1 person found this helpful
Try running my Premiere Pro Benchmark PPBM4 Then you can see how well your system performs with CS4 and maybe we can help with your configuration.
One immediate problem with your system is the video card, it is not a CUDA card and you will have to select a new nVidia card Adobe approved (or modifiable).
1 person found this helpful
Which Xeon Quad Cores are you running with? Do you often have time restrictive deadlines that you have to meet? I would definitely add another 2 Drive raid 0 since your working with AVCIntra. How many layers do you often edit with and what is your AE workload like?
Bill, I'll try to run that benchmark later today or tomorrow and get back to you with the results
EC, here are my specs taken right off the invoices:
Mobo: TYAN S5396WA2NRF Tempest i5400XT Dual LGA 771 Intel 5400 SS
Processors: 2 x Intel Xeon E5450 Harpertown 3.0GHz 12MB L2 Cache LGA 771 80W Quad-Core Processor
Memory: 20Gb Kingston 240-Pin DDR2 FB-DIMM ECC Fully Buffered DDR2 667 (PC2 5300) Dual Channel Kit Server Memory Model KVR667D2D4F5K2
System Drive: Cosair (Samsung, really) 256Gb SSD
Media: 2 x Seagate Barracuda XT ST32000641AS 2TB 7200 RPM 64MB Cache SATA 6.0Gb/s
GPU: SAPPHIRE Toxic 100269TXSR Radeon HD 4890 1GB 256-bit GDDR5 PCI Express 2.0 x16 HDCP Ready CrossFireX Support
Matrox Axio LE system (which I won't be able to use with CS5, which poses a problem for me in that PPro won't import clips that I've digitized with the Axio LE until Matrox introduces CS5 drivers which could be weeks or even months from now)
I have very restrictive deadlines. Last week I edited 5 HD spots in one day. Started at 9 a.m. and finished at 4...a.m.! That's why I'm willing to pay choke bucks for the fastest system possible (well up to $5K, let's say) and I'm wondering if it would be worth starting over with a current DDR3 config. That north side bus is so legacy and restrictive with my current set up!
I edit 4-12 layers, typically and many of those are text and graphic layers. Some AE Comps, though not regularly.
You suggested adding another RAID 0 because I edit AVCIntra. I thought this was a heavily compressed format and thus was way more CPU intensive than hard drive intensive. If I add 2 more drives, shouldn't I RAID all 4 together rather than to have 2 independent RAID setups of 2 drives each? (Harm, feel free to chime in on this one! Actually, please feel free to give advice to any of my posts.)
If you look about 20 lines down the Benchmark Results you will see a entry called "BillG Xeon x2 OC", this is my current editing system (soon to be replaced) but in it I temporarily overclocked my E5410's to run at 3.0 GHZ. They then are equivalent to your E5450's. It is a well tuned system with a big RAID array. That probalby would be the best you could achieve when you run PPBM4 I believe if you simply upgraded to a nVidia MPE CUDA compatible board you could run CS5 at least as an interim system while you figure out your best options for a new system. PS: I have no instability of CS4 on this machine which runs daily (non-overclocked). Since you run lots of layers the Quadro boards would be a good investment, otherwise it seems that the new Fermi GTX boards are working right now for several people.
Interesting question, does anyone know nVidia's schedule for Fermi Quadro's?
Message was edited by: Bill Gehrke
i know you dont want to here this but you need a new system.
the older Xeons are outclassed in many ways by even a single quad core i7.
particularly in the memory dept.
memory bandwidth on that system is like 7GB/s vs 28 on a single core i7
and 38GB on a newer dual Xeon.
CPU wise its beat as well.
with CS5 memory bandwidth is very important.
Bill, I think the best thing I can do right now is update to an MPE-certified video card and leave everything else alone to get started. I could spend 5 grand on a new system and maybe only realize a 10-15% increase in speed, when I will probably experience a much greater speed percentage increase by simply upgrading to CS5 with an MPE-validated video card. I might start out by buying the Quadro 4800 and sell it on ebay when the 480 (or something better, in the future) gets validated. Still wondering what I should do about my hard drive arrangement, but I think that I should leave that alone too, until I upgrade to CS5 and re-evaluate everthing.
Scott, you're correct: I didn't want to hear that, but I know you're right! What's a mother to do?.... Do you have a dual mobo & CPU recommendation? I can handle everything else (well okay, except the darn hard drive setup!)
A single I7 980 especially overclocked to 4.0GHz would be a considerable performance increase over that one. You will likely want a Quadro card with out the limitation at some point due to your editing style and project delivery times. What I would do is get an I7 system with a 285GTX for now and then plan on upgrading the video card once the Fermi Quadro's are out and supported. Then put the 285GTX in the old system and the Fermi Quadro in the new I7 system. Then you could turn the old system into a back up or render box.
AVCIntra has a high bit rate compared to any other compressed codec commonly used. The standard bit rate is 100Mb/s. It would definitely make use of a 2 drive raid zero when using more than 1 layer of video.
The earliest I have heard of Fermi Quadro's releasing is June. Keep in mind though they will be hard to get for 3 to 4 weeks or more after release. The release date will likely be a paper release and distribution will have real stock 3 to 4 weeks after. June is also just a rumor as the earliest date. It could still be later in Q2 or even Q3.
Okay, I think I've settled on the single i7 config. I couldn't justify the expense of going with two i7s on one mobo. Here's what I've come up with (if I've violated Adobe's forum policy by listing urls that identify vendors, like Newegg, I'll be glad to modify this):
Case: Thermaltake ArmorPlus(Armor+) VH6000SWA Silver Aluminum ATX Full Tower Computer Case
CPU: Intel Core i7-980X Extreme Edition Gulftown 3.33GHz 6 x 256KB L2 Cache 12MB L3 Cache LGA 1366 130W Six-Core Desktop Processor
Mo-Board: GIGABYTE GA-X58A-UD7 LGA 1366 Intel X58 SATA 6Gb/s USB 3.0 ATX Intel Motherboard
RAM: $1,000 24Gb of DDR3 2200 Brand TBD
Power Supply: Thermaltake Toughpower W0155RU 1000W ATX12V / EPS12V SLI Certified CrossFire Ready 80 PLUS Certified Active PFC Power Supply
GPU: PNY VCQFX5800-PCIE-PB Quadro FX 5800 4GB 512-bit GDDR3 PCI Express 2.0 x16 SLI Supported Workstation Video Card
$1,200 (approx. on ebay) http://www.newegg.com/Product/Product.aspx?Item=N82E16814133253&nm_mc=OTC-Channel&cm_mmc=O TC-channel-_-Video+Cards+-+Workstation-_-PNY+Technologies++Inc.-_-14133253&srccode=cii_724 0466&cpncode=18-6515286&DEPA=0&refer=channel&CMP=OTC-
System Drive: Corsair P256 CMFSSD-256GBG2D 2.5" 256GB SATA II MLC Internal Solid State Drive (SSD)
Media Drives: 2 x Seagate Barracuda XT ST32000641AS 2TB 7200 RPM 64MB Cache SATA 6.0Gb/s 3.5" Internal Hard Drive
Acquired http://www.newegg.com/Product/Product.aspx?Item=N82E16822148506&cm_re=seagate_2tb-_-22-148 -506-_-Product
56K Modem: Just kidding.
Total Cost (less RAM): $4,150 The GPU is the most likely component to change as time goes on.
Comments, computer experts? Again, if this list violates any Adobe Forum usage terms, I'll be glad to remove Newegg's name and/or other info.
Your estimate for 24 GB DDR3 2200 is extremely optimistic.
is already € 1,400 but only 1600 speed.
Your disk setup will be your major bottleneck. With the rest of your components, you will be waiting endlessly for your disks to finish. It seems a waste to get a FX5800 and an i7-980X when the disks are so amateurish.
Would 4 x 2Tb discs still be amateurish? I'd like to avoid going 5+ disc route (not counting my SSD I use for Sytem/App/Project Files). I no longer output to tape, so it's not a big deal if I don't always get perfectly smooth playback. I'm just hoping with the new system I've outlined, that it will take a lot less time to render (isn't that where the honkin' processor comes in?) In fact, I'd rather spend a little time rendering per edit session, than have 6+ HDs running in my system with the resultant additional cooling required and the noise it will generate just to avoid rendering. Not to mention increased chances for HD failure. As far as the RAM goes, I was only off by $300 based on what you listed: http://www.newegg.com/Product/Product.aspx?Item=N82E16820104170&Tpk=KHX1600C9D3K6%2f24GX
for AVCHD/AVC Intro, 2 sets raid o raids are more than enough
increasing the raid arrays do nothing.
anything less is not good either
Is that one set for media and the other set for preview (Render) files?
Okay, let me ask this then: why would I need a RAID 0 setup for my preview (rendered) files? Since the whole idea of rendering is to take multiple layers and render it down to 1 leads me to believe that the disk I/O requirements for rendered files must be minimal and thus could be handled with a single, dedicated-to-rendering drive. Comments? Anyone?
Felix, there are many opinions around, and justified different ways. I for one would prefer one larger array so I have less file management problems with my many smaller projects. I do know from my past experience in benchmarking that unless your third disk/array is just as fast as the project files disk/array that you will loose performance. That is why Scott and Eric are suggesting two arrays. Incidently rendered files are less and less important Even in CS4 I rarely rendered the whole timline, now with CS5 and with the right graphics board for the number of layers you have has me almost convinced to probably drop the Render Timeline score from the total score in my new PPBM benchmark.
Thanks for that Bill (and everyone else). I'm about to settle on the specs I outlined, but am willing to go with one 4-disc RAID array or two 2-disc RAID arrays. I guess the last question I have is does it make sense to go with a dual processor setup or does that kind of defeat the on-chip memory controller that the current CPUs feature?
2 sets of raid 0 are better than 1 set of 4 generally.
memory bandwidth is actually better on dual Xeon vs single
however the 6 core 980x was beating the 12 core 5660's (12 cores @ 2.8GHz)
in our tests....
Well in your opinion Scott and everyone else, do you feel the performace gain of dual Xeons is worth the extra $1200-$1500 cost? I've always heard you can't have too much processing power, but with 64-bit, MPE & DDR3, I'm just wondering if I'll really notice the performance gain. Again, most of my work is 4-12 layer High Def commercials, so it's not like I'm waiting all day on rendering...but it is taking long enough, where I don't mind investing in a new rig. Just not sure if the added cost of dual Xeon's is going to enable me to realize a significant performance gain over a single 980X. I'm already up to $4,100 with a single CPU-based system!
I do not, look at Scott's and Eric's benchmarks with CS5 and you will see that CS5 does not for some reason to be optimised for dual processor systems a single 6-core processor always beats any dual processor system system.
Bill, holy smokes! Thanks for that. Perhaps since "they" ditched the North Side Bus and put the memory management in the CPU, maybe there's some sort of speed hit with PPro in trying to synchronize the two CPUs. I don't mind saving $1,000+ dollars (the plus is for the beefier mobo required for dual) by not purchasing two of those expensive 980X's! Of course when I have this built, I'll report back my findings. I really owe you guys! Of course those findings will reflect my new RAID setup(s). Speaking of which, Harm I consider you the HD god! Can you give me final thoughts on how to Arrange the HDs? A couple of votes have been on dual RAID 0s. You seem to opt for a single, multi-HD array. I'm all eyes...
Take another look at my Benchmark Results and see if you find any systems at all that have two RAID systems plus the OS disk even on the list. If I had any input data that might confitrm that theory, I might be able to suggest such a configuration. Maybe I will have to put that on my long list of experiments to attempt to find that answer.
The benchmarks we did were with 2 2xDrive raid 0 arrays and 1 4Drive raid 0 array. There was little to no difference between them. Mind you if you are dealing with R3D or uncompressed then drive requirements change.
EC, thanks for that. I'll be working with AVCIntra and/or DVCPro HD. Those are the formats available to record to on the Panasonic HPX-370. Both have similar data rates. By the way, how do you delegate the data flow to the two independent RAID setups? (i.e. Media to one and Preview (Render) files to the other?)
With those Codecs, the 2x 2 drive raid 0 should be fine up to 6 to 8 video streams on tracks. If you are working with more often then you will likely want a bigger raid array configuration depending on how many drives you have to configure. We configure and test systems with the following.
1: OS Drive
2: Raid 0 array 1 - Project Files, Media, Preview Files
3: Raid 0 array 2 - Export Files, Media Authoring/Mastering Files, AE Cache, PS and Illustrator Scratch Disks.
If you want redundancy, then 1 larger raid 5 is the way to go but that requires more disks depending on what data rate you need for the amount of streams with those codecs.
One nice thing about a new virgin computer system. You can tune it and then easily try it with both configurations and see if you notice any difference with your media or measure any difference with benchmarks. Unfortunately playback stuttering is something that I have never been able to quantify. I would guess it might be media/project on one array and output files on the other. Since the prime new feature of CS5 being less dependent on ever (with the appropriate CUDA card) having to create any render files it may not be significant, but put them on the second array as unless you really heavily multitask. The preview files usage simultaneously with writing output files should be a rare occasion.
In the Benchmarks I tried the Preview directory on the second array to see if there was any performance change whether in Scrubbing or playback and there was none. Rendering might be a little slower but was so negligible I didn't notice it with my testing's. The export times would be faster with the preview files and export on different arrays. With CS5, rendering will be far less frequent and time intensive so having the Preview with the Media wont slow you down. Playback after render in CS5 is so light on the system, you can playback 4K R3D at full res and hardly push the CPU above 5 to 10 percent from what I have seen and most of the codec's bit rate now days is so low that playback and scrubbing really doesn't push the drives when playing from a file. This was more of a configuration issue when drives were slower and uncompressed was more commonly used.
Eric, please straighten me out on this statement.
The export times would be faster with the preview files and export on different arrays.
Isn't that true only if you check "Use Preview Files"? And isn't it true that for the best quality you do not want to use the preview files for final ouput?
Yes that is if you use the Preview files.
Honestly I always thought Media Encoder was pulling from the preview files anyway and you had to render the effects on the sequence before exporting. Obviously I was foolishly assuming and was incorrect. Thanks for the slap on the head.
Erc, you say "most of the codec's bit rate now days is so low that playback and scrubbing really doesn't push the drives when playing from a file." I'll be using AVCIntra and/or DVCPro HD. Both of these formats use 100 mbs. And if I have two or three layers of clips like these, plus titles, transitions, etc. won't I be pushing the HDs big time? Do you still think a 2 x 2-HD RAID 0 is enough? Right now I'm using HDV with a Matrox Axio LE and scrubbing's awful with a simple 2-drive RAID 0 for media.
8 bits = 1 byte. 100Mbits = 12.5MB(MBytes)per sec per video stream playing back. A 2 Drive raid 0 with the 1TB WD 64Meg cache Drives average read speed on the onboard controller is 195 to 215MB per sec. That gives you a very good idea how many AVCInta video streams you can playback at once. Mind you after the capacity of the drives exceeds 50% the raid 0 will slow down. So you do not want to archive to that array.