Xeon X5680s Now $45... October 2018

[danworks]

Finished my install … My speed more than doubled running Geek Bench

I'm very glad for you! This make me feel better

[James Steele]

Thank HC Markus. He's the one who shared this with the forum to start with.

[danworks]

Aaaarrrrggghhhhh

I'm still "touching iron" (knock on wood in US) and I will for a quite while but I've turned on the MacPro while crossing fingers, toes, guts and it did what it usually does …… showed me the desktop

Thanks to James that suggested me this upgrade and thanks to HCMarkus who opened this thread.

FYI, I've payed the Xeon £350 + £21 UPS shipping, which in American currency is more or less $540.

As soon as possible I'll post a real work environment to see how it behave with DP & friends.

ciao

[Gravity Jim]

The upgrade still has me giddy. Speed, response, and CPU power are all radically improved. Look-ahead plugs like Izotope Nectar Breath Control run close to latency free, real time effects and VIs exhibit better timing, USB MIDI input latency is gone. This is the best ever.

[toodamnhip]

Thank HC Markus. He's the one who shared this with the forum to start with.

Of course! Thanks HC

[danworks]

Considering that:

1- my machine was already a 12 cores - 2.4Ghz
2- numbers are not opinions
3- multi processing on DAW isn’t handled as on video apps

lets talk about general impressions and “real work environments”.

General impressions are pretty good, For example, I’ve recorded some screen movie with SnapzPro and the response was noticeably fast. Basically most of the operation I perform have a much faster reaction. That explains pretty well the geek bench difference fm my original 2.4Ghz machine on MULTI CORES (22249), and the hacked 3.3Ghz version (30237). We’re talking about almost 50% more power, which is a lot.

On real work environment things are a little different. As I said, as far as I know DAW do not use all the cores as video and graphic apps do. Bidule is a good example of how multi cores isn't really handled and DP is too; therefore even if the the faster feeling is there, is less noticeable.

Geek bench of the original 12cores SINGLE CORE - 2.4Ghz was 2124 while geek bench of the hacked 12cores SINGLE CORE - 3.3Ghz is 2680 That’s more or less 30% more and not almost 50% as in the MULTI CORES mode.

Here a couple of lil video I shoot before and after the transplant on the same project. A good way to check them out is to download them, put it close to each other and possibly let them start together.

First video is DP rewired with a quite large Bidule orchestral session, playing a 12 MIDI track of Synful Kontakt MachFive instruments.

Original 12cores
http://transfer.imusicproductions.com/m ... coresA.mov
Hacked 12cores
http://transfer.imusicproductions.com/m ... Hack_A.mov

The second one is just drum and bass playing Kontakt on VRack.

Original 12cores
http://transfer.imusicproductions.com/m ... coresB.mov
Hacked 12cores
http://transfer.imusicproductions.com/m ... Hack_B.mov

I don't want to cool down the enthusiasm, I'm so happy I did this upgrade and I'll never stop to thank you guys, but these are the numbers and this is the behavior within DP with these two lil examples.

ciao

[Gravity Jim]

You "hacked" (a better term would be "upgraded") an existing 12-core just for more clock speed? Well, I don't have time to watch the videos, but that was likely a waste of time and money. The enthusiasm you see in this thread is based on going from 8 cores to 12 ( or from 4 to 6). Performance gains from 12 CPU cores at 2.4 to 12 cores at 3.3 are probably negligible.

For me, the whole point of this exercise wasn't speed: it was to increase the number of CPU cores and therefore the number of VI instances I could comfortably run in a given project. If my machine was already a 12-core I wouldn't have bothered... while they're a decent indicator I didn't want Geekbench scores, I wanted more room for Kontakt and Omnisphere.

[HCMarkus]

It has truly been a pleasure assisting my fellow composers and recordists in putting together very powerful, yet cost-effective systems.

Danworks, I am wondering, are you are using multiple instantiations of your VIs in your demonstration videos?

I find Apple's Activity Monitor provides are more interesting view of CPU activity than the DP meter, as it shows per-core usage. When running DP on my Hex MP and observing Apple's Activity Monitor, I note that activity is shared across all physical cores fairly equally, with the virtual cores showing a reduced workload.

Because you were previously running a 12 core system, I would expect the increase in clock speed to translate directly to better performance, as any multi-core inefficiencies would be identical regardless of clock speed. The again, expectations based on theory aren't always fulfilled in the real world, are they?

[monkey man]

I bought the 2012 2.4 12 core machine to replace a 2008 2.8 8 core with a future CPU upgrade in mind once the chips' prices fell. I didn't know at the time that one would be able to install the X5680 CPUs in, say, a 2010 machine or whatever. If I had known this, I'd likely have bought a much cheaper S/H unit.

Still, I imagine there'd be some benefit/s in having the newer motherboard and architecture, possibly faster RAM and so on.

Jim, I've been disappointed with the "every-day-task" zippiness of the newer 2.4 12 core machine when compared to my previous 2008 2.8 8 core. I figured this must've been due to the fact that the "straight-line" speed (as opposed to luggin' a load up a hill) is probably more or less the same (slightly slower GHz but probably more efficient?). The switch to 3.3GHz or whatever should (hopefully!) recoup some of what I'd hoped would be a little extra snappiness for every-day tasks. If anything, the "new" 2.4 machine has been a little more sluggish than the ol' 2008 model; this has been frustrating and I'm very much counting on the upgrade's making all the difference. FWIW I simply swapped the drives over, as I always do, and have been using the same boot partition all along. Sure, a complete re-installation might well restore a sniff of snappiness, but I've been frugal with installations and the instant drop in performance after having moved to the 2012 model was unmistakable.

One other thing - wouldn't the raw speed increase of, say, 50% still translate to one's being able to run smaller buffers, increased snappiness within DP (the Holy Grail) or a proportionately-larger number of plug-ins / VIs in real-time? Am I missing something?

I'm not expecting miracles, even though when viewed, as it should be, through the lens of the upgrade from the 2008 model's completion, the performance bump should still be significant. This was always a long-term plan, and thanks to the Epic Fail that my MOTU AVB exploration has been, it's one that, to quote Darth Vader (eminently appropriate!), "is finally complete". Or something like that...

Oh heck, I'm getting ahead of myself; the chips and Arctic Silver are due to arrive in about 8 days...

[monkey man]

It has truly been a pleasure assisting my fellow composers and recordists in putting together very powerful, yet cost-effective systems.

Danworks, I am wondering, are you are using multiple instantiations of your VIs in your demonstration videos?

I find Apple's Activity Monitor provides are more interesting view of CPU activity than the DP meter, as it shows per-core usage. When running DP on my Hex MP and observing Apple's Activity Monitor, I note that activity is shared across all physical cores fairly equally, with the virtual cores showing a reduced workload.

Because you were previously running a 12 core system, I would expect the increase in clock speed to translate directly to better performance, as any multi-core inefficiencies would be identical regardless of clock speed. The again, expectations based on theory aren't always fulfilled in the real world, are they?

Geez, my post took so long to complete (internet connection kept dropping out!) I missed your sneaky lil' ol' entry, Sir Markus.

You kinda answered my question and (kinda) agreed with me... in theory at least! I too pondered the multi-thread / instantiation thing; my crappy connection tonight meant I couldn't watch the videos, but my first thought (after "oh crap, no!") in response to danworks' comments was a query as to whether his setup was configured for multi-out instantiations or multiple single-instrument ones. He may need to see that thread where MD 'xplaind it all. You know the one - the one I don't have a link for.

Again, thank you again, Sir Markus, in redundancy, advance and retrospect. Obviously I'll keep you informed even 'though you already know what the outcome will be. Yours in authentic redundancy... or is that redundant authenticity?...
Nicky

[James Steele]

For some reason, I thought danworks was using an 8 core machine. I don't think I'd have bothered to do the upgrade on a 12 core just for the clock speed boost. For me, going from my MacPro4,1 4 core to a 6 core was entirely worth the $200 expense.

[danworks]

@HCMArkus:

Multi instantiation. I know that on both Kontakt and MachFive multi instantiation in much more efficient that throwing a lot of instruments on one single instantiation. I agree with you that DP meter is very bad but that was a simple rough test just to have an idea before and after, and I've definitely got mine.

About "expectation", well I think I had just what I was expecting and I'm very happy with it, and I thank you again for having shared your know-how.

For me 50% more power on multi core mode and 30% on single core mode for a ridiculous amount of money and one hour of time, versus the cost of a new computer + new interfaces + ext boxes 4 SSD + re-installing everything fm scratch + whatever else happen, seems to me an amazing achievement and not a waste of money.

I've dig the subject of changing computer very carefully, spoke with many colleague, friends and software developer that have and use the new MacPro, and at the end of the day I've got the very same answer.

Apparently "to make music" it's much better to have a faster CPU than a larger number of cores. A couple of these guys with the MacPro 8 core 3.0Ghz, told me that they are incredibly happy with the results and with the facts that they don't have to be worried anymore about CPU strain. And one of them added:

"If I had to do it all over again, I might even just have gone for the 6-core and taken the 3.5GHz …"

Speed is very important, big time.

@Gravity Jim
It may be just a semantic matter but if you're not an official company tech and you open a computer, change the CPU with another one that isn't recommended by the company itself for that computer model, breaking all the rules and the warranty … technically it is not an upgrade

[Michael Canavan]

On real work environment things are a little different. As I said, as far as I know DAW do not use all the cores as video and graphic apps do. Bidule is a good example of how multi cores isn't really handled and DP is too; therefore even if the the faster feeling is there, is less noticeable.

That simply isn't true. DP handles multi cores pretty well actually. From tests I've run DP like almost all DAWs handles a single track on a single CPU though, so you can overload a 12 core machine by placing heavy plug ins all on the same track. If you evenly spread plug ins across dozens of tracks DP handles them 100000000X better than if you stack a few tracks. CPU is far more important for a single track, but multi core is in no way not "really handled" in DP, that's obvious if you have dozens of tracks. The latest iMac has the best single core performance, the i7 in it is great for single core operations, but it's also only a single four core.

[HCMarkus]

Redundancy expressed with grace can be a blessing Nick... its sorta' like backing up your files on another drive. Have a great day!

[danworks]

On real work environment things are a little different. As I said, as far as I know DAW do not use all the cores as video and graphic apps do.

That simply isn't true. DP handles multi cores pretty well actually.

Unless you don't work on project with 200 audio tracks and 600 plugins, plugs and audio tracks stop to be an issue ages ago. Many times I bring home average pop-rock project with a decent number of plugins on my 2009 MacBookPro with SSD, and I can work with it without any hiccups.

Virtual instruments are another story, and nowadays with the advent of the script for realistic legato slur portamento and other sophisticated programming, are "the problem". Spitfire Audio, Embertone, Kirk Hunter to name some, are great libraries but heavy on CPU.

DP with these VIs-libraries isn't the most efficient DAW out there, unfortunately I'm not the only one saying it, and for this reason a large number of composers that use orchestral setup environment, add to DP another host like Bidule (in rewire) or VSL to manage VIs. If DP was efficient enough with VIs, no one would even think to use two different software for a job that one single software could manage, right?

I respect your opinion and it would be great if it was true as you claim cos I could avoid a lot of headache … but I'm afraid it is not.

Lets see if something will change on DP9