Hi Marin,

I think we agree then. The emphasis is on sustained - i.e not interrupted or blocked by other devices.

That, of course, is what PCIe is seeking to address by QoS - Quality of Service. It means guarranteed response times to service requests fom a device. PCI never had such a concept.

The technology underlying this is that PCI is a shared bus, and never even set out to guarantee any specific response times. Soundcards needed that guarantee, and worked only because PCI (provided you didn't put anything greedy on the PCI bus alongside the soundcards) often managed deliver on time, even though it never promised to do so.

Note the word "often". Not always. That's the source of all our heart-aches over the last ten years.

My theory is that the PCIe designers simply built a PCIe - PCI bridge that performed to the standard that the PCI spec required, in ignorance of the fact that pro audio stuff depended on much better performance than the PCI spec provided for.


Peter

--------------------
PaQ

> The technology underlying this is that PCI is a shared bus, and never even set out to guarantee any specific response times.

> Soundcards needed that guarantee, and worked only because PCI (provided you didn't put anything greedy on the PCI bus alongside the soundcards) often managed deliver on time, even though it never promised to do so

This's not totally true Peter, PCI certainly does deliver on time, but it's bandwidth/access is shared between devices. Soundcards didn't need that guarantee because they simply added onboard buffers and that's why they worked, not because PCI often managed to deliver on time.

Regarding being blocked by other devices, this has caused lots of problems, but is usually due to poorely designed cards, not limitations with the PCI spec (not withstanding the fact it's a shared bus).

There's actually very few cards out there that properly implement the PCI spec, and also many mainboards that don't either (Via springs to mind).

It'll be interesting to see what real advantages (in terms of latency etc) PCIe brings.

Cheers

Edited by kris (12/01/06 07:05 PM)

Indeed kris - buffers keep the streamed audio going until the next 'delivery' of data arrives, and as the buffers get smaller these deliveries need happen more and more often, until eventually some other device asks for the PCI bus at just the wrong time and you get a hiccup (technical term that )

I for one will be extremely glad to be able to report on even one positive advantage of PCI Express for the musician, rather than the current host of teething troubles and stories of untapped potential.


Martin

--------------------
YewTreeMagic

Martin/Peter, have you heard any more developments on PCI-E support, UA, TC, RME etc. ??

I've spoken to a few companies that have promised support, but not a great deal of definite dates yet (partly due to the lack of supporting chipsets). It'll be interesting to see what kind of latencies we can achieve under Vista and PCI-E. Things aren't bad at the moment with XP/PCI, but once you start adding UADs and Powercores etc. etc...

Quote kris:

---

> The technology underlying this is that PCI is a shared bus, and never even set out to guarantee any specific response times.

> Soundcards needed that guarantee, and worked only because PCI (provided you didn't put anything greedy on the PCI bus alongside the soundcards) often managed deliver on time, even though it never promised to do so.

This's not totally true Peter, PCI certainly does deliver on time, but it's bandwidth/access is shared between devices. Soundcards didn't need that guarantee because they simply added onboard buffers and that's why they worked, not because PCI often managed to deliver on time.

---

Even with an on-board buffer, Soundcards need a guaranteed response time, or the buffer can overflow. PCI has no mechanism by which it can guarantee to meet that response requirement. PCI latency was an attempt to provide such a feature, but as an individual device could over-ride it or mis-behave, PCI can offer no guarantee in general.

This is not surprising. PCI is now an ancient design, and was never intended for any sort of real time operation... so I'm not criticizing PCI, simply observing that though it frequently did deliver on time, it was never contracted to do so. My suggestion is that the PCIe designers simply added a PCI bridge that provided what PCI had contracted to do, but no more. Soundcards need better than that, so the won't work on the PCI bus of many PCIe chipsets.



I have heard of some PCIe soundcards under test, and the story is that they are very stable. PCIe is designed for real time operation amd can deliver guaranteed response time (QoS) unlike all the other busses on offer. I expect that when we move to a proper PCIe environment, and stop trying to support PCI devices in a PCIe system and Windows provides native PCIe support we will look back in horror at the last ten years and wonder who we ever survived with our sanity intact.


If indeed we have.



Peter
Peter

--------------------
PaQ

Quote kris:

---

Martin/Peter, have you heard any more developments on PCI-E support, UA, TC, RME etc. ??

---

I haven't, but the NAMM show starts next week, so we may get some announcements there, and if not then, at the Frankfurt Musikmesse in late March.


Martin

--------------------
YewTreeMagic

Hi,

I have not heard anything either. But, as time passes I'm incresingly convinced that this is what will happen:

1. nVidia will either fix their PCIe - PCI problem with the release of a chipset for use with socket M2, or never bother. 50-50.

2. Motherboards with no PCI slots at all (just PCIe) will start to appear this year, possibly as early as late Q2, possibly on socket M2 mobos

3. PCIe Soundcards and FX cards will appear late Q1 or Q2. Nce a few manufactures have jumped the trest will follow within 6 months.

4. New builds using PCI will be a thing of th epast by the end of the year.


Peter

--------------------
PaQ

Hi,
I've just built myself a new music pc with the following spec:-

x2 3800+ dual CPU with Zalman CNPS9500 cooler
Abit AV8 Motherboard (Via K8T800 chipset)
2GB Corsair XMS3200C2PT Ram
XFX GeForce 6200 AGP passive cooled Video Card
Western Digital 250gb SATA2 H/drive
M-Audio Audiophile 2496 PCI
Akasa Ultra quiet 400w PSU
Antec P180 case

However I've overclocked the CPU to 2520mhz (from 2000mhz std) and this thing REALLY flies. According to Sisoft Sandra it is 10% faster than an x2 4800 and the memory bandwidth is higher too because its running at 210 rather than 200mhz and with CAS2.5 and low timings . It is rock solid stable at this, it will actually run "sort of" stable at 2600mhz but then it falls over very occasionally under heavy testing and even once is too often for me, so I settled on 2500mhz and torture tested it for two days without a single problem and now I'm very happy It is whisper quiet, the P180 case has a bit of soundproofing and very quiet fans (I'm using two case fans on their lowest settings) and with the zalman fan on its slowest setting I can't get the cpu temp above 49 deg even at extended 100% cpu duty, its usually around 35deg. BTW One thing I did notice is that some of the torture test programs only seem to run on a single core so can't take the processor above 50% duty, so beware!

I think that the Abit AV8 is another very good alternative to the Asus A8V as it has a pretty similar spec but includes a couple of firewire ports, the Via K8T800 chipset is AGP rather than PCI-e and seems well suited for DAW duties. I've had both Abit and Asus boards in the past and I think they are of equal quality, but the Abit possibly has the edge for overclocking. I have no crackles or pops from the audio at all, it runs any of the cubase SX 3.1 demo songs at about 20% cpu load whereas my old 2.8ghz pentium was almost flat out and the audio was breaking up badly. According to Dskbench I'm getting about 58mbps transfer rate (read and write) from the disk, however I partitioned 30gb as a separate audio drive, I then found that if I put the partition after the main one it is slightly slower, but if it is put before it (presumably on the outside edge of the disk) I get 70mbps read and write rate! This is from a SATA11 disk jumpered down to SATA 1 because the motherboard is SATA 1 only and without a RAID array.

This whole system has cost me about £850 and for performance it beats ready built DAW's costing twice the price or more. I know that overclocking frightens a lot of people, but in this case do consider the fact that the X2 3800 has exactly the same core as the x2 4800 (except for cache size) and therefore should be able to run at the same speed. The overclocking potential for the X2 4800 is a lot less, 10 to 15% seems usual rather than the 25% or more of the X2 3800, however every chip is different and there are no guarantees (in every sense!). If you can afford it then save a couple of days of hassle and buy an X2 4800 system, if not then this route may be worth considering if you need this level of performance.

Allan

Hi,

Well done...


Actually, you can run two copies of Prime95. It's in the help uder Affinity:

1. Consruct two shortcusts to Prime with the -A0 and -A1 switches added.
2. Set CPU affinity (in Options I think)



Peter

Peter

--------------------
PaQ

Quote:

---

Actually, you can run two copies of Prime95. It's in the help uder Affinity:

1. Consruct two shortcusts to Prime with the -A0 and -A1 switches added.
2. Set CPU affinity (in Options I think)

---

Ah! Didn't think of running two instances of Prime95, thanks for the help.

One thing that does need to be cleared up is if XP home supports dual core processors. From what I can find out it definately does NOT support dual CPU's, however it will support dual core Pentium but needs SP2 installed. I'm not sure if it supports dual core Athlons though.

I'm running XP pro so not a concern for me but would be nice to know the answer.

Allan

--------------------
Top Tip: RAPPERS. Avoid having to say 'know what I'm sayin' all the time by actually speaking clearly in the first place. (Viz Comic)

Yes, XP Home supports a single processor with multiple cores. Currently running a 4200 in XP home.

--------------------
ochremusic.com

I took a bit of a risk and bought a dual-core Gateway off the shelf at an electronics store. $850 U.S. dollars was too much of a price breakthrough for me to pass up, even though I knew I'd have to tweak and throw in a few new components.

Motherboard - FIC C51GU01 (Nforce 4)
CPU - Athlon 4200 X2
Graphics - Geforce 6100 (glorified integrated graphics), although I intend to upgrade soon
Audio Interface - Aardvark 24/96 (it gave me compatibility headaches at first but has stabilized with a little work)

Despite some of those potential pitfalls (generic MOBO, obsolete audio interface with notoriously bad drivers), performance has been nothing short of breathtaking. In Sonar 5, which is optimized for multiple CPUs/cores, I can run so many instances of the most CPU-intensive plugs I have (Perfect Space convolution reverb, Dimension Pro, Sonitus Multiband, etc.) that CPU-usage is effectively a non-issue for my needs. I upgraded from a 1.8 ghz Pentium 4, and I can run projects that had half of the tracks frozen/bounced (and were still maxing out my CPU) with no freezing at low latencies and still have power to spare for extra plugins.

One interesting note for musicians is that this off-the-shelf system doesn't have a CPU-fan - just a humongous aluminum heat-sink and a BTX-style airflow layout (i.e two large, fairly quiet fans in the front and back of the case), although I did have to replace the noisy stock PSU with a Seasonic and I'm going to quiet the noisey WD hard-drive with a copper SmartDrive enclosure.

Can you see the make and model of that humungous aluminium heatsink by any chance?


Martin

--------------------
YewTreeMagic

Hello Guys & Girls?
I am looking at Intel Core Duo specs and I can't see the words "64 bit" anywhere? Are they dual 64 bit cores or dual 32 bit cores with Hyperthreading ??I've noticed that the systems I have found so far all use XP pro,and not X64.Are they cheating??
Cheers
Jack

No luck getting the heatsink information, Martin - I opened the case, but I couldn't see any identifying information.
Here's the model - link.

To be clear, it's far from quiet out of the box, but if you're comfortable changing the PSU and a few other tweaks, it's serviceable. It was just a nice bonus not having to invest in a new CPU fan/heatsink, always a tricky component to remove.

Jack - yes, the Pentium D series is 64-bit compatible.


Martin

--------------------
YewTreeMagic

What would also be interesting is to see if anyone owns a dual core opteron (single dual core chip) to see how it compares to the X2 systems (other than the fact its stupidly more expensive).

Anyone got a dual proc dual core system?

Given the comparatively small number of single processor dual-core systems popping up on this survey I expect dual processor dual-core systems to be extremely rare.

If anyone's got one up and running do let us know how it performs though!


Martin

--------------------
YewTreeMagic

I dont mean dual dual core systems. I'm interested if anyone has a single processor dual core opteron machine running. One of the 939 dual core opterons (165, 170 and 175). Would be interesting to see hwo they compare to the X2's.

biohazard

They perform the same as an X2 at equal clockspeed (with equal cache), but can usually (1) be overclocked higher and (2) run slightly cooler.

That’s my current setup:

/*/ A8V Deluxe Rev. 2 BIOS 1017
/*/ Athlon64 X2 4800 no OC
/*/ 4x1GB OCZ4001024ELDCPE 200MHz CL2.5-3-2-5
/*/ 2x Samsung Spinpoint SP251 250GB PATA1 & PATA2
/*/ 1x WD Raptor 740GD SATA1
/*/ 1x Maxtor Diamondline III 300GB 7B300S0 SATA2
/*/ Matrox P750 with xp2k_110_01_114_se_u Latency 32 IRQ 16 shared with FW1394
/*/ UAD-1 V4.1 PCI_1 Latency 64 IRQ 16 shared with FW1394
/*/ UAD-1 V4.1 PCI_2 Latency 64 IRQ 17 shared with Marvell GB Lan
/*/ UAD-1 V4.1 PCI_3 Latency 64 IRQ 18 shared with Promise IDE/SATA/RAID deactivated
/*/ RME HDSP Digiface 2.943 PCI_4 Latency 255 IRQ 19 not shared
/*/ UAD-1 V4.1 PCI_5 Latency 64 IRQ 16 shared with FW1394
/*/ WinXP Pro SP2 latest patches
/*/ Sonar 5.0.1 PE

This setup works up to 97% UAD CPU load without any pops and crackles i.e. 28 1176LN 7 PPI’s per UAD-1 @44.1kHz 256/512/1024 latency.

Looping within Sonar requires a UAD-1 CPU load below 90% @44.1kHz 1024 latency to work reliable otherwise host CPU spikes up to 100% with dropout.

The switches /3GB and Userva=2900 in boot.ini seem to make no obvious difference but I’m not 100% sure how to find out. Physical memory of 3.407.084KB is recognised. Physical address extension is automatically switched on without having /PAE in boot.ini:
[boot loader]
timeout=4
default=multi(0)disk(0)rdisk(0)partition(1)\WINDOWS
[operating systems]
multi(0)disk(0)rdisk(0)partition(1)\WINDOWS="Microsoft Windows XP Professional" /noexecute=optin /fastdetect /usepmtimer
multi(0)disk(0)rdisk(0)partition(1)\WINDOWS="Microsoft Windows XP Professional 3GB Switch" /3GB /Userva=2900 /noexecute=optin /fastdetect /usepmtimer

HKEY_LOCAL_MACHINE\SOFTWARE\Universal Audio\Powered Plugins\UAD HypTopo\DMATurbo=1. Setting this one to 0 causes crackles and CPU spikes during looping.

During installation of WinXP SP2 it was required to press F5 when pressing F6 is possible to choose the right HAL (halmacpi.dll) for ACPI Multiprocessor PC. Otherwise WinXP has not recognised the dual core CPU with its two cores. Only one core showed up in device manager after installation! (See MS KB299340)

Using latest AMD processor drivers 1.2.2 from AMD seem to be mandatory for correct recognition and function of CPU independent from the fact if cool and quiet is used or not. During DAW use CPU power scheme should be set to continuous operation to avoid problems from lowering voltage and frequency.

Most important thing is MS Hotfix 896256 for multiple processors! Without this patch WinXP SP2 will not use core 0 and core 1 in a load balanced way. Core 1 got the higher CPU load without having the patch installed and dropouts occurred. (See MS KB896256)

Lots of useful info there Winnie!

I did a straight install of XP SP2 on mine and dual processors were recognised no problem, I'm using Abit AV8 and X2 3800 so not sure if that would make a difference? I'm not using any UAD's either. All of the regedits are useful and I'm busy working my way through them now.........

X2 4800 and no OC? Must be good for 2.6ghz at least

Allan

--------------------
Top Tip: RAPPERS. Avoid having to say 'know what I'm sayin' all the time by actually speaking clearly in the first place. (Viz Comic)

Very interesting info Winnie - thanks!

I'm using an A8V with an Audiophle 192 (eugh) in slot 3 and UAD-1s in slots 2 and 4. Can't seem to get the UAD-1s to be recognised in device manager in slots 1, 3 and 5, which will mean I'll have to start shuffling things about if I ever decide to get another UAD-1. (Really wish they'd build a Firewire chassis for UAD-1 cards.) I do have LAN and SATA enabled though, which I currently need, which might cause probs I suppose.

Hi csl,

I replied to one of your post over there:
http://www.chrismilne.com/uadforums/viewtopic.php?t=4102

If you haven’t seen it yet:
I fear you have got one of these A8V Deluxe with partly not proper working PCI slots 1, 3 and 5 if the UAD-1’s are not recognised there. I just put in the UAD-1 in PCI slot 1, 2, 3, 5, and the RME HDSP in PCI slot 4 and they have been recognised in the first step. I think it is better to put the soundcard into PCI slot 4 because it is not shared with any other device.

Thanks Winnie - I was hoping this wasn't the case, but I've been considering buying a replacement too. I've never had a duff mobo before - is this quite a widespread problem with the A8V or could I have done something to cause the fault? Are slots 1, 3 and 5 commonly problematic? (Oh, it's a v2.00 revision too.)

I don't know if any BIOS settings need to be set for things to work properly. The A8V BIOS seems rather more complex than the one on my NF7-S, so I may have missed a page or two.

I've currently set pnp OS to 'yes' and APIC enabled - this ought to be correct, shouldn't it? I hope it's not broken as I'm already rather overdrawn as it is...

--------------------
ochremusic.com

I read somewhere that a couple of people had problems with not properly working PCI slots. I had no problems so far (keeping fingers crossed).
I have this settings in the BIOS:
Plug and Play O/S: No
ACPI 2.0 Support: Yes
ACPI APIC Support: Enabled
IRQ 3-15: PCI Device
I have no idea if this will do something to the relevant PCI slots?

Thanks for all your detailed info Winnie!

Anyone who wants to read more about the Microsoft Hotfix relating to multiple processors can do so here:

http://support.microsoft.com/default.aspx?scid=kb;en-us;896256

However, unless I've read this incorrectly, it seems to be a Power Management issue with the two cores of a dual-core processor, where Windows XP gets confused (particularly with single-threaded applications) when one or other core isn't working very hard, and automatically throttles down the 'adaptive processing throttling' feature, thus reducing CPU performance.

However, as far as I can see, if you've already disabled CPU throttling (by using the 'Always On' power scheme) you shouldn't suffer from this problem anyway. Interesting that you got the problem Winnie - what power scheme were you using?

By the way, anyone who wants to read more about Power Schemes should have a read of my December 2003 PC Notes column:

www.soundonsound.com/sos/dec03/articles/pcnotes.htm


Martin

--------------------
YewTreeMagic

Hi Martin,

Quote Martin Walker:

---

However, unless I've read this incorrectly, it seems to be a Power Management issue with the two cores of a dual-core processor, where Windows XP gets confused (particularly with single-threaded applications) when one or other core isn't working very hard, and automatically throttles down the 'adaptive processing throttling' feature, thus reducing CPU performance.

---

Yes, that’s what I understand from this article as well if C&Q is not set to 100% power scheme.

Quote Martin Walker:

---

However, as far as I can see, if you've already disabled CPU throttling (by using the 'Always On' power scheme) you shouldn't suffer from this problem anyway. Interesting that you got the problem Winnie - what power scheme were you using?

---

I have read somewhere else that an alternative would be to switch off C&Q completely instead of installing the hotfix patch. I’m not sure if this will turn out to be 100% true. Having the CPU working with 100% power scheme running one instance of Prime95 on both cores without having the hotfix patch installed showed more CPU load on core 1 than on core 0 in an unbalanced way. After the hotfix patch installation the picture looked different – more balanced CPU load on core 0. This can be monitored with the ‘AMD Power Monitor’ together with ‘AMD Dashboard’, Taskmanager and ProcessExplorer.

My resume on this is that this hotfix patch for multiple core’s/CPU’s different performance states like dynamic voltage and frequency scaling in response to system workloads works very well together with the AMD processor driver and C&Q concept of AMD. It is funny to measure how the system’s power consumption is reduced through a few mouse clicks. Unfortunately this does not really make sense for DAW use because the increase of voltage and frequency takes some milliseconds which introduces the risk of getting crackles again. DAW users have to pay the bill for the energy due to 100% power scheme.

Another aspect is that neither WinXP nor SP2 have been designed for this dual core application so that I can imagine that MS has included some further improvements under hood for the handling of single-/multi threaded applications in this hotfix patch which wasn’t there before because of the Linux competition regarding better scalability. But this is speculation.

Hmm. As the hotfix will be included in SP3 (as stated by MS on that page), I'd guess it's an essential patch for improving the way Windows handles dual cores. Plus there's no mention of AMD processors specifically, just 'dual core', which might suggest it could be useful. It's quite difficult to tell, with the page being rather vague, though I've heard quite a few folks now suggest it has nothing to do with C&Q.

But as usual, if it ain't broke...

Agreed - Cool 'n Quiet can't be instantaneous, as it's responding to CPU demand. Perhaps Steinberg and other audio application developers should investigate providing an option to link C&Q to their Transport Play buttons, so musicians get the benefit of cooler, quieter operation in-between takes, but disable this feature one you start recording or playing back, to avoid audio glitching

I'm interested that you still got better performance after installing that hotfix even thought you were using the 100% power scheme - as you say, perhaps other unannounced tweaks are included in that patch.


Martin

--------------------
YewTreeMagic

Hi Martin,

I'd just like to say that I've been getting SOS since 1987 and have greatly enjoyed your excellent columns over the years when you have been active. I actually stopped subscibing early last year but just took out a limited esub....mainly to access the Scan Computer X2 PC I've been using Scan since the early MicroMart days....a really excellent hard working crowd they are too. Also I couldn't wait any longer to see the K2 review!!

Anyway the point is that I'm also looking to build another PC and will probably 'risk' AMD! I thought that the following thread would be a worthwhile addition to this sticky~: -

http://www.soundsonline-forums.com/showthread.php?t=2061&highlight=ath lon

I hope this helps

Frank

PS This Steinberg multiprocessor pdf is also pointed at by the thread.....very very interesting..... Like!....Hyperthreading should be turned of with dual core processors....(I'm sure that you knew this but I didn't!!)

ftp://ftp.steinberg.net/Download/General_Documents/Multiprocessing_Tech_In fo.pdf

Hi,

That HotFix refers to mobile CPUs, though as CnQ is a development of mobile PowerSaviong technology it may apply to AMD desktops as well.

Frankly, I think CnQ is a waste of space. If you disable it an X2 will still run cool enough to make it quiet. The Point is that CnQ is actually only the Cool bit. Quiet simply exploits the Cool to turn the fans down. The bios does that (provided you tell it to), and it will still do it if CnQ is disabled.


Peter

--------------------
PaQ

Quote Martin Walker:

---

Thanks for all your detailed info Winnie!

Anyone who wants to read more about the Microsoft Hotfix relating to multiple processors can do so here:

http://support.microsoft.com/default.aspx?scid=kb;en-us;896256

However, unless I've read this incorrectly, it seems to be a Power Management issue with the two cores of a dual-core processor, where Windows XP gets confused (particularly with single-threaded applications) when one or other core isn't working very hard, and automatically throttles down the 'adaptive processing throttling' feature, thus reducing CPU performance.

---

I've been thinking about installing this hotfix - I've been getting some issues (possibly related to kontakt2) with the CPU load spiking up to 100%. I read somewhere (can't find the link anymore) that certain applications are having problems with the timers reading funnily on dual cores (which certainly could cause the app to think the cpu usage was spiking when it wasn't) and that this hotfix fixes it.

And if you look at the microsoft page it says the hotfix also addresses:

Quote:

---

The synchronization of the processor Time Stamp Counter (TSC) registers across processors when you use the ACPI Power Management timer on multiprocessor systems.

---

and more detail:

Quote:

---

Correct TSC synchronization
On some operating systems, the processor TSC may change the rate at which it counts. Additionally, the processor TSC may stop counting when specific processor power management features are used. On computers that have multiple processors, the TSC is typically the operating system hardware timer that supports calls to the kernel KeQueryPerformanceCounter function. When TSC does not increment monotonically, system components that use the kernel KeQueryPerformanceCounter function may not work correctly. To address this problem, Microsoft makes it possible for the ACPI Power Management Timer to be used as the operating system timer that supports the kernel KeQueryPerformanceCounter function. However, some programs may directly access the TSC by bypassing the Windows timer APIs. The multiple-processor Hardware Abstraction Layer (HAL) makes sure that the TSC registers on all processors on a multiple-processor computer remain closely synchronized. Therefore, access by system software that may be directed to different processors does not return different results. This change makes sure that the multiple-processor HAL continues to correctly synchronize the TSCs across all processors on a computer, even if the ACPI power management timer is used as the operating system hardware timer.

---

So, I think I'll try it next time I see a spiking type problem...

--------------------
Midipicks - the all new MIDI Guitar forum...

Ello...

This is my dual dual-core (quadcore) system built around May, June...

2x AMD DualCore Opteron 280s (4 x 2.4ghz)
Supermicro H8DCE Dual Opteron Nforce4 pro motherboard
4gig (4x1gb) Kingston HyperX EEC registered DDR400/3200
1x Western Digital RAPTOR 36gb 10k SATA drive
4x Hitachi 500GB Deskstar SATA drives
Matrox Millennium P650 128 PCI-Express graphics card running dual output at 1920x1200 DVI on
2x Dell 24-inch 2405FPW LCD
Magma 7 slot 32bit/33mhz PCI expansion chassis
RME FireFace 800 Soundcard
4x UAD1 dsp pci cards (in magma chassis)
3x TC Powercore MK2 pci cards
Seasonic S12-600 EPS12/SSI PSU

http://web.ukonline.co.uk/ten.dj/ti5.jpg

Using it in cubase sx3.1.1 and windows xp sp2. Have tried it with xp64 and it runs fine, but no uad or poco 64bit drivers yet so its not much use.

I originally had the Tyan K8WE 2895 motherboard which had some teething problems but worked fine after the 1.1 bios. I sold it for the SM H8DCE which has a better spec and much better future proofing, and it was up and running perfectly in no time at all.

The only problem I did encounter was running uad cards directly on the motherboard. They work fine in a magma chassis but give extremely odd results when directly on the board. It is wierd because it doesnt have the dreaded amd 8131 chipset that has caused a ton of problems on other dual cpu boards. I am actually hoping ua will come out with a version 2 which is either pci-e or firewire (or a choice of both) with much more power on each card. The current one is looking a bit long in the tooth and with all the new plugs announced we need MORE POWER!!!

Quad cpu support in cubase is simply stunning. I can load huge amounts of vsti, effects and audio tracks all at low latency without it even breaking into a sweat, and the responsiveness over two 24inch monitors at quite a high res is also very impressive (I was expecting some slow down but not a thing!!)

On my old daw (p4 3.4ghz) my biggest project was bouncing around 85% on the cpu meter, on the new daw it was steady at 35%...nuff said

I dabbled in clocking the 280s from 2.4ghz stock to 2.75 stable with ntunes mem freq adjust. Just wish the board supported voltage settings as might be able to squeeze a bit more out of em.

Yes, it definately cost a pretty penny but having the freedom to work exactly as you want without corner cutting or thinking about restrictions is quite liberating and has increased production a tremendous amount. Shouldnt have to think about upgrading for a while either

ten

Hi Ten,

I think you've just won the "mine is bigger than yours contest", that's some tool you have there!

I'm sure it did cost a pretty penny, but the other way to look at it is how much you have saved against the traditional hardware route, you would need a big room to fit it all in and you would be looking at a six or seven figure sum!

Surprised to see as many UAD and powercore cards, and even more surprised that they work happily without glitches as they must be causing a lot of traffic on the Pci buss. I'd have thought that with that sort of power you could forget about dsp cards and run all of the fx with vst plugins alone, but as you say it is pretty future proof and had a huge amount of power in reserve.

Did you manage the 2.7gh overclock with standard vcore voltage? Pretty amazing if you did, most processors will only go about 10% over without more volts.

Top job

Allan

--------------------
Top Tip: RAPPERS. Avoid having to say 'know what I'm sayin' all the time by actually speaking clearly in the first place. (Viz Comic)

Hi Allan,

The uads and powercores were actually left over from my old daw. I collected them over the years getting some good deals on ebay and from the readers adds here. I also got a magma to put them all in. I was hitting the top of the cpu meter easily with native plugs, so I wanted to offload as much as possible to the dsp cards.

I had 3 mk1 powercores to begin with but recently took advantedge of the upgrade to mk2s TC were having, and your right, I really dont need them all now I am thinking about selling off the UADS and magma completely and just going totally powercore as I love the reverbs and Sony plugs for it, although UA have just announced a bunch of new plugs so I might just hang on a while

I was never really impressed with native plugs, especially reverbs, compressors and EQs, so would always turn to the uad/poco alternative. Although now native stuff is beginning to catch up and with the cpus of today they have a lot more headroom to play with (although I must say nothing comes close to the MD3 plug yet!)

About the overclock, yes it was at stock voltage. I had to limit the mem bandwidth in the bios to 166, then slowly raised up the freq in ntune (I would really rather do everything from in the bios though.) It actually went a bit higher than 2.75 but it crapped out a few times so I backed it off. I only did it to test out clocking with Opterons, though and now I run completely at stock (boring!) Would be nice to have access to voltage controls in the bios and some phase change cooling and then see what we can do

Ten

Out of interest how much did that quad core set you back? I wont ever believe you if you hit a limit with the machine. You wont need to upgrade for fricking years now.

That system has to be worth a few pics in the magazine under the heading of... You lucky bar steward..

--------------------
nibbled to death by an Okapi http://www.soundclick.com/tubilahdog

small story

The project started out about this time last year. I was simply going to go with a dual cpu setup using opteron 250s which I got from ebay for a good price. I wanted to wait for the new breed of pci-e motherboards with onboard sata and the Tyan K8WE looked like it was going to be the first dual cpu with pci-e available. It was orginally scheduled for a late Dec release, but as always this was well overshot and got pushed back and back and I finally got hold of it around March. I was slightly worried as early reports of single cpu nforce4/pci-e boards with audio apps were very poor and even RME recommended not using them.

Thankfully dualcpu nforce4pro boards were not affected in the same way and actually gave very good results. I had some teething problems with the K8WE but after a couple of bios revisions the K8WE was up and running perfectly (and I still recommend the board totally for anyone looking at this kind of system with pci-x slots.)

Then AMD announced they were releasing dualcore opterons and GAS took hold 4 cpus in one machine I thought to myself.....after coming from only a single cpu setup just a couple of months before....yes please!!! Of course I took a gamble getting these as nothing in audioland took advantedge of dual dualcore cpus yet, but I couldnt pass on the opportunity I was given to aquire 2x 275s for half the UK street price from a chap in the states.

Then the SM H8DCE was announced and released in the same month (impressive!) This has 8 onboard sata ports, much better board layout and 4 pci-express slots compared to the K8WE's 2, so I sold the K8WE and bought one which was up and running without a single hiccup.

Unfortunately I still didnt have any software to take advantedge of dualcore cpus, and cubase ran like a complete dog with them. Thankfully Steinberg announced dualcore support with the next version (3.1) which came out a while later and ran simply amazingly.

280s were announced and I got another good deal on those so sold up the 275s.

The most recent addition is a virus TI which is awesome (software still has some teething problems.)

/story

I got some very good deals on the parts from ebay and purchasing from the usa.

Some examples are:

2x Opteron 280s for £1100 (cost about double that here)
4 gigs ddr 400 ram for £350 (same stuff costs £900 here!)
Fireface was £500 (about £850 here)

All of the above were purchased on ebay from peeps in the USA. Obviously you have to do the old import tax trick But I saved literally thousands of pounds purchasing 2nd hand kit from the states. I wouldnt of been able to afford it otherwise! Of course I have no warranty on a lot of the kit but thats the risk you take for getting it cheap as dirt.

The uads and powercores I picked up over the years from ebay and readers adds, some for as cheap as £100 from the usa!!!

The Dell 24inch flatscreens had an offer on at the time of buy one get the 2nd half price, so I had to go for it really

I worked out if I paid for the whole system new UK street prices it would of come to around £8000-£10000 (including dsp cards.) As it was this came to around £3500-£4500.

couple more pics.
http://web.ukonline.co.uk/ten.dj/9.jpg
http://web.ukonline.co.uk/ten.dj/10.jpg
http://web.ukonline.co.uk/ten.dj/11.jpg
http://web.ukonline.co.uk/ten.dj/12.jpg

ten