1 person found this helpful
If you see the results in the top 10% of the PPBM5 results list (indicated by D9), you'll find that all of the LGA 1155 systems on that list are overclocked. In fact, in that portion of the list only three of the i7-980X systems and four of the dual-Xeon X5680 systems are at their stock speeds.
And only 93 systems scored in the top 25% (indicated by Q3 or higher) in that list (as of 13 April 2011). Of those, the top-scoring stock-speed i7-970 system scored 216 seconds - not much faster than my top-scoring stock-speed i7-2600K system's score of 220 seconds. That 970 used a fast but expensive GTX 580 card but a single non-RAID drive for the media/project drive while my system has two 1TB 7200RPM drives in a RAID 0 but only a GTX 470 card. That difference would have widened somewhat had the disk subsystems of both systems been equal. And my score would have been much closer to the stock i7-970's result had I used a GTX 580 instead of the GTX 470.
As always fast and great in your responce
It's actually your system and the competing 970 at stock speed that has been my "for and against" when i am to finaly decide if im going to stick with the i7 2600K i have already built, or change it to a 970. I can do both within the budget i got, but then i cant buy an external backup system or UPS. My customer must buy later (wich i dont prefer) if i go for the 970.
I know how easy the SB is to overclock and how stable it is starting to become. If i send out a system to a customer and i overclock it, i gotta be shure that my client also can adjust these settings. And compared to OCing a Nehalem, SB is a piece of cake - for anyone. You only have to change the turbo multiplier for all oc's up to 4.5. If your going 4.7 or higher, just a few more settings have to be tweaked. That make sit possible to learn a customer to maintain a overclocked system because it will not req more than 1 hour to teach the customer how to keep a stable overclocked SB platform, in my opinnion.
But if i'm going for a 970, the best i can do is make a slight overclock and keep it that way in order to not have to rush out offering support if the system becomes unstable after long hours of stress/workload.
So for me an easy overclockable 2600K is better than a hard to overclock 970...for the end customer.
But then again it's the mather of pci lanes and memory. I am close to jumping on a 1366 platform, i just need to know if i have made a good choice or not with the SB platform. My customer is not even used to CS5 yet, he is going to start learning the system by traveling to Sweden and attending a CS5 course after easter. And this is a "starter system" for him (he will upgrade if he has to, enthusiast SB/Ivy Bridge) but i still want to get most for my 4K+ budget now.
Still waiting for 16GB of memory before i can test the current system i've buildt on the bench. I so want to go for an overclocked i72600K that is overclocked to 4.5GHz, rather than a costly 970 that is not overclocked because i cant depend on that my customer knows how to maintain an overclock on that CPU.
From what i could tell you mean that a i7-2600K system perform just as well, IF overclocked, as a 970 system, no mather if the latter got a better PCI/mem system...or am i wrong?
Btw, sry for my bad English, hope you understand what i'm writing
1 person found this helpful
More than 25% AE work or working with AVCIntra, R3D 2K or Cineform 2K or higher- I7 970 or higher
Mostly XDCam, AVCHD, H264, DVCPro HD, or Canon XF codecs - SB2600 or higher
R3D 4K or more than 50% AE work - Dual Xeon 5650's or higher.
That should give you a good general outline to work from.
My own concerns with the new SB CPUs are indeed for the missing PCI lanes. I personally do not consider it wise to be editing any solid state media with anything less than a RAID 3 controller. That means an Areca card, and assuming you also want GPU acceleration, that means you need more PCI lanes than are available on the current SB mobos.
So while the performance of the new CPUs is nice, especially for the price, the platform itself is lacking. I might build a gaming rig using an i7 2600K, but for editing, I really want those extra PCI lanes. That means waiting for Q4 2011 and the second set of SB CPUs.
Actually, Jim, there is an i7-860 (an LGA 1156 Lynnfield system that's the predecessor to the LGA 1155 Sandy Bridge platform) fairly high up on the PPBM5 list that used an Areca PCIe-x8 card and a GeForce GTX 460 graphics card that's been forced to run in x8 mode because of the Areca card. That system scored 80 seconds in the AVI Disk I/O test (this is due to RAID 5 instead of RAID 3) and a shorter-than-expected 6 seconds in the MPE Export test. What that tells you is that the GTX 460 GPU does not take full advantage of even x8, let alone x16, bandwidth.
To all of you guys, thanks a lot for helping me, it is really aprissiated. All of the info is making me learn more so this is so great and fun.
The one thing that concerns me is that a lot of people are using the PPBM5 benchmark as a reference when setting up a new system. If the SB platforms lack of PCI lanes is so dramatic to the performance, shouldn't the Benchmark reflect that better? Right now it looks like it favours SB over a lot of 1366 platform just based on the performance, but if it really aint that in the real world, things must be adjusted in the test - in my opinion.
But if Randal is right and the cards don't even utilize more than x8 lanes, the performance gain is not needed exept for amount of memory on a 1366 platform.
I guess It all depends on how and what he (my customer) will be using Premier Pro for, I have absolutely no experience with using the software - I will get my hands on after easter when my customer has bought it. I am just a builder trying to get a grip on what i must take into considderation when building a system like this.
I am however leaning towards 1366, Even if 2600K performs a lot better per core, the 6 cores on an 970 will have an impact in some programs that have good use for 6 cores. I really look forward to learn more about the software.
But the question still remain - Is the PPBM5 Benchmark really showing wich system is good or not? If the SB platform is so much worse than a 970 on a Nehalem platform, the benchmark dont reflect that fact - maybe Randal is right and the number of PCI lanes doesent mean so much...I cannot tell by the benchmark, there are to many 2600K on the toplist to say its a "less performing" system.
That system scored 80 seconds in the AVI Disk I/O test (this is due to RAID 5 instead of RAID 3)
I disagree with your conclusion "this is due to RAID 5 instead of RAID 3", and it is my PPBM5 score that you are quoting; please consider:
1) Here are the PPBM disk I/O scores from my newer (JES3) rig:
- 12 WD 1TB 7200 rpm drives configured as RAID 0 - score = 66 seconds (testing, benchmarking, learning)
- 8 WD 1TB 7200 rpm drives configured in RAID 5 - score = 68 seconds (current configuration for this multi-purpose home PC)
So, would RAID 3 help me much? I don't think so.
2) CPU speed (processor family and clock speed) do indeed impact the PPBM5 Disk I/O score; see:
"...CPU speed: important and surprisingly affects all 4 PPBM5 tests..."
Enjoy your posts!
Unfortunately there is allot of incorrect assumptions about HDD configurations. Those have already directed people to purchase configurations that have done little to the performance or have actually caused further complications or headaches and at significant cost. The main consideration with HDD configurations is reaching the optimal transfer rate for the codec of material that you are dealing with. Once you are beyond that then the performance gain is minimal to non existent. So once you meet the optimal drive consideration for the codecs that you deal with, the configuration should be based around what kind of data redundancy you require in partnership with a backup strategy. The standard rule of thumb is any data that costs you or the company significant loss needs 2 data recovery/fail safe points minimum. So if you have the data on the work drives in the system, you would want atleast 2 points of recovery from there should those drives fail. Redundant Raid arrays give you a single level of fail safe but the operational time is the primary reason for raid arrays beyond performance. The second fail safe would be an external backup or image to pull from if the raid fails. Now if you are running a raid 0 for performance, then you would want 2 external backup/image points to pull from since the raid 0 does not count as any protection. If you maintain those fail safe points then raid 0 has little downside compared to the performance gain until you exceed 3 drives. At that point the risk of drive failure from 1 of those drives is high enough with the current drive failure rate that you want a redundant raid volume instead to protect the array. That is the point you compare the performance/redundancy pro's and cons for Raid levels 3, 5, 6, and 10.
The summary and application of all of this is that a 2 drive raid 0 of mechanical drives or a single SSD gives you the optimal transfer rate for AVCHD, XDCam, H264, Canon XF, HDV, most Pro Res versions, DVCPro HD, AVCIntra, M-Jpeg, most DNxHD versions, and 1080P Cineform. R3D, Cineform Raw, Pro Res 444, and uncompressed will utilize greater transfer rates depending on the amount of layers. Beyond that your drive considerations are and should be based around redundancy/backup points. That is why the SB2600 platform is more than enough for 50% of workflows since the raid array requirements are really based on data protection more than performance. If you are configuring large raids for performance with any of the codecs listed first then you are throwing away money. Consider carefully what your data protection requirements should be for your requirements and then decide whether large raid arrays make sense financially or not. That will then lead you to the platform required for your configurations.
The information in your reply was really helpfull, thanks a lot for sharing this with me and others that find it usefull.
I am gona stick with the SB platform. Next up is running the benchmark on it. And when it comes to hdd/backup/setup, I think this "Little Big Disk" will have it's impact in the nearest future: http://www.lacie.com/no/products/product.htm?id=10549
Agian, thanks for your help, all of you!