@ X79 - 2.30.1900 - RAID w/ Single Disk Stalls + Quad Bandwidth
Posted 08 April 2012 - 04:44 AM
I originally came to ask about X79 quad channel bandwidth, and if that would ever be updated to reflect the quad channel performance. But, before posting here tonight I installed the latest final to check if that had been corrected before I made a thread about it, and I've just ran into another issue after updating to the latest final so I better include that as well.
I do see you've been working on SMART data detection on X79, so that may have something to do with this, the issue is now with a single disk in RAID mode it stalls out during the initial loading at "Scanning RAID devices". This did not happen in the previous version I was using, the older 2.20.1800 version. It's also takes many attempts to kill the process in task manager when this happens.
So I think you need to test with single drives using in RAID mode on X79. Note, I do not have ANY RAID arrays built, only single drives in RAID mode. However, if I reboot and connect a secondary drive before windows loads, or hot-plug one after windows has loaded (Or AIDA64 is stalled), then it only stalls for a short time - maybe 30-45 seconds and eventually loads. I have waited up to five minutes for it with only single drives connected, it never gets past the scanning for RAID arrays. This is with the 184.108.40.20611 Enterprise drivers, if that matters. I have not tested if the same occurs with 11.5 or earlier drivers, but I can if you need me to just let me know!
Now, for my memory issue/request. On X79, using dual channel 128-bit the Cache and Memory Benchmark results are the exact same as using quad channel 256-bit, these results should not be the same or even similar, they should reflect the 256-bit quad channel mode and they do not. Other programs such as SiSandra reflect the memory bandwidth differences as shown below.
So, can this please be corrected in a future update? Surely I can't be the first to notice or ask about this right? It would be great if this could be fixed quickly as I'm doing an X79 review right now, but I know it's probably something you can't just fix overnight or you would have already.
If you need any CPU-z or R/W-Everything dumps, ect. from me just let me know - Thanks!
GA-X79-UD3 - BIOS F10 / F10c
3x C300 SSD's
16GB Mushkin Memory
The RAID thing isn't a huge issue for me right now, but it should be corrected of course. But it would be really awesome if you could get the Cache and Memory Benchmark updated soon, thanks!!
Thanks for all your hard work, love this app!
Posted 08 April 2012 - 05:36 AM
But, there is some issue there because I did reboot a few times previously and it wasn't helping, and I also noted that "Force" making it load by hot pluging a drive was causing other issues (Windows would stall rebooting/shutting down, and detection of newly connected USB Drives post hot plug and AIDA64 load would fail)
So, not sure what's going on, but it seems to have corrected itself though? It's not my system itself, I know that, I review hardware and it's not my first day at any of this, so it wasn't just some instability or anything like that (Just letting you know)
I take that back, another few reboots and it's stalling again on RAID, so something is up with it when single disks are used in RAID mode for sure!
Posted 08 April 2012 - 07:20 AM
BTW, since at Intel RAID scanning we're doing everything "by the book", and full SMART details are unavailable even when it loads up fine, all we can hope for is that Intel will soon release a proper, fixed RST Enterprise driver.
BTW #2, have you tried to use the latest Intel RST Enterprise driver 220.127.116.1116 WHQL? Maybe it would help about the slowdown.
As for the quad-channel memory bandwidth results, they are actually accurate. The single-threaded memory bandwidth performance -- that AIDA64 currently measures -- doesn't scale when going from 2-channel to 4-channel configuration on X79 systems. We're already working on multi-threaded memory bandwidth benchmarks: they will be able to properly measure the performance improvement due to 3+ memory channels.
Posted 08 April 2012 - 02:18 PM
So if it's possible, please use the latest RSTe WHQL driver of 18.104.22.16816. Hopefully it fixes the issues on your system. If not, then you can still disable the RAID member enumeration in the AIDA64 Preferences
Posted 09 April 2012 - 04:24 AM
As for the memory, why does it scale with triple channel then on X58? I understand it's a single threaded benchmark, but how does that apply to xx memory modules in use? Just wondering to try and understand, since it seemed to scale fine with single, dual, then triple on X58 - or at least I think I remember it working properly when comparing like that on X58 (Maybe I'm wrong haha!)
Anyway, thank you again for the quick reply!! I'll look into the RAID issue with your tips, and I'm really looking forward to the new multi-threaded memory benchmark!!
Keep up the great work, your program is the best!!
Tested RAID member enumeration disabled and that instantly fixed the issue! I will update the drivers as I should anyway, then I will re-enable this and see if it still happens or not. I'll let you know.
Updated drivers to 22.214.171.12416, and re-enabled the above setting and everything works quickly as expected!! So thanks for the heads up, it seems the old drivers were causing issues with AIDA64 RAID member detection when no actual arrays were built, all working perfect now - thanks!
Can you give a guesstimate on when the multi-memory benchmark version might be ready for testing? I could beta test for you if you'd like, I'm friends with Stasio - so maybe he already has posted a beta on this and I didn't notice yet?
Posted 09 April 2012 - 07:16 AM
Core i7-3960X with Quad Channel DDR3-1600 and X79 chipset has a theoretical memory bandwidth of 51200 MB/sec. AIDA64 measures 16788 MB/sec memory read bandwidth (cca. 33% of theoretical bandwidth).
So as you can see, with only a single thread it's not possible to get close to even half the theoretical memory bandwidth of such platforms. That explains why in single-threaded applications (like most 3D games) it doesn't matter if you use more than 2 channels for your memory While in heavily multi-threaded applications multi-channel memory configurations look much better.
We don't yet have a beta of the multi-threaded memory benchmarks. We expect to roll the new memory benchmarks out in the 2nd half of 2012. It will be a major change, since all memory scores will change in the switch, so the new benchmarks will debut with a major AIDA64 version update -- probably v3.00. BTW, in the first half of 2012 we'll update other benchmarks like CPU PhotoWorxx to implement SSE and AVX optimizations.
Thank you for your feedbacks on the Intel RSTe driver issue.
Posted 22 April 2012 - 04:26 AM
CPU:Intel i7 920 @ 3.8(D0), Mainboard:Asus Rampage II Gene, Memory:12GB Corsair Vengeance 1600
Video:EVGA Geforce GTX 680+ 4GB, Sound:Creative XFI Titanium Fatal1ty Pro, Monitor:BenQ G2400WD
HDD:500GB Spinpoint F3, 1TB WD Black, 2TB WD Red, 1TB WD Black
Case:NZXT Guardian 921RB, PSU:Corsair 620HX, OS:Windows 7 SP1
0 user(s) are reading this topic
0 members, 0 guests, 0 anonymous users