Intel i5 / i7 / i9 DAW Performance Reports : [Archive]

TAFKAT

12-21-2008, 08:31 PM

Hey All,

The Intel i7 is the biggest architectural shift for Intel since the Pentium 1. Gone are the FSB , Northbridge - Outboard Memory Controller, and also the added arbitration across multiple dies , multiple L2 Caches , replaced by a Native Quadcore, Shared Cache Architecture , Onboard Memory Controller , Tri Channel DDR3, and a new QuickPath interconnect architecture.

This is a whole new ball game , for Intel at least.

Hyperthreading also makes a return.....

The approach is now very similar to AMD's approach, and of course we will get the AMDians shedding a few tears that Intel have copied them, and fair enough. The difference being that while AMD have been sitting on their hands, Intels new architecture has hit the ball out of the park again, and Hyperthreading II was the biggest surprise of all when I ran thru the benches..

O.K, here are some results for the last 3 generations of Intel chips, all @ 2.66 Ghz :

QX6700 / Q9450 / Dual X5355 / Dual X5430 / i7 920 HT Off / i7 920 HT On

http://www.dawbench.com/images/dawbench-core2 v i7-4.JPG

I have only shared these results with some DAW builders that I trust and some developers and tech journalists so far.

The results speak for themselves, the i7 has hit the ball out of the park again, and HT is actually working this time.

The low latency results are incredible, I can't wait for the Dual Socket versions of these things, and the single Octocore will be insane.

Keep in mind this is with Triple Channel DDR3 1333 , it will be better again with DDR3 1600.

Anyhow,

The BIOS's are still a little green, so I won't be marketing my i7 DAW solutions until early 2009 , but the dev system is proving pretty solid with the latest BIOS.

I will not share publicly the specific motherboard I have selected as the best of the bunch , so please don't be offended when ignored if a request is made , I do need to keep some cards close to my chest..

Peace

Vin Curigliano
AAVIM Technology

Animus

12-22-2008, 05:08 AM

damn that's pretty fraking impressive. Must not be tempted to build another computer! Resist the gearsluttery!

Jcschild

12-22-2008, 10:08 AM

Well since Vin took care of the effects side of things here is the samples side of things

http://www.adkproaudio.com/images/k5benchmarks.jpg
since i know it will get asked here is the project specs.

Kontakt samples playback
This test is more geared toward Composers and heavy samples users.
There has been alot of discussion about what is best for samples playback and if more cores are better.
Samples playback is very memory intensive, memory speed and bandwidth are effected by several things.
GHZ of the processor, Front side bus of the processor, memory type (DDR2/DDR3) and of course platform (chipset and processor architecture). Little or no improvement is found with Quad core vs dual core.
A faster GHZ dual core will out do a slower quad core for samples playback. DDR3 1600 is better than DDR2 800.
Ideally when considering an overall picture of a project a fast GHz quad is best as the Quad cores do have a great effect on effects usage, overall project playback and high track counts @ high rez.
Some thoughts: the AMD is a very affordable and well performing answer for a samples(slave) box or an overclocked dual core.
The Core i7 is outstanding as a main system, or the more afforable overclocked Core 2.

The test: (not downloadable but duplicatable)
all done with XP Pro SP3, Fireface 400, 4gig ram) (core i7 was 6gig ram, vista 64)
the samples set was over 2gig)
6 Instances of Kontakt3 in rack All using K3 included Vienna instruments Default dfd settings All running live, no freeze, no bounce MIDI tracks running to every patch 1 Violin ens14 (all) Viola ens10 (all) Cello ens8 (all) Bass ens6 (all) 2 Piccolo (all) Flute (all) Oboe (all) FrenchOboe (all) EnglHorn (all) Clarinet (all) 3 TrumpEns3 (all) FrnHornEns4 (all) Trombens3 (all) Tuba (all) 4 Violin ens14 (all) Viola ens10 (all) Cello ens8 (all) Bass ens6 (all) 5 Tenor Recorder (all) Piccolo Recorder (all) Low Whistle (all) 6 Timpani Hits.

Scott
ADK

TAFKAT

12-22-2008, 10:16 AM

Hey Scott,

Thanks for sharing those Mate.., some really interesting results that help cover the bases across the various aspects of DAW system / session loadings

BTW: Is that a ball sailing out of the park I see... LOL

Man, can't wait for the Dual Socket Xeon versions of these things, I have only seen some Server specific test results so far, but they were insane Here (http://www.anandtech.com/weblog/showpost.aspx?i=532)

http://images.anandtech.com/graphs/amdshanghai_112208120910/17881.png

I know they are not relevant to Audio specific stress's , but Man are they impressive.. :-)

Peace

V:

Animus

12-22-2008, 11:27 AM

Welcome to the forums Scott! Glad to see you here.

btw, I am still loving the hell out of the UAD2 Quad I got from you. Now where is Fatso? :o

TAFKAT

01-12-2009, 07:49 PM

Hey All,

I have been seeing various reports on other forums with some members who have recently built i7 systems complaining about various performance issues, some related to high DPC latency issues, and one guy even going so far as to claim that its a well known and inherent issue with current Intel chipsets ?

Say What ?

I only have one thing to say to that.., actually a picture is worth a thousand words..

http://dawbench.com/images/i7-dawbenchdsp-dpc.jpg

i7 / X58 / XPSP3 System running at 032 samples , RME HDSP , DAWbench DSP Test Session, 130 x 8 Band RXC Multiband Compressors , DPC latency at 13.

:D

psvennevig

01-12-2009, 08:57 PM

Sweet mama!

Jcschild

01-13-2009, 11:31 AM

yup thats what mine look like!:icon_eek3:

FYI INtel is sending me a Xeon i7 system!!!

TAFKAT

01-13-2009, 03:57 PM

Hey Scott,

It seems that secret Kentucky Straight Snake Oil we sprinkle on the systems is doing its job then... :D

Re the i7 Xeon system.., cooool, let me know when ya have it in house..

leggy

01-21-2009, 10:08 AM

Interesting review by Martin Walker of an i7 system built by a UK company called Scan, who supply computer components but also build complete systems, in the latest Sound On Sound. As we've seen in the benchmarks posted so far, properly built and setup these machines offer a massive leap in performance at low latencies. Vin and Dawbench get a namecheck too!

Rob

TAFKAT

01-21-2009, 10:39 AM

Hey Rob,

I was wondering when that review would hit..:D

I helped Martin flesh out some of the performance aspect of the review while he was finalising that part of the report, and supplied him with some data to be able to compare the system to not only another i7 system so he knew he was on the same playing filed, but also the earlier systems architecture..

You'll notice that all the numbers bar the review system are mine, and the review i7 systems results will be close to my i7 numbers.

The review hasn't hit the eSub yet, and my hard copy won't hit until next week , so I am looking forward to reading it..

Interestingly when I posted my i7 results at SOS forum, they barely caused a ripple, it will be interesting to see if the review wakes the membership up there.. :sleeping:

TAFKAT

02-27-2009, 08:48 PM

Time to focus on some more interesting stuff than the Steinberg Soap Opera..

The Dual Socket i7 Xeon platform , aka Gainstown , is getting closer to release.

More Info Here (http://www.theinquirer.net/inquirer/news/173/1051173/dual-core-nehalem-systems-pump-bandwidth)

http://www.theinquirer.net/img/11647/2tylers.jpg

Judging by the spec and some early reports, the arbitration issues that were needed to be navigated in the previous dual socket architecture are gone, so we should be seeing some incredible low latency scalability on these systems.

I am in the process of negotiating some Engineering Samples to try and get a bit of a head start on some development and qualification for DAW / video application. There are not many of these in the Wild outside of the bigger OEM's, but its looking like I may be able to corner some advance hardware shortly.

Stay Tuned..

leggy

03-01-2009, 01:10 AM

Good luck Vin! Be very interested to hear about this.

Rob

psvennevig

03-03-2009, 02:39 PM

TAFKAT

03-03-2009, 03:38 PM

Hey Pal,

Hopefully I can get some advance hardware to be able to compare some notes..

BTW: Just checked out the specs here in Oz for new Nehalem Macs..

They now also have a single Socket system as well with a Xeon derivative of the desktop i7 , Max 8GB of RAM, and starting at only $4495.00.., the Dual Sockets are starting at $5899.00

I can see the stampede now.. :wink:

LEX

03-03-2009, 03:56 PM

Sam

03-03-2009, 04:07 PM

Those are Aussie Prices right?
yep :)

TAFKAT

03-03-2009, 04:11 PM

Just did a conversion from the US store as well, they are about 10% higher when converted..

US is $2499.00, $3299.00 / AU conversion is $3900.00 / $5155.00

They maybe adding GST here on top.

That single Quad is a bargain... :rotfl:

Daryl

03-03-2009, 05:05 PM

Well, I just speced up a machine that will do what my current one does. Apart from the fact that there aren't enough hard drives (what a PITA) it came to £9.5K. What a bargain. Anyone who pays these prices either has too much money, or is an idiot....! :icon_eek3: :icon_lol:

D

leggy

03-03-2009, 06:21 PM

TAFKAT

03-03-2009, 06:24 PM

I just had a conversation with a close friend this morning who is a Macnoid of the highest order, and when I noted the single i7 system being so expensive, he says, "who cares, no real Mac user will buy a single socket system".. , oh of course I say, real Mac users only use Octocores, because they do scale so well on OSX..., and I'll make a note that single i7's would just be a waste of time for you .. :icon_rolleyes:

Daryl

03-03-2009, 06:33 PM

Well you can't really compare these dual CPU machines to yours. From the preliminary info we're seeing it looks possible that they will outperform any current machine by quite a margin. And of course nobody with any sense would buy their memory and hard drives from Apple - way too expensive Especially if you're going for 32gb :)

Buuuuuuuuuuuuut, if you don't buy everything from Apple, then it's not really a Mac. I mean, 3rd party bits and bobs can't possibly be as good as something that comes from Apple.

Seriously I wasn't trying to compare my machine to the new Macs, although my machine will out perform any Mac using OSX, simply by the fact that I can use a 64bit OS, whereas there isn't such a thing for OSX. Yet.......... :>o

However, the biggest problem for me is not the lack of power, it's the darned inefficiency of ASIO performance. Who knows whether or not this is improved much with the new chips.

D

TAFKAT

03-04-2009, 02:35 AM

Steinbergs Official Word on Intel i7 and Hyperthreading Here (http://knowledgebase.steinberg.de/71_1.html)

Dated 3rd March 2009 :

If you are running a single Pentium 4 CPU system, enabling Hyper-Threading results in an increase of processing power.

If you are running a Xeon HT multi-CPU system or a Pentium Extreme Edition dual-core system, you should disable HT in the BIOS.

The same applies to the new Intel i7 processors that also feature Hyper-Threading. We are currently investigating using an Intel i7 with enabled HT in conjunction with our sequencers. Until our tests are still ongoing, we recommend turning off HT on these systems as well.

Whats there to keep investigating.. ?

What can I say except as per usual they have proven they are so far off the ball , it is embarrassing.

That information is a as useless and misguided as their stance on multicore scaling being broken on 8 Cores systems across the board.

While Steinberg continue to flap around in the dark, those of us who have actually been developing and testing these systems in the Real World know that Hyperthreading on the i7 has absolutely no correlation to the previous Hyperthreading implemented on the Netburst architecture.

My advice is to ignore any and all advice from Steinberg in regards to anything to do with DAW scaling performance.

Michael Tibes

03-04-2009, 09:02 AM

I bet they're just trying to cover their asses until they checked things out - which might take a little while as we know :sleeping:

Michael

TAFKAT

03-04-2009, 09:24 AM

I bet they're just trying to cover their asses until they checked things out - which might take a little while as we know :sleeping:

Michael

Hey Michael,

I suppose my question would be is why they haven't already dotted the i's. Its clear that they do not even have an i7 system at their disposal, if they did have, there wouldn't be anything left to continue investigating.

The fact that they have simply tagged the i7's onto an article that is 3-5 years old , and has no correlation to the current architecture is ridiculous.

I know for a fact that Cakewalk / Digidesign development and QA had i7 engineering sample systems for many months before they were publicly available, so any fine detail was sorted before release. It seems the combined resources of Development, QA and BETA can't even manage to get some qualified data even months after release.

You have to wonder.

psvennevig

03-04-2009, 06:45 PM

I agree.

SB clearly can not have tested the i7s with HT on. It does work.
As Vin says, the new HT have nothing todo with the old Netburst HT.
They just kept the name (maybe Intel shouldn't have).

P

Michael Tibes

03-05-2009, 02:02 AM

Hey Michael,

I suppose my question would be is why they haven't already dotted the i's. Its clear that they do not even have an i7 system at their disposal, if they did have, there wouldn't be anything left to continue investigating.

The fact that they have simply tagged the i7's onto an article that is 3-5 years old , and has no correlation to the current architecture is ridiculous.

I know for a fact that Cakewalk / Digidesign development and QA had i7 engineering sample systems for many months before they were publicly available, so any fine detail was sorted before release. It seems the combined resources of Development, QA and BETA can't even manage to get some qualified data even months after release.

You have to wonder.

You have to wonder sums it all up. I can just guess that their testing procedure is so complicated / long lasting, that they won't officially say 'it works' before they tested every program with the I7s. Feels like a lame way to just say something instead of 'we haven't tested it yet so proceed at your own risk' - which we anyway do :098:

I can't imagine them being cut off from Intel / AMD supplies. Anyway I've given up on taking their hardware recommendations too serious already during the PIII era when my system - which 'officially' had to be crap - was having about 20% more vst power than their highly recommended system. As far as I remember they never had a detailed hardware recommendation again afterwards, which seems to be a fair solution considering the possible amount of people which will just screw everything and complain afterwards.
Whatever, I got used to the frustration about the fact that the last remainig bits which would make cubendo 'perfect' in my opinion will perhaps never be implemented though they seem so obvious to me. Probably everyone has a different perspective on perfection.

BTW, a really long time Logic / Pro Trolls user recently came to my studio and after a day or two he mentioned my system's stability was 'remarkable'. This was with HT on...

Michael

StevenG

03-05-2009, 06:36 AM

I have an i7 in my editing room with HT on. Probably the best rig I've had since I've been using Nuendo.

Crashes have been related to either Virtual Katy's memory usage problems, or Audio Ease's crap VST implementation (speakerphone, though a great app, is the dodgiest plug I have).

Sam

03-05-2009, 06:41 AM

StevenG

03-05-2009, 06:52 AM

Thats inspiring to hear Steve! I was wondering how these things were faring out in the real world......what you you have in that i7 box as far as video/audio hardware? Do you have the expansion slots fairly full up?

This rig is an overclocked i7 920 (OC to 3.2 Ghz). The audio card is an PCI RME AES32, and video comes out through a BlackMagic Design Decklink SP (PCI). The graphics card is a MSI Nvidia Geforce 9500GT. I'm running 3 gigs of RAM in triple channel mode. The OS is win Xp SP3 32 bit.
Also I installed a used 3Com 996BT server NIC that I had in my previous rig. These are invaluable in getting good performance out of our Gigabit network (we have all our projects and audio on a dedicated Dell server with 3.5Tb RAID 1+0 storage).

As I mentionned in the Gigabyte mobo thread, you need to watch out for the slots on the board, as Gb had the stupid idea of putting various coolers and heatsinks or components that make some of the slots unuseable.

Sam

03-05-2009, 07:26 AM

Thanks mate - things are looking bright then huh!?

I am about to pull the trigger on an RME AES32 with a Lynx Aroura 16 next week so good to hear they are playing nice...

I will wait until the dual socket versions settle in and the mobo's mature a little before I go there...but it definately seems like a logical step to take in 2009 :)

TAFKAT

03-05-2009, 09:04 AM

Sam,

On the point of the Dual Socket i7 Xeons platform,

While Apple are frothing on that they are available, neither Supermicro nor Tyan , the 2 traditional 3rd party Dual Socket Workstation motherboard vendors have not even breathed a word of their product line, and Intel are holding their cards very close to their chests as well, I am still waiting on some engineering samples from Intel , so I think the dual socket systems a re a while off yet execpt for the iFruits.

Something that does peev me , but as a close friend noted this week, he suggested they may buy a few more units than me... :wink:

Maybe so, but its interesting that HP, Dell , etc still haven't been allowed to bring their units to market either...

The Single Socket motherboards have matured nicely, but as Steve has mentioned, some of the slot layouts are a little less than intelligent..

@ Steven,

Interesting to see you are running an overclocked unit in a mission critical environment , not that there is anything wrong with that. I would be interested to know just how much you are taxing the unit ?

You wouldn't have time to run some DAWbench DSP -RXC numbers for me would you, I would love to see what the system spits back.. ?

The test is a lot quicker than my earlier benches btw..

BTW: I also agree that the i7 have been the best systems I have ever run Cubendo on, Hyperthreading is scaling better than 4-8 physical cores on the earlier architecture as I mentioned in my O.P, pity Steinberg haven't dotted the i's there and managed to get a unit in house , times must be tight.... :eusa_whistle:

:009:

StevenG

03-05-2009, 09:16 AM

@ Tafkat:

I did keep a close eye on CPU temps during the first few days, but as all our computers are in a climate controlled machine room, I'm not worrying too much about overheating.
The sound editing project I'm on right now has around 100 tracks, but no plugins at all for the moment, and I'm not at a stage where I'm writing automation all over the place. The CPU sits at 3%... And that's with the Decklink card running. The 920 CPU is a 2.7 GHz chip, I'm OCing it to a safe 3GHz for laughs.

I wont have time to benchmark until in two weeks time, as I'm out of the studio recording foley next week (on Pyramix, yuck!!). But please remind me to do some testing, I will gladly oblige.

Steven

TAFKAT

03-05-2009, 09:30 AM

Hey Steven,

Thanks for that, I'll be interested what the system does throw at low latency packed with the Decklink as well, which does raise DPC latency a touch.

3% CPU usage , take it easy Bro, don't want to run too close to the edge there.. :D

Pyramix eh, I have a few clients using it , mostly for Orchestral recordings and I think its great for certain areas. Still a little grey in others, especially in the Native arena..

BTW, Please call me Vin.. , I think we have known each other long enough now.. :wink:

Beat Productions

03-06-2009, 10:43 AM

This rig is an overclocked i7 920 (OC to 3.2 Ghz). The audio card is an PCI RME AES32, and video comes out through a BlackMagic Design Decklink SP (PCI). The graphics card is a MSI Nvidia Geforce 9500GT. I'm running 3 gigs of RAM in triple channel mode. The OS is win Xp SP3 32 bit.
Also I installed a used 3Com 996BT server NIC that I had in my previous rig. These are invaluable in getting good performance out of our Gigabit network (we have all our projects and audio on a dedicated Dell server with 3.5Tb RAID 1+0 storage).

As I mentionned in the Gigabyte mobo thread, you need to watch out for the slots on the board, as Gb had the stupid idea of putting various coolers and heatsinks or components that make some of the slots unuseable.

Steve:
Not to hijack, but what temperatures are you getting overclocked to 3.2? My I7 setup is in an isolated box and even with a new Thermaltake heatsink I can't overclock without getting into the 70s celsius.
BTW I LOVE my new It, gigabyte UD3R and 6 gigs of patriot 1600 ram.

Jcschild

03-09-2009, 06:51 PM

Hey Vin,

RE the Xeons.. well i have had this POS dual i7 here for what 5-6 weeks maybe longer.
we are finally shipping it back to Intel and have asked for a ready for release board and processors.

multiple bios requests emails with engineers etc... not worth the hassle...

we still have not been able to run a single benchmark...
the first "alpha" litteraly bios took 30 minutes from reboot to post to windows.

we tried using it (installing windows from scratch) got past the installs to trying memory configs and a few other things...
more bios more intel making wonder whos awake over there...

we gave up looks like i will be benchmarking a stinking apple first :icon_twisted:

Scott
ADK

TAFKAT

03-09-2009, 08:14 PM

Hey Scott,

Thanks for the heads up Mate.. :-)

I am hitting some brick walls at my end as well getting some joy from them.

Hopefully the board they can muster will a lot closer to a release product, that is if/when they actually wake up from their slumber..

Re the MacPro's , yeh that maybe the best option for now.. , try and get at least the 2.66GHZ unit if possible, but I suppose what ever you can get will be better than what we have now...

Peace

Jcschild

03-23-2009, 01:39 PM

well now Steiny is telling people to turn off HT on Core i7
http://www.cubase.net/phpbb2/viewtopic.php?t=114864

psvennevig

03-23-2009, 02:20 PM

I can't understand why they recommend this? To me it feels like they haven't really tested it. OR they have used a very early i7 sample from Intel, with a Intel mobo or something.

Very strange. And also that they posted this to the KB this early is also not common.

Curious to what plugin causes the problem :-)

P

TAFKAT

03-23-2009, 04:06 PM

Hey Pal, Scott,

We spoke about this earlier this month on this thread when I first posted a heads up about the so called knowledge base article..

I made a personal note not to comment on anything the geniuses at Steinberg post on other forums, but this is just beyond ridiculous , and it is causing major confusion in the market place with prospective i7 clients for me..

He is simply monkeying the official company line, and it really is about time Steinberg employ someone outside of the over worn brain trust they have continually dragging the place down.., and get some fresh blood that actually know their arse from their elbow on technical matters .., as it stands, this is all I see..

http://www.dawbench.com/images/4wm.jpg

I am willing to bet they have done NO testing whats so ever , or have any qualified data to show that Hyperthreading needs to be switched off on i7 systems , if they did , they would be able to give a detailed response , not offer up the offal they keep spitting out.

It is beyond ridiculous that not only do Steinberg continually hide their heads over the bad , but they also can't acknowledge the good , this is just a cherry on top of the whole x-scaling is still broken on 8 cores in XP/Vista fiasco..

Arghhhhhhhhhhhhhhhhhh... :headbang-1:

Someone link this thread over there please.. , time for the regular C.net guys to get some clearer detail on this..

Jcschild

03-23-2009, 06:16 PM

i just knew you would enjoy that! i love making your day!
:wink:

Daryl

03-23-2009, 06:59 PM

I would imagine that the SB tests have been done by "top men", no less.

D

ROCKINROG

03-23-2009, 08:54 PM

:rotfl: :crap:

TAFKAT

03-24-2009, 04:30 PM

i just knew you would enjoy that! i love making your day!
:wink:

Yeh, thanks... Mate... :-)

I just read the reply to the request for more details.., the response is typical when in fact they have nothing , and not even worth quoting here .

In short its B.S.

At least we know now that they have i7's in house, and the genius who is piloting them , who can't resolve spiking at low latency so of course it has to be the i7 Hyperthreading... :icon_rolleyes:

Funny no one here is experiencing the same thing related to Hyperthreading on the i7's.

I suspect this is also the genius who tacked the i7 Hyperthreading info onto the 5-6 year old Netburst Hyperthreading article..

These reps are Steinbergs own worse enemy, and an industry joke IMO.

Anyone here who is going to Messe needs to make it clear to them that they are not doing themselves or the end users any favours by maintaining this approach.

I know if I was attending , I would probably get arrested as I doubt I could keep a civil tongue..

Jcschild

03-24-2009, 04:50 PM

I would probably get arrested as I doubt I could keep a civil tongue..

now i find that hard to believe and completely out of character for you.
:rotfl::rotfl:

Fredo:smash: <-vin

TAFKAT

07-23-2009, 03:27 AM

Quick heads up..

Just putting the finishing touches to this years x-scaling shoot out between Cubendo 5 / Reaper 3 / Sonar 8 ( yeh I know Sonar 9 is only a few months away )

I'll have results for XPSP3 and Win7 32 / 64 RC1 , before anyone asks - I'll be leapfrogging Vista again.. :-)

Preliminary Win7 results are a touch above XPSP3 , very impreesed with that for an RC build of the O.S, and very early days in regards to me learning the tweaks ..

psvennevig

07-23-2009, 08:36 AM

Cool!

And this proves I guess that Win7 will/can be a major success for Microsoft also as a OS for a DAW.

Finally the common OS for a Windows DAW will be a x64 based one.
Bye bye stupid memory problems and 4GB limit for i/o mapped mem spaces for drivers...

Pål

efernan

07-23-2009, 04:52 PM

Windows 7 is performing better than Windows XP?

Wow, that would be a surprise.

Which versions (x32 / x64)?

Thanks!

TAFKAT

07-23-2009, 05:33 PM

Hey Efernan,

Both - performance is identical across both 32 and 64 bit versions with C5 and my DAWbench tests session.. :-)

My testing is still preliminary, but at 032 samples Win7 outperformed XPSP3 by a small amount , all very smooth. There was a nice solidity to the system running close to the edge that Vista never seemed to have, so I can see this as being the next DAW O.S of choice for many..

Still early days, its still not without its quirks, but because I am still dealing with an RC , it wouldn't be fair to share them just yet..

Heres one interesting one I will share for those poking around with Win7 RC.

On initially booting the system , when it first lands on the desktop , quickly call up Task Manager and make note of the memory usage, which is the same of Vista - around 900 to just over 1 GB of RAM at idle, yeh you read that right.., now just wait another 30 seconds, there is some minor thrashing from the drive and boom, the memory usage is halved.. LOL

I still have no idea what process or service is tying up that extra 500 odd MB on boot. Heres the rub, start tweaking and it stops the memory purge, even something as simple as removing IE8, yep, you can remove IE8 now, at the expense of using an extra 500 MB... LOL

Its no doubt going to be interesting ..

Daryl

07-23-2009, 06:09 PM

I still have no idea what process or service is tying up that extra 500 odd MB on boot. Heres the rub, start tweaking and it stops the memory purge, even something as simple as removing IE8, yep, you can remove IE8 now, at the expense of using an extra 500 MB... LOL

So how will that work in Europe? Will that mean that because we don't have IE8, we won't need to remove it, and will therefore have a lower footprint than the rest of the world? I can't wait to try all this out, when I get my next studio upgrade. :D

D

TAFKAT

07-23-2009, 06:36 PM

Hey Daryl,

Nooo, Soory Mate you misunderstood me.., un-installing IE8 actually stops the memory purge after booting on the test system I was tweaking, so you actually are using 500MB if you remove it... :icon_eek3:

So the European OEM builds will not have IE8 at all.., hmmm, need to get a hold of them for a look see..

TAFKAT

08-19-2009, 07:34 PM

Quick heads up,

Intel will be releasing i5/P55 the replacement for the Core2Quad / P45 platform in early September.

The new platform is based on Lynfield series CPU's , which differ from the Nehalem in regards to only having Dual Channel DDR3 instead of Tri Channel, and also have the PCie x16 lanes on die as well.

To confuse matters Intel will also release i7 branded variants of the new chip, meaning those models have HT where the i5 does not. Idiots. !

The P55 system architecture differs from the X58 in that it does not use QPI, but instead the earlier DMI-PCIe based interconnect , so the platform is a mixture of Core i7 CPU architeture and the earlier style interconnect.

It will be interesting how it performs for audio. I suspect better than the Core2Quad/P45, but not as well as the i7/X58

Numbers soonish, stay tuned. !

NDA lifts September 8.

TAFKAT

09-08-2009, 03:31 AM

NDA Lifted today..

http://www.dawbench.com/images/dbdsp-amd-v-intel-0809.JPG

Intels Lynfield /P55 platform is a refresh of the Nehalem chips first released with the i7's, memory is Dual Channel down from Tri Channel , PCIe x16 is now on-die eliminating the Northbridge.

The changes haven't really resulted in a major change in DAW performance, at least on the incremental benches , with the i5 750 that has no HT performing right on par with the equally clocked i7 920 with HT Off - both running at 2.66 GHZ.

The i5 750 systems are the direct replacement for the Core2Quad / P45 , and easily outperform the older architecture by a significant amount. It also has no trouble accounting for AMD's latest and greatest running at 3.40 GHZ. If anyone was wondering what ever happened to the old AMD V Intel Forum wars , I think the above graph is evidence enough .. :D

evanrabby

09-21-2009, 01:25 AM

any more tests or news on the i5?

to clarify: the ENTIRE NORTHBRIDGE is on the i5 cpu!

and doesent it outperform the p45/core2Quad by almost 50%?

i actually have a 965 chipset with q6600 quad core, and doesent the i5 DOUBLE that performance? and at low latencies, its even better?

TAFKAT

09-24-2009, 02:46 AM

Hey Evan,

Not much more to test on the i5-750 to be honest, it performs equal to an equally clocked i7-920 , with HT Off , and yes , the entire Northbridge is now on die.

Interestingly the 1156 "i7" variants are not scaling as well with HT as the i7 1366 chips, which may have something to do with the lack of QPI.

Still have no idea why the idiots called anything on the 1156 platform i7, they were originally all to be called i5's.

Anyhow,

Overall they are a great improvement on the previous gen Core2Quad chips, which amazingly will be EOL'ed a lot quicker than first anticipated.

You will definately see a healthy boost over your Q6600 rig.

:009:

evanrabby

11-21-2009, 06:22 PM

well, my first failed motherboard is the foxconn inferno katana.

now, for stability i already had to turn off everything in device manager i didnt need, lots of SYSTEM devices like 'motherboard resource' and unused pci express ports, and i always tweak windows to run almost no services, (ie only 19 processes running at 168 mb memory) try every slot, etc.

no matter what it will pop on graphics operations with reaper on low latency streaming at 64 samples when my test project was at 30% cpu only. my only problem i have ever had with my daw setups is window moves and resizes/graphics dropping out audio.

the issue is, i can drag a window around for a minute or so and then itll pop only once, and not in time with any movements, but it never drops out if im just adjusting parameters because theyre small graphic operations. so i can use the system, im just not happy.

nuendo is very stable, and will never get anywhere NEAR more than 30% on the asio meter even on m largest insane audio tracking project, but also exhibits tiny little pops at the 60% asio meter or more with that dawbench test.

also the dawbench count was similar only at 4 ghz overclock (a very stable overclock that will run up to 95 deg. without crashing).

my only thoughts are: is it the nvidia 7100 gs/driver? i thought a higher end card would be worse but mabe ill have to try.

and , mabe the foxconn p55a-s, the low end version, is tuned more for stability. oh well, i dont have the time to fully sus out another board... mabe ill find a fix.

xp sp2 or 3
motu pci-324 or 424 card (using 2408 mkII or 24i but its only the card that matters)
or m audio profire 26x26
2g ddr3 1333 ram at spd settings cas 8 at 1600 oc
nvidia geforece 7100 gs
foxconn inferno katana
i5 750 at 4ghz
ocz solid state hard disk, 30 gig
wdd 1tb hdd for record but i dont record with reaper only stream.

evanrabby

11-21-2009, 10:12 PM

the i5 has qpi. it shows up in device manager as 2 'quick path interconnect' devices. IE the memory controller is on die for sure, thats why we cant overvolt the ram or itll fry the chip past 1.6v, and my system crashes iv i go higher IE 1.7 or 1.8 volt.

it seems that only thing on dmi/southbridge is the non 16x pci-e, pci, lan and hdd controllers?

TAFKAT

11-22-2009, 12:53 AM

Re QPI on the i7 / X58 and i5 / P55 , I should clarify that what I was referring to was QPI external to the ondie controllers.

With the X58 the PCIe x16 controllers are on the Northbridg connected via QPI, I was also under the impression that the Northbridge to Southbrdge was also QPI , but I was incorrect , its DMI-PCIe. In that respect the P55 system with its PCIe x16 controller ondie connected via QPI, and the Southbridge connected via DMI-PCIe is essentially the same. Its simply a one chip configuration over the 2 of the X58.

I have done further testing with HT versions of the Lynfield chips, and at identical clockspeeds to the X58 platform, are on par using the DAWbench DSP test at all latency settings. I have no issue scaling the P55 systems to 95% ASIO using the RME reference cards. The results should scale accordingly to clockspeed , if they do not, then there are other subsystem variables at play.

evanrabby

11-22-2009, 01:45 AM

ahhh, so its true: the i5 is really identical, or even better considering theres no northbridge to cool off...

well i was definitely happy with the relative performance, i can run at 64 sample buffer with the same cpu power used as 128 sample buffers on my q6600, and i can now run 32 sample buffers as well, at a smaller project size, never possible on reaper before. id say its giving a almost doubling of power due to efficiencies.

ok, so i assume the subsystem variable is my motu audio card. its an assumption, but ill now have to find me an rme card thatll take 3x 8 channel lightpipe in and re test... mabe its time for me to buy the ***expensive*** rme solutions....

and its too bad, i bet its alot to do with the drivers too...

evanrabby

11-22-2009, 02:30 AM

ok i have one thing to add:

have you ever tested anything on a MOTU audio card? [edit, the pci-324 or 434 internal not the firewire 'crap'] they are a very common interface, and not really un-professional either relative to studio use, and i have considerable money invested in motu gear.
:sulkoff:
im curious if they are not on par in terms of asio headroom.

Vinark

11-22-2009, 03:47 AM

TAFKAT

11-22-2009, 07:50 AM

ahhh, so its true: the i5 is really identical, or even better considering theres no northbridge to cool off...

Both chips are based on Nehalem , so the performance as long as the platform holds up will be close to identical , the dual channel v tri channel memory is not showing any quantifiable difference in this particular test with my development system, that doesn't mean that it won't in sample specific tests , or the P55 is a better platform than X58. I have had reports from test partners where an i7 860 / RME combo was not scaling anywhere near my i7 920's , so it is sub system specific. The lack of Northbridge is one advantage in regards to cooling.

Personally I still prefer the X58 , higher memory expandability , possible upgradability to i9 - I'll say possible until its confirmed, no change to X58 is sheduled , also the socket design/platform is proven , P55 is still a babe in the woods.

Re MOTU, well the 424's didn't work at all on my X58 reference systems for some reason , but they do work on the P55 , they do not scale anywhere near the RME's at 032 samples which is what I tested the 424/2408 Mk3 I have as a qualification platform. I didn't note the exact figure as I was more interested in qualifying that they actually worked at all than the exact comparable figures at the time I tested it , but I will do so when I get another P55 system in front of me, I am currently just closing off some X58 systems. I don't keep a P55 system around as a development system.

The MOTU's held up O.K on the previous Core2 platform up against my other reference cards at the time which were RME and Lynx, but on the latest platform they are not scaling into the 90+ regions under ASIO, tapping out way before , also I should note, I don't have any of the video related issues you reported, and if I returned identical results clocking the systems to 4GHZ, I would be concerned.

evanrabby

11-25-2009, 06:25 AM

well, i get better results in nuendo by setting affinity to 3 of 4 cpus.

then i can run the asio with the motu 324/424 right to 100%! i mean, i get no dropout until the meter touches max, and also the task manager is NAILED to full on 100%

with dawbench nuendo, i can get about 90 going at 64 samples, and strangely, its no different at 128, actually its even less stable at 128 samples! 256 sample buffers allows about 110 or so.

so im finding its true these systems work excellent at low latencies, whereas older q6600 would show closer to double the cpu at 64 samples, here i get mabe 5% more cpu used...

TAFKAT

07-12-2010, 10:36 PM

Quick heads up,

I have a new Xeon W3680 Hexacore on the bench at the moment with an RME AIO card, so I thought I would run up some numbers on the DAWbench DSP RXC bench @ the stock clock of 3.33GHZ ( these chips are unlocked so can easily overclock on air and stock voltages, but its not my system so I am playing nice )

O.K Results:

Xeon W3680 @ 3.33GHZ

12GB PC12800 DDR3 - Tri Channel

RME HDSP AIO : Driver 3.083

032: HT OFF - 178 RXC : HT ON - 213 RXC

064: HT OFF - 188 RXC : HT ON - 225 RXC

128: HT OFF - 196 RXC : HT ON - 236 RXC

256: HT OFF - 200 RXC : HT ON - 240 RXC

Simply astounding performance .

Can't wait for some of the cheaper lower clocked version hit.

Jcschild

07-15-2010, 09:41 AM

your numbers match mine (980x) other than 32buffer! you got 12 more than me :-)
its nice to see validation!
come on now 4GHz get to it!

FYI the 970 launches the 18th (3.2GHz 6 core)
Scott
ADK

TAFKAT

07-21-2010, 05:21 PM

Yeh,

We are always within a few plugs , not sure about the 032 results, maybe I still have a few aces up my sleeve... LOL

Anyhow

Latest Intel/AMD Chips side by side..

http://dawbench.com/images/dawbenchdsp-amd-v-intel-10.jpg

AMD's new Hexacores are not even accounting for the current Intel Quads, while the Intel Hexacores have just further extended the performance variable - so nothing much has changed