Nope.
The 2 tests that failed are free ones developed a long time ago.
The 2 that passed are from Apple and a paid for one that was revised recently.
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.
MP 7,1 Which LRDIMM 2933 64GB RAM for 2019 Mac Pro - Samsung vs Micron vs Hynix?
- Thread starter paulchiu
- Start date
- Sort by reaction score
Nope.
The 2 tests that failed are free ones developed a long time ago.
The 2 that passed are from Apple and a paid for one that was revised recently.
Yeah , that test did not pass . It seemed to freeze up .
That's a nice field display . Does Marshall make a DP version ?
Froze after 2:45 hours is all we got. No error logs option. I’m not too worried over it as 2 long compressive memory tests using TechTool Pro 11 and 12 passed without errors. The only caveat is that TechTool Pro could only allocate 736 and 741 each time of the 768gb
I think the short but official Apple hardware tests gave me confidence more. Strange, but I’ll leave it at that.
Yes. Not sure the model number but there certainly is. This display is near 10 years old. Still works, but only 1080p, so I could not select 1 pass for the overnight test on memtest86.
when I started video work a decade plus ago, this marshal with canon 5D‘s was the go for 1080p capture.
On occasion , I have lost the channel if I were to turn off , unplug or place the monitor in sleep mode during one of these extensive tests . It's not fun to lose the results of a test like this . So , what I do is connect them to high quality , but old , displays . If the screen gets a burn , that's the cost of the operation , sadly .
I have four ancient 23" Eizo fluorescents with DP , DVI and VGA ports for my benches at my shop . That takes care most of my needs . If I need a HDMI port , I'll risk one of my nice 27" 4K Dells . My MP7,1 has the 580X card that only has two HDMI ports . I'm not overjoyed about that .
[automerge]1580151990[/automerge]
If I were you , I'd still find a way of getting memtest to pass , at least with one pass with the entire memory configuration . Maybe a larger HDMI screen so you can access all the user preference choices ? Don't risk your XDR display . This test should not have frozen , the first time you tried . I have had very good luck running memtest as a bootable device . My only complaint is that it can take forever and you can only perform just this test on the System at the time . Running memtest in an OS session just isn't very thorough ( although it can be useful when you run other tests concurrently for a System wide stress test ) .
Last edited:
I am with you on your last point. It is not satisfying to not have that Memtest86 test complete. I think it may have been the stick as it is an old freebie 4GB USB2 throwaway.
I have the AMD Radeon Vega II with 32GB and not happy it did not support both the Marshall at 1080P and Pro Display XDR connected to Thunderbolt 3 at the same time. It's either one or the other.
Once, you disconnect the Pro Display XDR during USB-boot run, the process halts.
I agree that running one pass and seeing the good result is reassuring. In my case with 768GB, I think even with the USB-boot, not all of the RAM will be tested. Definitely more than the 741GB TechTool Pro 12 tested, but not all 768GB.
Will be going to the closet for an old 4K display later, if this stock market recovers.....
OK, tried running Memtest86 on a stick again and finished test 1 to test 11 fine but test 12 refused to even start.
Rebooted and selected just test 12 (Random 128 bits) and it stopped at 1 seconds
again no errors
I searched the web and others have experienced this with older Zeon processors.
Which processor do you have? It can't be any different than mine.
My lovely MP7,1 has had three different processors in her short life already . Boy , kids do grow up quick these days .
She started out with an 8 core W-3223 from the factory , went to a 24 Core Gold 6212U I had dropped shipped from Intel until the NVRAM reset killed her and now is gracefully aging with an used 28 Core W-3275M shipped from China that I installed myself .
My only complaint is memtest takes a long time even with my humble 32GB of memory . I managed to reduce the amount of time it takes by building a custom USB NVMe M.2 SSD thumb drive loaded up with a bootable memtest86 , but it still takes hours . The memory always passes . It's factory installed ram .
Yes, I did not go your route and paid a royal premium for my W3275 28 core from Apple. I calculate probably 1800 more than necessary. But I figure as a long time AAPL owner, I should set a "good" example. LOL
No, I am just lazy and don't want to expanse the extra effort to save on the CPU plus scared of screwimng things up on the install.
I digress
running the last test 13 (hammer) alone and after 46 minutes, still at only 2% done.
I really need to use my Mac for work tonight and not have to wait until 3AM
I see that you bought the M386A8K40CM2-CVFC0 revision, is there any difference between B0/C0/G0 or the one w/o the suffix?
I asked because the Samsung DDR4 Guide only mentioned it as buffer but no other info and the one w/o the suffix (M386A8K40CM2-CVF) has a higher price tag than the ones that do on Memory4Less ($314 vs $299).
I ordered M386A8K40CM2-CVF without the suffix at memory.net
Did not look into the difference between the suffices.
I searched just now for M386A8K40CM2-CVFC0 at memory4less and found this:
https://www.memory4less.com/samsung-server-memory-m386a8k40cm2-cvfc0 at $528????
This is weird.
Could not find m386a8k40cm2-cvfc0 at memory.net now.
Sorry could not help you more.
I see. Seems like they just hiked the price for some reason, I made the order yesterday and it was $299.43.
Care to explain why a 56KB executable that don't write anything to the media it boots from will run faster with a NVMe thumb drive? Even if you are using the memtest86.com fork that have HTML reports, I don't get it why would be faster.
NVMe is a significantly faster drive , both with IOPS and sequentials . So therein must lie the answer .
Come to think of it , I did use different interfaces with the two drives . But I don't think it would matter .
For the cheap USB thumb drive I used the type A USB port on the I/O Card .
For the USB-C NVMe drive , I used the TB3 port on the top of the Mac . It's a convenient place to rest a larger and heavier drive .
Maybe next time I run the test ,I'll connect the NVMe drive to the type A port on the I/O card to see if performance is affected . I have both type A and Type C 10 Gbps adapters for the drive , after all .
Still don't understand… it's a 56KB executable that is run from memory and can be loaded even from CDROMs or from the EFI. NVMe will do absolutely nothing for a RAM test.
Well , that's reality . The test both boots and runs faster with NVMe . So there's some part of the test that takes advantage of a faster drive . Here are the screen grabs documenting the difference . The test that took 10 hours is with the cheap USB thumb drive . The test that took 4 hours is with the NVMe drive . Ceteris paribus . Only the drive is different . Have you tried different drive types and recorded the elapsed time to completion of a standardized memtest ?
Memtest86+ is made to be run from anything, it's a very tiny executable (source code is a 215KB tar file) that is loaded once to memory and never access the disk again. Some PC motherboards even have it from the firmware, it can be loaded from an UEFI DXE, from PXE and most Linux distributions have it as a GRUB menu item. I usually load it from CDROMs or very tiny USB keys. Older versions can be load from a floppy disk.
PassMark Memtest86.com fork can write a HTML report at the end of the test, even to a server, but nothing is run from the disk, there aren't tests to be read from the disk, like the original version that is open source, it just write/read/compare different patterns to or from memory.
Your time discrepancy between tests has other motives. You can check the code yourself, it's here: Memtest86+ V5.01 source code (.tar.gz)
Last edited:
I'm a hardware technician , so I don't understand software very well . Code is all Greek to me ... The same software package was also used to create both test drives . The only variable is the drive type . What I really care about is that the NVMe drive is quicker and there is a pragmatic reason to use it with all my system Builds .
Another quick question, if I may: how long did it take for memory.net to ship your order?
I end up cancel the order from M4L as it seems pretty shady and got lots of bad rep on the Internet.