It’s:
Slot 0: 4GB
Slot 1: 2
Slot 2: 2
Slot 3: (empty)
Slot 4: 2
Slot 5: (empty)
Guess I could buy 2 more 2GB DIMMs…
Might not be a bad idea, that would eliminate one possible source of problems and make sure your system is firing on 3 channels.
I investigated getting an i7 system and I am pretty sure that the memory should be in pairs for dual channel, hence dual channel kits but having 2x2 Gb plus another 2 Gb with one 4 Gb does’nt add up to any dual or tri channel combination though it does add up to 10 Gb.
You could try just the 3x 2Gb as tri channel, but I would firstly check I am right and then ask why HP configured it like this, it seems strange to use such a processor with this configuration.
That said, though it may give you more headroom I dont think this is the reason you have 80% load on one core and 30% on the other eleven and my feeling is that spikes on that core are causing the problems.
Running the steinberg power scheme turns off core parking apparently and that may help a little if you have not already done it — be aware that the warning about heat is real with C6,mine goes up 11 c really quickly, but with C5 there is only a very slight increase, certainly in C6 there is an improvement in latency but yet again I would be surprised if it made the load even across all cores.
It is very easy to check out the ram issue. Just remove the 4gb module and run the 3X2gb in tri channel (check wich slots they should be in). If that doesn´t change anything you can disregard this from now on.
that ram config is completely wrong, I’m surprised it’s running very well at all like that.
slot 0
slot 2
slot 5
should be populated if you have 3 sticks, otherwise populate them all as long as they’re in set’s of 3 i.e 3x1 gig 3x2 gig or 3x4 gig in the corresponding places.
as mentioned above get rid of the 4 gig stick for now and fill the 0,2,5, slots with your 2 gig sticks.
MC
p.s. I’m selling my matching 12 gig 6x2 DD3 tri channel ram, if you’re interested P.M me
What happens with your CPU load if you open the project and switch some of the instruments off and back on again? Is the imbalance lessened or shifted to another core?
When I tried the DAW bench projects a while ago I was quite surprised to see the variance in numbers of plugins I could run before overload just by restarting them.
Even if you checked the DPC with DCPlat I’d recommend you to try LatencyMon in case you haven’t. I also got very fine numbers in DCPlat despite having big dropout problems but after running LatencyMon I could pinpoint a problem with a kernel mode driver showing an odd behaviour. The latency shown was still low, but with a distinct pattern. And sure enough, reverting the driver (graphics card in my case) put everything back to normal again.
/A
So the RAM needs to be matching in every other slot?
Yes and slots are normally color coded.
Blue and white x3 here
http://www.gigabyte.com/fileupload/product/2/3571/3385.jpg
I would still want an explanation from HP though.
Try using one instance of Omnisphere for each instrument. Also split Trillian into two instances.
One thing I did was turn off the multicore balancing in Kontakt (which is the bulk of my instruments in this project) and the problem did subside–at 384 samples latency I hit no overloads and core 1 was somewhat less loaded when Kontakt was set to balance to 6 cores.
This is a very heavily-loaded project. If people are able to get i7 920s down to 64 samples, though, I should be doing better.
Update:
I borrowed a colleague’s Steinberg MR816csx and did an A/B test with the ProFire 2626.
I set them both to a latency of 512 samples and booted Cubase 6 with my “heavily loaded” project and hit play. After reading (http://forum.recordingreview.com/content/yamaha-steinberg-mr816-csx-review-67/) how the MR816 performed superbly at super-low latencies and used very little CPU, I was expecting my ASIO load to drop like a rock when I tried the MR816.
Not true. The ASIO load was actually worse by about 20% with the MR816 than with the ProFire 2626.
So with that test behind me and still seeing ASIO overloads with my top-of-the-line i7 980x CPU running at about 25%, I am on to the RAM issue. Since my RAM is installed as 4GB and 3X2GB DIMMs, I ordered 2x4GB DIMMs so that I’ll have 3 of each. I know I could have just pulled the 4GB DIMM, but it was only $100 for the additional 8GB. I will report back on this over the weekend.
I do have one very curious thing to report: turning off hyperthreading reduces the ASIO load. I have heard others say that hypertheading has improved their performance. When I ran the DAWBench test, it was pretty much a wash (hyperthreading on/off in the BIOS) so I left it on. But with my “heavily loaded” project it definitely helped stabilize spikes and also reduce overall load to turn hyperthreading off.
One further update.
Installed a proper tri-channel RAM configuration for a total of 18GB:
0: 4
1: 2
2: 4
3: 2
4: 4
5: 2
I did not notice any appreciable change.
So it seems that the change that really had the biggest positive impact was turning off hyperthreading, which I discovered, is recommended by Steinberg: Windows+C6: does Steinberg recommend hyperthreading? - Cubase - Steinberg Forums
Hi lights
I have not got cubase 6 so maybe admin can change post to guest
just want to say Omni and trill and rmx are pretty big heavy waits and would say be careful because Omni can bring your system to the ground,
there are some patches in Omni if you do not edit they can make your asio screem for help it can be quite easy to forget what we are using when we have powerful systems…
novaburst
keep up the great music
Actually, while I’ve heard people say that I find it not to be the case. ‘Cold’ heavily used all 3 Spectrasonics instruments and barely rated a blip on my ASIO meter. Kontakt seemed to be far worse on the load than Omni, Trillian, or RMX. I can tell by how much benefit I get when I freeze each instrument.
I will continue to experiment but it does seem like Hyperthreading should be avoided.
Well guys, since I’ve had hyperthreading turned off, Cubase has performed like a champ. I am finally “done” (are you ever done in the digital world) with this bear of a project (to be released tonight I hope… rendering now) and I have unfrozen every channel, piled loads more VSTs of every kind onto it and I just can’t get the ASIO meter to even come close to the red. I must have at least 12 instances of Kontakt, some with multilayer, multi-gigabyte instruments, Omnisphere, Stylus, Trillian, and Groove Agent One, 2-4 vocal tracks, some routed to additional busses for more effects, and about 8 audio tracks for percussion.
I want to caveat by saying I am running this project at 512 samples latency. However, this is an insanely complex project and, to be fair, i haven’t actually tried to go lower for this one.
as i said great work, do you use wavelabs to master your work and convert it to mp3 or is it strait from cubase to the player then converted
keep it up
The site I use, bandcamp, accepts only WAV, so I send the 44.1/32bit WAV to them and they automatically generate all of the variants they need to stream and enable download. That way customers can choose whatever format they want–up to lossless.
I also put all of my “mastering plugs” (e.g., limiter, EQ, etc.) on the master channel in Cubase. So aside from the final fade (which I did in Sound Forge), the piece you hear is pretty much exactly as it sounds when I hit play in Cubase.
I tend to mix INTO the mastering plugs, so I always have the limiter, etc. active. I find trying to do it any other way results in me wanting to remix after I turn on the limiter.