the frames then writes the rendered data. That was just a test to show that it would utilize a drive higher than Sata 3. I already pointed this fact out to you multiple times. The fact that you continue to twist my words around just so you can continue to spread disinformation because your afraid of hurting your ego just shows you shouldn't be in IT.
This is my current read/write speed averaged out using 2x 512GB 850 evo ssd's striped as raid 0 in a Sata 3 pcie card. Theoretical max speeds should be 800MB/s. It says the fastest it went during the test was 751MB/s Write and 767MB/s Read
Last edited:
Nova bench has been bust for years and doesn't produce any real world result comparable to a RAM disk.
I see commercials and campaigns being edited for the world's top brands daily. Look at this simple observation.
The nMP has replaced the cMP in many many studios. It doesn't have enough local storage for large 4K/RAW footage and most editing teams use shared resources on a network. They are accessing video content through the gigabit Ethernet Adapter that can barely manage 100MB/s bandwidth.
How do they manage it? How has anyone managed that for so many years? Because the apps are smart enough to utilise very little bandwidth when they write compressed or low res files to temp/swap/cache. The only issue they face is slower project loading times but once the content has loaded it doesn't hold them back.
The other issue is capacity. If you try to use m.2 drives or even a local RAID in your cMP you will run of space pretty quickly. M.2 were originally intended for slim laptop computers, not video editors. The best use you can make from them currently is as a temporary working drive to load smaller projects from.
I completely support your desires for having obscenely fast computers. But I've been there, I tested these m.2 RAIDs in real world applications. I posted gelatinous benchmarks comparing nMP, cMP, OS X and Windows. The biggest change you can make to speeding up your workflow is...moving to Windows (even on the same cMP, playback was much quicker, video acceleration is superior and Media Encoder rendering times was 4X faster).
Last edited:
With a 4x ACHI blade m.2 raid you can get up to 4TB of storage which could be more than enough.
You keep saying you can't get more than 2GB/s on a cMP no matter how big your raid is.
Proof you can go over 2GB/s here:
http://barefeats.com/hard210.html
Then when i get proof you say everything is a bottleneck and the components are outdated... More hogwash...
10,666 MB/s is the theoretical maximum transfer rate for ddr3 PC3-10600 SDRAM according to basic computer science.
My Geekbench results shows it going to 9.5GB/s:
PCIe speed is 5GT/s per lane so running a x16 card (16 lanes) will have a maximum transfer rate of 8GB/s.
x8 would be 4GB/s
x4 would be 2GB/s
Are you sure you didn't put your card in the x4 slot?
To say that your limited by a 100MB/s network card is hogwash. You can buy 40Gb/s and 100Gb/s cards that will plug right into your pci x8 slot and give you up to 8GB/s transfer rate over a network with a compatible switch and router.
What was your setup? I've asked this before but you never replied. It would be good to know your card model and how many m.2's you had raided together and the what model the mac pro was.
Last edited:
Synthetic test and theoretical performance limits of individual components. This is what you do again and again. And you even started a Geekbench thread a year after everyone gave up on it.
We have moved on to real world performance gains. Stop looking at all these insane benchmark sets ups and synthetic benchmarks. They just make people consumers feel insecure if they don't have the latest, greatest, fastest 'thing'.
The absolute best thing you could do for Adobe video performance is move to Windows, even on your Mac.
The second best thing you could do is make your own work more efficient. There is always room for more personal efficiency.
The third thing you can do to save your time and speed up your work is stop going on the Internet to insult professionals with all this silly multi gigabytes per second nonsense.
Ugh your so ignorant. You think computer science math isn't real? You keep tossing this pseudoscience not backed by anything i can find.
You didn't even answer any of my questions or look into my post. You replied within about a couple minutes of posting and just tossed back deconstructive garbage as always with no facts. I think you probably just glanced and started ********ing again.
It seems to me you put your pcie m.2 controller in a x4 slot or didn't have a large enough raid.
Last edited:
Thanks for this laughable answer that has NOTHING to do with the apps you are trying to accelerate. If all the pros in the world who are editing the biggest movies, ads and music videos come here to read your gigabytes per second posts they will be laughing all over the place. Because they're producing the best stuff in the world (without these silly upgrades and theoretical speeds) while you go on the internet to argue about upgrades that the software can't take advantage of.
I already explained if you actually read my post. After effects is 64 bit multithreaded software capable of operating at full capacity. If your loading say a 1TB raw video file from a 4TB Raid going at 8GB/s it will make a considerable improvement in processing that data vs a sata 3 raid running at 750MB/s. So what's your ******** going to be now?
Please show us how you're going to build your 4TB 8GB/s raid in a cMP. Even you did that using Samsung UBX drives those chips would throttle under sustained load (if you didn't know about this issue you are out of this tech debate....out)
I have said 20 times to you that indeed read speeds can improve project load times, but once a project has loaded you are encumbered by one major thing: your temporary cache files and background rendering.
That's true one major thing that slows editors down more than "project loading".
Your own tests that you published showed the background rendering was writing those temp files at 25MB/s. Can you demonstrate to the readership that you need 5GB/s read write speeds when nearly all the pros in the world don't have these upgrades?
That's my last word to you. Blocked.
Just a quick recap of the things you said since i started posting:
I said Sata 3 is outdated..
Your Response:
Sata 3 is still practical for any enterprise workstation as anything more won't make a difference.
My response:
Sata 3 has a max transfer rate of 750MB/s. M.2 has a max transfer rate of 8GB/s in a PCI 2.0 x16 slot. It will make a noticable difference.
Your response:
You can't even use 750MB/s in After Effects.
My response:
I post benchmarks showing it does utilize 750MB/s Read 25MB/s Write and and can utilize more. I point out my AE is still loading the footage for processing but not writing an output file yet hence the 25MB/s write.
Your Response.
Haha loser 25MB/s write. M.2 can't go over 2GB/s on a cMP.
My Response:
http://barefeats.com/hard210.html
Your Response:
Haha loser 25MB/s write. You should use PC. 100MB/s Ethernet adapters are the best out there.
My Response:
I explain everything to him again like a child and inform him i can't work on a PC even though he says newer tech is useless. and there are 40Gb/s and 100Gb/s ethernet adapters.
Your Response:
Haha loser 25MB/s write.
Ugh! Will you guys please take your Urinary Olympiad to the parking lot so we can maintain civility in this thread?! (Where's MVC when we need him?!!)
Last edited:
Zaithe,
I think Soy is invoking the weak link in a chain argument while you seem to focus on one element in the chain.
Why don't you run whatever tasks you do through you slower hardware and compare with your new PCIe hardware?
I think Soy is invoking the weak link in a chain argument while you seem to focus on one element in the chain.
Why don't you run whatever tasks you do through you slower hardware and compare with your new PCIe hardware?
For Windows, may be yes, install a card and then you are good to go. But I doubt if there is any 100Gb/s network card that can be used in OSX without driver issue and with reasonable price.
I'm still waiting for my 2x m.2 drives to come in the mail. Won't get the other 2 until 4 months.
Main thing that stood out to me is that he said the cMP couldn't handle over 2GB/s transfer rates using m.2 raid 0. Saying that applications wont really utilize the 750MB/s available in Sata 3.. Not true..
Just left and right he kept finding something to argue with me about and i would constantly have to correct the disinformation he was spreading.
Looks like you can get a 40Gb/s adapter for mac it just will cost you about $1000+
https://www.attotech.com/products/t...underlink/thunderbolt3-to-40gbe/TLNQ-3402-D00
But there are cheaper 10Gb/s options. I just wanted to show that him Saying you were stuck with 100MB/s is incorrect.
Some people don't see the truth, even if there hiting her face. Barefeats show that over 5 GB/s are possible with the cMP 2010
Zaithe don't waste you time with SoyCapitanSoyCapitan.
But one simple question to Soy: WHY do you think 2 GB/s is the max.?
Zaithe don't waste you time with SoyCapitanSoyCapitan.
But one simple question to Soy: WHY do you think 2 GB/s is the max.?
He said the ram's maximum transfer rate is 2 GB/s when DDR3 1333mhz dram has a maximum theoretical output of 10,666 MB/s.
He said none of that is true that the only real ram benchmarks is creating a ram drive and running a disk transfer speed test on it.... Supposedly it's better than ram transfer speed benchmarks on novabench, geekmark, diglloyd or any other benchmarking software out there according to him.
Don't get me wrong Soy is a smart guy but he seems to throw around stuff only backed by himself and is not open to criticism.
In fact, it's a question being in my head for a long time. Why the RAM drive is so slow? Obviously, the 4x SM951 RAID 0 can beat the RAM drive. Is it mean that the SSD RAID array is faster than the RAM? Or it's simply because they are designed for different functions. So, for storage, SSD RAID 0 can do better. But to work with CPU as temporary memory, the RAM still way faster then SSD RAID?
I found that in real world. There is almost no benefit by using the RAM drive to make big ZIP file, or running games from RAM drive, etc. I suppose the loading time can be greatly improved, but the result is not.
I found that in real world. There is almost no benefit by using the RAM drive to make big ZIP file, or running games from RAM drive, etc. I suppose the loading time can be greatly improved, but the result is not.
unRARing files benefits greatly from fast drives. The difference is immediately noticeable between a regular SSD and a blade type SSD.
Since I don't have PCIe SSD. Do you mind to make a quick test about on unRAR PCIe SSD vs RAM drive?
I really want to know if there is any Real noticible benefit to run a RAM drive if the system already has a good SSD. Thanks in advance.