View Full Version : i7 980x Now or Wait for Sandybridge?
Renat Zarbailov December 28th, 2010, 02:09 AM Hey folks,
I am curious what your thoughts are on whether or not to wait for sandybridge in the 1st quarter of 2011 or build an i7 980x system now...
I edit AVCHD 17mbps in Premiere CS5.
If going the i7 980x route now, I am considering the following components:
ASUS Rampage III Formula LGA 1366 Intel X58 SATA 6Gb/s USB 3.0 ATX Intel Motherboard ($300)
12GB RAM 3X4GB sticks (CORSAIR XMS3 12GB (3 x 4GB) 240-Pin DDR3 SDRAM DDR3 1333 Desktop Memory Model CMX12GX3M3A1333C9) ($170)
CORSAIR CMPSU-620HX ATX 12v v2.2 / eps 12v v 2.91 620Watt ul fcc power supply (I will actually take it out of my current system)
MSI N240GT-MD1G/D5 GeForce GT 240 1GB 128-bit GDDR5 PCI Express 2.0 x16 HDCP Ready Video Card (..and unlock it with a CUDA hack) ($90)
Haven't decided what C: SSD drive to go with, want to stick with 128GB, any suggestions which brand/model is best?
Haven't decided what CPU cooling to go with, I loved my previous build for it's silence, I used CPU COOLER SCYTHE USA|SCMNJ-1000 R.
Also, haven't decided what case to go with, I loved Antec case for its silence that I have once used to build a system a few years ago for a friend, I think it was called P6.
Any suggestions/comments?
Truly appreciate your input,
Happy upcoming New Years!!!
Randall Leong December 28th, 2010, 08:58 AM Hey folks,
I am curious what your thoughts are on whether or not to wait for sandybridge in the 1st quarter of 2011 or build an i7 980x system now...
I edit AVCHD 17mbps in Premiere CS5.
If going the i7 980x route now, I am considering the following components:
ASUS Rampage III Formula LGA 1366 Intel X58 SATA 6Gb/s USB 3.0 ATX Intel Motherboard ($300)
12GB RAM 3X4GB sticks (CORSAIR XMS3 12GB (3 x 4GB) 240-Pin DDR3 SDRAM DDR3 1333 Desktop Memory Model CMX12GX3M3A1333C9) ($170)
CORSAIR CMPSU-620HX ATX 12v v2.2 / eps 12v v 2.91 620Watt ul fcc power supply (I will actually take it out of my current system)
MSI N240GT-MD1G/D5 GeForce GT 240 1GB 128-bit GDDR5 PCI Express 2.0 x16 HDCP Ready Video Card (..and unlock it with a CUDA hack) ($90)
Haven't decided what C: SSD drive to go with, want to stick with 128GB, any suggestions which brand/model is best?
Haven't decided what CPU cooling to go with, I loved my previous build for it's silence, I used CPU COOLER SCYTHE USA|SCMNJ-1000 R.
Also, haven't decided what case to go with, I loved Antec case for its silence that I have once used to build a system a few years ago for a friend, I think it was called P6.
Any suggestions/comments?
Truly appreciate your input,
Happy upcoming New Years!!!
If you're going to use CS5, the i7-980x that you're planning to get is the better choice at the moment. The initial Sandy Bridge platform (Socket LGA 1155) will replace the LGA 1156 "mainstream" platform, and as a result will still be limited by the total number of PCIe lanes (24 theoretical maximum - 16 on the CPU, plus up to 8 on the P67 chipset's PCH), quad-core maximum, and only a dual-channel memory controller. If you must wait for Sandy Bridge, wait until the fourth quarter of 2011 when the LGA 2011 (and possibly its uniprocessor/dual-processor equivalent LGA 1356) CPUs get introduced.
In addition, the RAM that you selected is not the best choice, even at $180: It is only 9-9-9-24 at 1333 speed.
Third, you will need at least two or three hard drives in addition to the SSD: Adobe advises against running any of its Creative Suite programs on a system equipped with just a single drive (SSD or HDD) for everything. If you ignore this advice, your system may run slower than a Celeron system for video editing. (Remember, unlike SCSI or SAS, SATA is still only a half-duplex interface in which data can travel in only one direction at a time.)
Fourth, I do not recommend an Antec gaming case (such as the Nine Hundred series or the Twelve Hundred): Despite its stock cooling performance, the case is surprisingly cramped on the inside, especially from rear to front. This needlessly complicates the connection of internal SATA devices to the motherboard (especially those with SATA ports mounted along its front edge, which might get blocked by the fixed, nonremovable hard drive cage mounting bracket), and in case you want to upgrade to a higher-end GPU, you may find that the card to be a tight squeeze or not even fit inside the case at all. Plus, its stock fans are far from silent at their highest speed settings (which might be required for hot-running systems such as these).
Pete Bauer December 28th, 2010, 09:05 AM Your question is a tough one to answer since none of us has any experience or video benchmarking with Sandy Bridge. I'm not the most knowledgeable about comparing architectures, and would defer to our resident experts. But here are a couple considerations based only on internet reading:
- Sandy Bridge will initially be available only on LGA 1155, meaning only dual channel DDR3, and up to quad core processor, until at least Q3 2011. PPro loves RAM and is happy to have more cores. So whether the greater efficiency of the Sandy Bridge architecture will overcome a loaded 980X for editing or rendering in PPro is a question mark.
Either way, though, it looks like you'd be able to rebuild your system into LGA 1356 (triple channel mem) or LGA 2011 (which I haven't found much info on but is supposed to supercede 1356) plus a CPU with six or eight physical cores by late in 2011.
If you're ready to build and edit now, I'd say build a 950 (quad core) system -- parts are already readily available and the architecture is mature (but not end of life). It'll do a good job with AVCHD and you can save the $600 vs the 980X that you can use toward a Sandy Bridge upgrade toward the end of 2011. If you're not too cost-sensitive, then do stick with the 980X, realizing that in less than a year you might be wanting to replace it and the Mobo.
If you're not in a big hurry to get editing and really want to try out Sandy Bridge, we'll be eager to hear how it goes. Based on what (little) I understand of the architecture roadmap, though, a later upgrade toward the "top end" won't be particularly cheaper either way. Regardless, upgrading later from a Sandy Bridge or a 1366 system built now would require both a new mobo and processor (plus some more RAM to fill out the triple channel slots) just like a move from 1156 to 1366 does now.
EDIT: See also Randall's comments, posted while I typed.
Allan Tabilas December 28th, 2010, 11:51 PM cpus and motherboards are already on sale now in other countries
Intel Sandy Bridge CPUs and motherboards now on sale in Malaysia, what Consumer Electronics Show? -- Engadget (http://www.engadget.com/2010/12/20/intel-sandy-bridge-cpus-and-motherboards-on-sale-in-malaysia-wh/)
Since Sandy Bridge is just days away (US "official" release in first week of January in Vegas for CES show), I'd wait two weeks for the benchmarks to come out. Anandtech already had a preview from this past August comparing performance versus equivalent # of cores and similar clock speed at Anandtech (Sandy Bridge versus Core i7 980x [westmere], Core i7 880 [lynnfield])
The Sandy Bridge Preview - AnandTech :: Your Source for Hardware Analysis and News (http://www.anandtech.com/show/3871/the-sandy-bridge-preview-three-wins-in-a-row/8)
I have a Core i7 920 (socket 1366) myself, and also considering buying on the initial quad core Sandy Bridges like the Core i7 2600K, which clock-for-clock will be 10-30% faster than any equivalent clock speed on a Nehalem (1366) or Lynnfield (1156). From the Anandtech preview, the 980x is still faster in encoding benchmarks due to more cores. The 980x will be slower on an equivalent clock basis. Since you can probably get a Core i7 2600K, a P67 motherboard from Asus/Gigabyte/EVGA/MSI etc, and 16GB of ram (4 x 4GB sticks, unfortunately only dual channel) for the price of a single 980X cpu, the first Sandy Bridges seem like a smart choice.
Or you could wait until Q4 2011 for the 6 core or 8 core Sandy Bridges.
Randall Leong December 29th, 2010, 08:11 AM I forgot to mention that the initial Sandy Bridge platform will still be (relatively speaking) ill-suited to a hardware RAID card: The way most P67 motherboards are designed, most motherboard manufacturers will be eating up some of the PCH's lanes with an onboard USB 3.0 controller (this is possible because the P67 PCH's PCIe 2.0 lanes will now be full-bandwidth instead of the P55's half-bandwidth lanes), a PCIe-to-PCI bridge controller (for legacy PCI cards), an extra SATA or eSATA controller (to add legacy IDE/PATA drive support and two addtional SATA ports) and an onboard PCIe NIC from Realtek or Marvell (since the Intel PHY chip that does not need a PCIe connection remains expensive to motherboard manufacturers). That leaves only four available PCIe lanes on the P67 PCH - and most mobo manufacturers opt to include two or more PCIe x1 slots. That leaves only the CPU's integrated 16 PCIe lanes - but putting a hardware RAID controller that requires a PCIe x4 or x8 slot into a PCIe x16-length slot would drop the main PCIe graphics card slot's bandwidth from x16 to x8. As a result, the initial Sandy Bridge platform will still suffer from the very same problems that plague current LGA 1156/P55 platforms: not enough available (as opposed to total) PCIe lanes for serious video editing.
Why Intel still hasn't come up with a chipset that natively supports USB 3.0 is a mystery (although the P67 PCH does natively support two SATA 6 Gbps ports and four SATA 3 Gbps ports). But a chipset from either Intel or AMD that natively supports USB 3.0 will not arrive until Q3 2011 at the earliest.
Allan Tabilas December 29th, 2010, 10:06 AM Should you need a massive number of PCIe lanes (raid controllers like Randall mentioned, video cards, etc), MSI has their solution called the Big Bang Marshal -- 8 PCIe x16 slots (four of which are full x16 bandwidth). It's likely that other manufacturers will follow suit (Asus with the RoG line, Gigabyte, EVGA etc). It doesn't use NF200 chipsets for more PCIe lanes like say the EVGA SR-2 Classified mobo or Gigabyte X58A-UD9 [those mobos are for 1366 sockets), but uses the Lucid Hydra chipset for the massive lanes.
SemiAccurate :: MSI shows off its Big Bang Marshal board (http://semiaccurate.com/2010/12/28/msi-shows-its-big-bang-marshal-board/)
Randall Leong December 29th, 2010, 12:18 PM Should you need a massive number of PCIe lanes (raid controllers like Randall mentioned, video cards, etc), MSI has their solution called the Big Bang Marshal -- 8 PCIe x16 slots (four of which are full x16 bandwidth). It's likely that other manufacturers will follow suit (Asus with the RoG line, Gigabyte, EVGA etc). It doesn't use NF200 chipsets for more PCIe lanes like say the EVGA SR-2 Classified mobo or Gigabyte X58A-UD9 [those mobos are for 1366 sockets), but uses the Lucid Hydra chipset for the massive lanes.
SemiAccurate :: MSI shows off its Big Bang Marshal board (http://semiaccurate.com/2010/12/28/msi-shows-its-big-bang-marshal-board/)
The biggest problem with such motherboards is that most such boards will cost much more money than what most of the current X58 motherboards cost. These Lucid Hydra-equipped motherboards are expected to cost $600 or more, completely eliminating any cost advantage of Sandy Bridge versus current i7/X58 platforms.
Allan Tabilas December 29th, 2010, 12:41 PM Agreed, the MSI Big Bang Marshal is likely a $400+ mobo, or any other mobo that features 1) Lucid Hydra) or 2) one or multiple NF200 chips). Still, those are options should you need a massive number of PCIe lanes or expansion slots for say a high end raid card. A high end Areca raid card can exceed $1k (or a 980x for that matter).
Panagiotis Raris December 29th, 2010, 01:06 PM sandy bridge is the mid-grade replacement; NOT the 1366 replacement. that will be the 1356 chipset.
The i7 980X for now will still be editing king, and the X58 will still be the top choice for editing until it is supplanted by the (whatever it is, X68 etc) 1356 replacement.
Mark Johnson December 29th, 2010, 05:35 PM Why Intel still hasn't come up with a chipset that natively supports USB 3.0 is a mystery (although the P67 PCH does natively support two SATA 6 Gbps ports and four SATA 3 Gbps ports). But a chipset from either Intel or AMD that natively supports USB 3.0 will not arrive until Q3 2011 at the earliest.
Expect Intel to roll out Light Peak and skip USB 3.0.
Randall Leong December 29th, 2010, 06:48 PM Problem with LightPeak is, will the public be ready to shell out such expense for such proprietary technology when it is introduced? LightPeak might not be backwards compatible with the millions of existing USB devices, and might require the purchase of an entirely new input device that's priced far higher than current input devices cost. But what I stated is mostly speculative.
In any event, it is way too soon to tell whether Intel's going with a largely proprietary technology that will not be licensed at all to other companies or have its licensing costs so high that nobody wants to pay anywhere near that much for it (or cannot afford to) will flop (like Intel's flirtation with RDRAM and its early insistence of going its own way with 64-bit computing that is largely incompatible with existing x86 software) or succeed in the marketplace.
Panagiotis Raris December 29th, 2010, 10:38 PM agreed; lightpeak licensing will likely cost oodles compared to SATAIII or USB3 devices. funny that they have suddenly jumped into SSD's, and are dabbling in RAM as well.
chances are they will keep proprietary rights and choke their own potential, then 2-3 years later it will begin to take off.
for most users, SATAII is fast enough, and SSD's are at the extreme end. USB 3.0 will appeal (and does) to those on the higher end spectrum, same with SATA III and SSD's.
LightPeak on standard (platter) drives will be pretty damn useless for most though, and until SSD's drop a LOT in price, chances are LightPeak will start and stay on as a high end user alternative. Look at SSD's (esp RAID config SSD's) at the moment; heck even PCI-E SSD's are still somewhat experimental yet available. But if you care, and IF you can afford it, then you probably understand how to utilize it. MOST PC users do NOT.
case in point... I bought my folks a new 6 core AMD system for xmas, to be used as a home media center PC. replaced a quad core AMD which is now relegated to web surfing and a windows media center system for the office/bedroom area. Both PC's had single disk hard drives, 7200RPM SATA II. I wasnt there when they set it up (trapped in NYC) but i did get a call...
The FIRST thing my mother said was 'its just as slow as the old one' because she was transferring media onto a 5400RPM external USB 2.0 drive and it was maxing out at 32MB/sec. She was pissed because she thought it was doing something wrong, as it advertised higher USB 3.0 speeds, she even plugged the external into the correct USB 3.0 slot, and when she went to look for songs or movies it took just as long to seek/find/play as before, not knowing the bottleneck was the 5400RPM USB2.0 external.
AFTER i set up the RAID5 on both machines and the gigabit networking, again, she said they were both 'just as fast' as each other.
Now i did install a total of 10 hard drives in the two PC's (5 disk RAID5) and network them via their gigabit connections, so NOW the transfers are MUCH faster, but the point is, she, an average user/consumer, would not have NORMALLY utilized the potential of her system had i not set it up and shown her a few bits. She was used to seeing my personal PC boot in NO time and transfer files at 250MB/sec; i run SSD's and RAID0, the average user will stick with SATA II and 7200RPM single disks.
LightPeak, SSD's, SATAIII, USB3.0, and RAID are still beyond the average user; as such, they really do not mean much to the average user. As such, it is probably not a high priority until fast enough (read, SSD's) storage comes WAY down in price. Even SATAIII isnt much faster than SATAII in a platter drive.
As flash storage prices drop and capacity increases to the point where it is a viable alternative to the standard platter disk, and read/write speeds hike up close to those of PCI-E SSD's, THEN LightPeak will suddenly be the high end, and slowly trickle down.
For most people, again, they will NEVER saturate SATAI, let alone SATAII transfer rates without either going for an SSD or a RAID setup, and again, in that case they are no longer the 'average' user. LightPeak is 3-5 years away from economical viability, in my estimation. At the earliest.
On the other hand, i am extremely interested in a PCI-E x4 SSD... And doubling my 4x80GB disk 7200RPM 2.5" SATA II RAID0 setup into an 8 disk RAID0 or two 4-disk RAID0's... (i back up weekly to externals). Chances are ill bite the bullet and buy a pair of 120GB Corsair Force SSD's in RAID0 and enjoy 570MB/sec read and 550MB/sec write speeds, and donate the slow Intel X25M 80GB SSD's (currently in RAID0 to my folks or old machine.
Randall Leong December 29th, 2010, 11:45 PM case in point... I bought my folks a new 6 core AMD system for xmas, to be used as a home media center PC. replaced a quad core AMD which is now relegated to web surfing and a windows media center system for the office/bedroom area. Both PC's had single disk hard drives, 7200RPM SATA II. I wasnt there when they set it up (trapped in NYC) but i did get a call...
The FIRST thing my mother said was 'its just as slow as the old one' because she was transferring media onto a 5400RPM external USB 2.0 drive and it was maxing out at 32MB/sec. She was pissed because she thought it was doing something wrong, as it advertised higher USB 3.0 speeds, she even plugged the external into the correct USB 3.0 slot, and when she went to look for songs or movies it took just as long to seek/find/play as before, not knowing the bottleneck was the 5400RPM USB2.0 external.
This just shows to highlight that both the host controller and the bridge chipset in the device must be capable of USB 3.0 transfers. Otherwise, the USB standard of the slower device will be the limiting factor.
With that said, my recently acquired 5900RPM USB 3.0 external drive has a physical transfer speed that's just about as fast as the fastest of the current-generation 7200RPM SATA hard drives (when plugged into the correct USB 3.0 port). Only the relatively slow access times betray the external's 5900RPM heritage.
Renat Zarbailov December 30th, 2010, 12:10 PM If you're going to use CS5, the i7-980x that you're planning to get is the better choice at the moment. The initial Sandy Bridge platform (Socket LGA 1155) will replace the LGA 1156 "mainstream" platform, and as a result will still be limited by the total number of PCIe lanes (24 theoretical maximum - 16 on the CPU, plus up to 8 on the P67 chipset's PCH), quad-core maximum, and only a dual-channel memory controller. If you must wait for Sandy Bridge, wait until the fourth quarter of 2011 when the LGA 2011 (and possibly its uniprocessor/dual-processor equivalent LGA 1356) CPUs get introduced.
In addition, the RAM that you selected is not the best choice, even at $180: It is only 9-9-9-24 at 1333 speed.
Third, you will need at least two or three hard drives in addition to the SSD: Adobe advises against running any of its Creative Suite programs on a system equipped with just a single drive (SSD or HDD) for everything. If you ignore this advice, your system may run slower than a Celeron system for video editing. (Remember, unlike SCSI or SAS, SATA is still only a half-duplex interface in which data can travel in only one direction at a time.)
Fourth, I do not recommend an Antec gaming case (such as the Nine Hundred series or the Twelve Hundred): Despite its stock cooling performance, the case is surprisingly cramped on the inside, especially from rear to front. This needlessly complicates the connection of internal SATA devices to the motherboard (especially those with SATA ports mounted along its front edge, which might get blocked by the fixed, nonremovable hard drive cage mounting bracket), and in case you want to upgrade to a higher-end GPU, you may find that the card to be a tight squeeze or not even fit inside the case at all. Plus, its stock fans are far from silent at their highest speed settings (which might be required for hot-running systems such as these).
Thanks everyone for your kind input on this.
Randal, you mentioned about avoiding the RAM I selected for the config I listed, what RAM do you think is ideal for the mobo I chose?
Also, as far as RAID 0 for the CS5 edit drive, do you recommend going with a RAID card or using the chosen motherboard's RAID capability. If going with RAID card, which card is great to get that is valuable.
One more thing, can anyone recommend a silent case for this build?
Thanks again everyone!!!
Randall Leong December 30th, 2010, 01:55 PM Thanks everyone for your kind input on this.
Randal, you mentioned about avoiding the RAM I selected for the config I listed, what RAM do you think is ideal for the mobo I chose?
Also, as far as RAID 0 for the CS5 edit drive, do you recommend going with a RAID card or using the chosen motherboard's RAID capability. If going with RAID card, which card is great to get that is valuable.
One more thing, can anyone recommend a silent case for this build?
Thanks again everyone!!!
That RAM is okay, but you can fit in faster RAM for a little bit more money. A good example is Corsair's new Vengeance series RAM, which typically are rated at 1.5V VDIMM @ DDR3-1600 speed with 8-8-8 or 9-9-9 latency timings at 1600. (It's not that I'd recommend avoiding RAM rated at 9-9-9 at 1333 speed; it's just that you can get faster and/or lower-latency RAM for just a little more.)
As for RAID 0, if you're going for just a two-drive RAID 0, stick with the onboard Intel RAID.
Allan Tabilas January 3rd, 2011, 02:32 AM Benchmarks are out for the $320 Core i7 2600K (3.4Ghz, 4 cores / 8 threads) vs $1000 Core i7 980x (3.3Ghz, 6 cores / 8 threads)
Sandy Bridge 2600K , 2500K, and 2100
AnandTech.com - The Sandy Bridge Review: Intel Core i7-2600K, i5-2500K and Core i3-2100 Tested (http://www.anandtech.com/print/4083)
Sandy Bridge Core i7 2600K vs Core i7 980x (only) in multiple benchmarks (video, gaming, productivity, etc)
Bench - CPU - AnandTech :: Your Source for Hardware Analysis and News (http://www.anandtech.com/bench/Product/287?vs=142)
At least in regard to say the x264 2nd pass 3.03 benchmark, the Core i7 980x is almost 30% faster (transcoding a 720P 4Mbps MPEG-2 source to h.264)
In regard to 3D rendering, the Core i7 980x is 18% faster on the Cinebench R10 multi-threaded test. On the single-threaded test, the 2600K is 18% faster than the 980x.
I'll find another website review that uses, say Premiere Pro and its Mainconcept h.264 encoder in Adobe Media Encoder. Eventually somebody will post results on the PPBM5 benchmark site.
Randall Leong January 3rd, 2011, 08:11 PM It looks like even the i7-2600K is only at about the same level of performance in CS5 as an i7-950 (stock to stock). But the 2600K beats the 950 in everything else.
Allan Tabilas January 4th, 2011, 01:52 AM I only found one or two websites that benchmarks the new Sandy Bridge Core i7 2600K and Core i5 2500K with regard to Adobe Media Encoder or Premiere Pro CS5. The vast majority of reviews for PC hardware usually test either 1) x264 3.03 or earlier 2) Handbrake with some version of x264 3) WMV 4) Cinebench and maybe 3DS Max benchmarks.
Intel Core i7-2600K, Core i5-2500K and P67 / H67 Motherboards Review - Media Encoding (http://www.hardwareheaven.com/reviews/1091/pg8/intel-core-i7-2600k-core-i5-2500k-and-p67-h67-motherboards-review-media-encoding.html)
Video Encoding
Our first media encoding test uses the latest CS5 software from Adobe. This software accesses all of the cores/threads available on our processors and is therefore an ideal application for media enthusiasts. To test performance we run a 1920x1080 file (custom recording) through the encoder set to 1080p 24 HQ using the H.264 Blu-Ray codec within the application.
They tested with a Core i7 875K (2.93GHz , close to the Core i7 950 of 3.06GHz), which has a deficit of 16% stock clock speed from the 3.4GHz of the Core i7 2600K.
In Hardwareheaven.com's test, the 2600K had a time of 1m32s, and the 875K had a time of 1m43s, only 11 seconds difference, or about 12% difference (and this is before 875K clock speed deficit, which makes the 2600K result less impressive) --- But you have other benchmarks where the 2600K exceeds the 875K, 950, or 980x.
i7 2600K versus i7 950
Bench - CPU - AnandTech :: Your Source for Hardware Analysis and News (http://www.anandtech.com/bench/Product/287?vs=100)
What's more impressive is the Sandy Bridge mobile parts, including the 2nd from the top model, the Core i7-2820QM, which is supposed to be equivalent to the previous laptop CPU, the Core i7-820QM. The mobile i7 2820QM is actually FASTER than my desktop i7 920 in many of the benchmarks, which is crazy --- I'd have to check if the original laptop Core i7 720QM,740QM, or 820QM were faster than the Core 2 Quad 9xxx or Q6xxx desktop CPUs.
http://images.anandtech.com/graphs/graph4084/34964.png
AnandTech.com - Intel?s Sandy Bridge: Upheaval in the Mobile Landscape (http://www.anandtech.com/print/4084)
I'll see if I can find more benchmarks. Kitguru.net has a benchmark with Premiere CS5, but for some reason the benchmark graphic only includes the 2500K cpu, without comparing it to either last generations' Core i7 or even AMD processors. Weird.
Intel Core i7 2600K and Core i5 2500K review | KitGuru (http://www.kitguru.net/components/cpu/zardon/intel-core-i7-2600k-and-core-i5-2500k-review/16/)
I think the OP (Renat) will be happy with the Sandy Bridge 2600K at $320, as I'm not sure that the $1000 980x is worth 200% or 3x the price premium for likely less than 30% performance improvements with CS5 (and less parallel workloads will favor the IPC of Sandy Bridge). Even my old i7 920 handles Renat's workload of 17Mbit AVCHD pretty easily at stock 2.66Ghz speeds. And should Renat need something exotic like dual or triple full bandwidth PCIe 2.0 x16 slots, there will be P67 Sandy Bridge options from Gigabyte, MSI, Asus that can supplement it with either Nvidia's NF200 or LucidLogic's exotic option (whether X58, P55, or new hotness P67). Renat's original option of going X58 / 980x route included the Asus Rampage III board, which already has the "exotic" NF200, and is $300ish.
Future Sandy Bridge sockets
http://pc.watch.impress.co.jp/video/pcw/docs/360/112/6.pdf
For myself, I'll upgrade my i7 920 and old Gigabyte X58-UD5 motherboard (original launch date from 11/2008, no USB 3.0 or SATA 6G) to this. Maybe Q4 2011 will not only have the X58 replacement (called Socket B2 or LGA1356 versus LGA1366 Nehalems or Westmere) equivalent Sandy Bridge, but hopefully a six or eight core (hopefully) extreme edition compared to the current Core i7 980x.
Claire Buckley January 4th, 2011, 08:01 AM Hey folks,
CORSAIR CMPSU-620HX ATX 12v v2.2 / eps 12v v 2.91 620Watt ul fcc power supply (I will actually take it out of my current system)
And after you've added those HDDs and the other bits and bobs don't forget to up your PSU to at least 850 watts or more.
Steve Kalle January 6th, 2011, 03:12 AM The one important thing y'all are missing is the new 'Quick Sync' instruction set designed to dramatically increase video ENCODING speed. Check this out: Quick Sync Vs. APP Vs. CUDA : Intel?s Second-Gen Core CPUs: The Sandy Bridge Review (http://www.tomshardware.com/reviews/sandy-bridge-core-i7-2600k-core-i5-2500k,2833-5.html)
This makes CUDA look like a Celeron processor.
I wouldn't go running to get a new SB CPU but wait until Adobe CS6, which, I hope, should have Intel's new acceleration enabled.
However, for those who encode to multiple formats, I would consider building a PC with a new SB i7 just for encoding.
Pete Bauer January 6th, 2011, 08:02 AM I only skimmed the same Tom's article yesterday so correct me if I misunderstood, but from what I recall I'd agree with Steve that Quick Synch is of possible FUTURE benefit but won't help current pro NLEs. Don't have time to re-read the article this morning, but my recollection is that software has to be specifically coded to take advantage of Quick Synch (which current pro NLEs wouldn't be) and if you use a graphics card rather than the on-chip video, Quick Synch would be disabled along with the rest of the chip's video subsystem?
Also, I was left wondering if editing functions on the timeline would be able to access Quick Synch similarly to how effects can be accelerated by CUDA to give that delightfully smooth editing experience we can now enjoy. In other words, it isn't entirely clear quite yet that software access to the Quick Synch function is broad enough to do more than hard-coded renders. TBD, unless someone picked up additional tidbits about QS?
Quick Synch probably won't be of practical use to us for a while; maybe in a year or two, or maybe not. Sandy Bridge overall is a nice step forward in Intel's architecture but I think that for a while yet the i980X will still be the best single CPU desktop solution (at a handsome price). If you're in the i950 or lower budget range, the early benchmarking indicates that current Sandy Bridge setup might come close to competing without being a slam dunk.
I'd sure love to have a new Sandy Bridge to replace my old daily use computer, though!
Randall Leong January 11th, 2011, 07:23 PM I only skimmed the same Tom's article yesterday so correct me if I misunderstood, but from what I recall I'd agree with Steve that Quick Synch is of possible FUTURE benefit but won't help current pro NLEs. Don't have time to re-read the article this morning, but my recollection is that software has to be specifically coded to take advantage of Quick Synch (which current pro NLEs wouldn't be) and if you use a graphics card rather than the on-chip video, Quick Synch would be disabled along with the rest of the chip's video subsystem?
Also, I was left wondering if editing functions on the timeline would be able to access Quick Synch similarly to how effects can be accelerated by CUDA to give that delightfully smooth editing experience we can now enjoy. In other words, it isn't entirely clear quite yet that software access to the Quick Synch function is broad enough to do more than hard-coded renders. TBD, unless someone picked up additional tidbits about QS?
Quick Synch probably won't be of practical use to us for a while; maybe in a year or two, or maybe not. Sandy Bridge overall is a nice step forward in Intel's architecture but I think that for a while yet the i980X will still be the best single CPU desktop solution (at a handsome price). If you're in the i950 or lower budget range, the early benchmarking indicates that current Sandy Bridge setup might come close to competing without being a slam dunk.
I'd sure love to have a new Sandy Bridge to replace my old daily use computer, though!
The PPBM5 results list now includes two i7-2600K Sandy Bridge systems. A Sandy Bridge 2600K @ 3.9GHz with only 8GB of RAM and a more reasonably priced disk subsystem actually beat my overclocked i7-920 (3.7GHz) with 12GB of RAM on the PPBM5 list. Overclocking my 920 or 950 to 3.9GHz delivered very little performance gain versus the additional heat that such overclocking of such older-design CPUs produced. The config that's heavily overclocked (4.7GHz), equipped with 16GB of RAM and a hardware RAID card with eight SATA hard drives came out very highly on the list, if not at the top of the list - this despite the relative lack of total PCIe 2.0 full-bandwidth lanes in that platform (remember, the LGA 1155 platform effectively has only 20 PCIe 2.0 lanes total because four of the 24 theoretical maximum PCIe 2.0 lanes are taken up by a USB 3.0 controller, a PCIe-to-PCI bridge controller and an additional SATA controller). The trick to achieve this level of performance is to get a motherboard that has one PCIe x4 slot in addition to the PCIe x16 slot, and not to allow that PCIe x4 slot that the hardware RAID card is plugged into to drop to x1 bandwidth by installing anything into the PCIe x1 slot(s).
Steve Kalle January 11th, 2011, 07:48 PM The PPBM5 results list now includes two i7-2600K Sandy Bridge systems. A Sandy Bridge 2600K with only 8GB of RAM and a more reasonably priced disk subsystem actually beat my overclocked i7-950 on the PPBM5 list (both CPUs were at roughly equal MHz speeds).
I think too much emphasis is put on the PPBM results because they only test tasks most of us run overnight. However, it is a great resource/guide for PPro users and it does a fantastic job of explaining the results and how to apply these results to your own PC. I think a footnote should be added on the home page telling people that using a dual-core cpu, 2GB ram and 1 drive WILL cause many PROBLEMS in addition to installing tons of crapware. Too many people have user crap PCs and with crapware and then complain about it, which has given earlier Premiere versions a bad rap. Not to mention, faulty hardware such as video cards and ram are difficult to test because they can appear as very random issues. I bring this up because I have had both a bad stick of ram and just recently, a bad Quadro. The bad Quadro caused problems mainly with AE, which then caused me to think that Adobe made an unstable program. During the last week of using a new Quadro (got to love HP warranty-talked online at 4pm and had a new Quadro the next morning at 10am), I have not had a single problem with AE or any BSODs.
I think a great test would be testing the quickness of playback of various formats which should provide a good evaluation of both the processing power and disk subsystem for certain formats. Furthermore, if there was a simple way to test everyday editing tasks, that would be awesome. For fellow AE users and even Encore users, a Dynamic Link test/benchmark would be great. Plus, just AE benchmarks would be great because that program really uses system resources much more heavily than any other Adobe app. One easy test for AE is hitting zero '0' to render a comp of at least 20 seconds to see how quickly and how many of those 20 seconds the computer can cache in ram. Hitting zero is a very common task for AE users.
Adriano Moroni January 12th, 2011, 01:02 PM Hi, here we speak about quickness of i7 2600k but how much will it be stable for making video editing?
Allan Tabilas January 14th, 2011, 11:05 AM For those waiting or bought a Sandy Bridge cpu and are looking at RAID cards, the Tech Report did a quick test with a Marvell SATA controller on the PCIe 2.0 x1 slot (theoretical bandwidth 500MB/s) and a OCZ SSD. Unfortunately they did not test the x8 slot with an Areca raid card.
A quick look at chipset PCI Express performance - The Tech Report - Page 1 (http://techreport.com/articles.x/20241)
I have both a Core i7 920 and i7 2600K, so I'll try to run PPBM on it.
Randall Leong January 15th, 2011, 11:00 AM Unfortunately they did not test the x8 slot with an Areca raid card.
This is because the PCIe x8 slot on P67 motherboards shares bandwidth with the primary PCIe x16 slot that's used for primary graphics cards. Thus, if any card is plugged into that x8 slot, the primary PCIe x16 slot also drops down to x8 mode.
And keep in mind that the LGA 1155 Sandy Bridge platform has a maximum total of only 24 PCIe lanes. However, most motherboard brands (including Intel itself) eat up four or five of those 24 PCIe lanes with onboard USB 3.0 controllers, PCIe-to-PCI bridge controllers and extra SATA or eSATA controllers (and often an onboard PCIe NIC). That leaves only 19 or 20 PCIe lanes available for expansion cards.
Third-party repeater chips such as Nvidia's NF200 can add additional PCIe lanes to the LGA 1155 platform - but they fail to solve the limited total bandwidth of the LGA 1155 CPU's integrated PCIe controller. They effectively convert 16 PCIe 2.0 lanes into 32 PCIe lanes that are artificially restricted to PCIe 1.0 bandwidth.
Panagiotis Raris January 19th, 2011, 12:50 PM i am skipping sandy bridge and waiting for the X58/socket 1366 replacement. not worth the money at this point.
Randall Leong January 19th, 2011, 01:04 PM i am skipping sandy bridge and waiting for the X58/socket 1366 replacement. not worth the money at this point.
Actually, none of the other current CPU platforms are worth anywhere near their current street prices, for that matter. If anything, all of the other platforms provide much worse performance-per-dollar ratios than the Sandy Bridge at this point (at least at their current street prices). The i7-950's price, for example, seems to be the starting point for the higher-end platform - but then, I would have needed to spend more than $300 on a motherboard that can overclock the CPU above 3.6 or 3.7 GHz without requiring a massive increase in the CPU voltages. And at my maximum "safe" overclock for the entry-level LGA 1366 CPUs on such lower-priced (~$200) motherboards, the system still performs slower than even a slightly-overclocked Sandy Bridge CPU in most applications. That is exactly what makes the LGA 1366/X58 platform a relatively poor value if one is to buy or build a completely new system today. But then again, it is currently the only viable choice if one wants to add both a better sound card and a discrete hardware RAID controller card (the latter is required for LOVE, a/k/a "Laughing Over Video Editing", disk setups) into the same system without having to spend an astronomical amount of money. Do that with current Sandy Bridge platforms and either the disk performance or the MPE GPU performance will suffer.
In other words, the LGA 1366/X58 platform is still a waste of money if one is planning to stick with a KISS ("Keep It Stupidly Simple") disk setup for the foreseeable future. However, it is currently the only non-astronomically-expensive choice if you are going with a KISS setup with the intention of upgrading to a LOVE setup within the next several months (or whenever funds permit).
However, if you already own a Core 2 Quad or better, you might as well hold on to it for a little longer unless you really need an up-to-date system right away (most likely because your current system has been having serious stability issues that cannot be easily solved).
Panagiotis Raris January 19th, 2011, 07:53 PM i understand that; i bought my components out of desperation because my home PC, which was a 2007-ish Q6600 with 8GB of RAM died and it was more cost effective to upgrade. i only used it for light video editing, but mostly as a home entertainment system and web surfing; so it was more than adequate. also the sandy bridge stuff was not out yet. I should have explained that i meant in my position i am skipping sandy bridge.
IF i was in the market for a new home PC, i would go sandy bridge, same if one of the work machines died. as it stands, we will upgrade our work machines in about 8 months, hopefully to 8 core systems, and i will take one of the i7 980X systems for my personal PC.
For our business systems we have i7 980X machines; one with 12GB RAM the other with 24GB RAM. both overclocked to 4.1Ghz. My sig system is only a web surfing/home entertainment system that gets light video editing done on it. I think i paid $375 for the motherboard and processor shipped, and $180 for the ram shipped, and donated the Q6600 system case/PSU/RAM/processor/accessories (dead motherboard) to one of my editors who only had a core 2 duo laptop.
The i7 980X is HORRIBLY overpriced considering what sandy bridge is capable of; one could almost build an equal sandy bridge system for the $1000 pricetag. and the i7 970 is even worse; $800 and most cant seem to get it reliable over 4.0Ghz.
I LOVE your 'KISS' acronym by the way!
Randall Leong January 20th, 2011, 10:57 AM What I stated in my previous post clearly underscores the point that a PC that's used for serious video editing should not be used for anything else. You see, the older i7-920/930/950 is slower than even an i5-2500 for everyday applications when the CPUs are at their stock speeds - but then again, the i5-2500K runs at a stock (non-turbo) 3.3GHz versus 3.07GHz for the i7-950. What's more, a dedicated video editing PC should be semi-permanently disconnected from the Internet (this means connected to the Internet only if any part of that system needs software or firmware updates) since that PC normally should not have any antivirus or anti-malware software installed at all whatsoever, especially if that PC is normally connected directly to a broadband modem with no firewall whatsoever. This will minimize the risk of the dedicated editing PC being infested with malware and being hacked by outsiders.
For many users, the only solution is two separate PCs - one dedicated to video editing, the other for everything else.
Panagiotis Raris January 21st, 2011, 09:44 PM again, my machine (in my signature) is a dual purpose unit; one profile/boot partition for editing (with internet disabled, no installed anti-virus, and optimized configurations) the other for everything else.
As of yesterday, i have revised my hard drive setup and reinstalled everything (PITA). Again this is my home PC that i can and use for work.
Drive 0 - 2X Intel X25M in RAID 0 on Marvell SATA III, Windows 7 64 bit, CS5 suite and all software
Drive 1 - (intel ICH10R 2 disk RAID 0) Media/projects
Drive 2 - (intel ICH10R 2 disk RAID 0) Pagefile/Media Cache
Drive 3 - Preview files - OCZ Enyo USB 3.0 128GB
Drive 4 - Export files - OCZ Enyo USB 3.0 128GB
SATA II/III port list...
Marvell SATA III
RAID 0 configuration...
Port 0 - Intel X25M 80GB SSD
Port 1 - Intel X25M 80GB SSD
Intel ICH10R SATA II
RAID 0 configuration... Intel ICH10R
Port 1 - Fujitsu Lifebook 80GB 2.5" 7,200RPM HD 8MB Buffer
Port 2 - Fujitsu Lifebook 80GB 2.5" 7,200RPM HD 8MB Buffer
RAID 0 configuration... Intel ICH10R
Port 3 - Fujitsu Lifebook 80GB 2.5" 7,200RPM HD 8MB Buffer
Port 4 - Fujitsu Lifebook 80GB 2.5" 7,200RPM HD 8MB Buffer
CD/DVD/BD-ROM devices...Intel ICH10R
Port 5 - Pioneer BDR-206 Blu Ray burner
Port 6 - Pioneer BDR-206BK Blu Ray XL burner
I also have 5 1TB USB 2.0 drives as backups, one for each logical RAID 0 drive, so im not concerned about data loss or having to re-create a RAID 0 array. My girlfriend is also quite adept at computers, so its not a bother should we lose an array or have to reconfigure the machine. We also have two 2TB USB 2.0 drive for music, movies, etc as it does function as a media center. I just installed the BDR-206BK yesterday as well as the two USB 3.0 OCZ Enyo SSD's which are now the preview and export disks, respectively, so i will run PPBM for CS5 later tomorrow on the current i7 950 processor, and later on the i7 980X when i receive it from a friends' computer part-out. The work machines are P6T's with RAID controller cards; this is a personal machine.
It is perfectly stable at 4.3 Ghz whether used as a media centre or an editing machine, and thus far is just fine in all respects for 4-10 days of continuous use, which is the maximum time we have left it on since we built it in late November.
Scott Chichelli January 25th, 2011, 08:36 AM to the OP.
yes buy a sandy bridge now skip the X58 unless you need a raid controller card.
(more than 2 sets raid 0)
for $700 less you get the same performance of a 980x..
16 gig ram is more than plenty unless doing AE with PP open.
Scott
ADK
Randall Leong January 31st, 2011, 07:02 PM to the OP.
yes buy a sandy bridge now skip the X58 unless you need a raid controller card.
(more than 2 sets raid 0)
for $700 less you get the same performance of a 980x..
16 gig ram is more than plenty unless doing AE with PP open.
As of today (January 31), Intel has recalled the P67 and H67 chipsets due to a flaw that causes slow degradation of the performance and reliability of the four SATA 3.0 Gbps channels. (The two SATA 6.0 Gbps channels are unaffected.) Therefore, Intel has suspended shipment of the chipsets to its motherboard partners. The chipmaker will start shipping a "fixed" revision of the chipsets some time in February.
This is a warning to those who are considering a Sandy Bridge system.
Scott Chichelli February 3rd, 2011, 03:24 PM all boards have been removed from the channel replacement ones in about 4-6 weeks... well worth the wait.
Scott
ADK
Steve Kalle February 3rd, 2011, 03:43 PM to the OP.
yes buy a sandy bridge now skip the X58 unless you need a raid controller card.
(more than 2 sets raid 0)
for $700 less you get the same performance of a 980x..
16 gig ram is more than plenty unless doing AE with PP open.
Scott
ADK
Where are you getting your performance numbers from? The most improvement I have seen with the the chips is in the 15-25% range for non-video tasks and less for video encoding.
Scott Chichelli February 4th, 2011, 04:25 PM from in house testing...
Premiere Pro CS5 Version 5.0.3 Testing
Video material - AVCHD 1080P 24 Frame Each Cut to 30 minutes of material
Export Codec - H264 HDTV 1080P 24 Preset Default
4 Effects per Layer - Fast Color Corrector, Brightness & Contrast, Video Limiter, Sharpen
Each Layer Scaled to 50% for 4 frame PinP view.
I7 2600 3.4GHZ Turbo to 3.8GHz (stock speed)
8GB Blackline 1600 CL 9
460GTX
4 WD 1Tb Sata 64 Meg Cache 600 Drives in 2 Raid 0 arrays
3 Layer - 37:35
4 Layer - 40:49
16GB Blackline 1600 CL 9 (stock speed)
570GTX
3 Layer - 36:17
4 Layer - 40:05
I7 2600K 3.4GHZ Turbo to 4.7GHz (OCed)
16GB Blackline 1600 CL 9
470GTX
3 Layer - 31:35
4 Layer - 34:35
980X at 3.33GHz (stock speed)
24GB Blackline 1600 CL 9
470GTX
3 Layer - 37:51
4 layer - 41:06
I7 980X 4GHZ (OCed)
12GB Blackline 1600 CL 9
570GTX
3 Layer - 32:30
4 Layer - 35:25
Scott
ADK
Steve Kalle February 4th, 2011, 07:01 PM Your variables are all over the place. You go from 24gb ram to 12gb and then go from a 460gtx to a 570gtx to a 470gtx. And you compare 4.7Ghz to 4Ghz....
You use WD drives in Raid 0? I hope you don't send this to customers.
Randall Leong February 4th, 2011, 09:16 PM Your variables are all over the place. You go from 24gb ram to 12gb and then go from a 460gtx to a 570gtx to a 470gtx. And you compare 4.7Ghz to 4Ghz....
You use WD drives in Raid 0? I hope you don't send this to customers.
I agree that there are too many variables involved. However, I can compare the stock-speed i7-2600K with only 8GB of RAM and only a GTX 460 with the stock-speed i7-980X with 24GB of RAM and a GTX 470. As it turned out, the 980X is a tad slower - but that's primarily because its stock speeds are a bit lower than the 2600K. And that's despite the 2600K having much less RAM and a somewhat lesser GPU. In Turbo mode, the 980X runs at 3.46GHz with all six cores operational while the 2600K runs at 3.5GHz with all four cores operational. So, the i7-980X needs to be overclocked slightly just to beat a stock-speed i7-2600K, based on Scott's testing.
Scott Chichelli February 5th, 2011, 10:44 AM you guys cant read?
980X at 3.33GHz (stock speed)
24GB Blackline 1600 CL 9
470GTX
3 Layer - 37:51
4 layer - 41:06
16GB Blackline 1600 CL 9 (stock speed)
570GTX
3 Layer - 36:17
4 Layer - 40:05
both stock speed no ocing..
more ram in the 980x yet it lost
same on the oced versions...
$300 vs $1000
like i said $700 more to get the same performance.
the only downside is if you need a real raid array.(8+ drives) then you have to go to X58 or Xeon platform.
the stock 2600 beats all stock 900 series.
only the 970/980 can hang with it.
Scott
ADK
Scott Chichelli February 5th, 2011, 10:50 AM Your variables are all over the place. You go from 24gb ram to 12gb and then go from a 460gtx to a 570gtx to a 470gtx. And you compare 4.7Ghz to 4Ghz....
You use WD drives in Raid 0? I hope you don't send this to customers.
yes they are varied and for a reason...
note the 460 with 8 gig ram vs the 470 with 16gig ram.. barely a difference..
yes 4.7 vs 4GHz thats the point a $300 processor can OC to 4.7GHz whilst a $1000 processor can hit only 4 ghz. (which does speak to the power of 6 cores vs 4)
imagine if we hade a 6 core sandy bridge! like the ones coming in Dec. this is just a preview... these present sandy bridge chips are the "Budget line" ones...
cant wait til the end of the yr..
\
oh and yes the majority of systems we ship have 2 sets raid 0 with the WD 64meg cache drives
the higher end systems 8+ drives raid 5,6.
Scott
ADK
Steve Kalle February 5th, 2011, 11:07 AM So, you are telling me that a 570gtx has absolutely nothing to do with these numbers when the other is using a 470gtx? With 4 scaled layers and 4 effects per layer, the video card is involved quite a lot.
I'm assuming these tests did not have MRQ checked. Even though the output is not smaller (1080 to 1080), the files are being scaled so I would think that Adobe's bicubic implementation would still be used to scale each file and produce a better image. Frankly, without MRQ checked, any benchmarks are useless to me because the encoding is done at a LOWER quality than any professional would use. If you haven't seen the image quality difference, then do some tests. Adobe's MRQ (in a CUDA system) finally equals the quality of external encoding software such as TMPGEnc. Hell, even in broadcast SD analog. I can see a noticeable quality difference when I use MRQ and when I don't use it.
Don't you know about the major problems with WD drives in Raid and them spinning down constantly? Randall can speak to this issue and how changing drives actually increased both his workflow and encoding speeds.
Btw, even with the high-k gate in the Intel CPUs, such high speeds will certainly reduce the life of these CPUs as far more electrons are stripped off. When Intel releases a CPU above 4Ghz, then we will know that they have solved this problem.
Scott Chichelli February 5th, 2011, 11:51 AM Steve,
you are mistaking the TLER issue in a parity drive arry with standard raid 0
and the tler issue can and will occur in a parity raid with seagate and any other brand that is not an "enterprise" drive. the enterpirse drives have a different firmware.
the spin downs and tler issues do not occur in raid 0, 1 or 10
give me a little credit dude... i have been doing this a very long time. (98)
We have to warrnty these systems i ship as well. not to mention our rep.
as to the 470 vs 570 depending on what you are doing will determine the performance gain.
realistically the 470 would suite most. the 570 is quieter and is what the 400 series was supposed to be originally. just more cores not a architectural change.
Scott
ADK
Scott Chichelli February 5th, 2011, 11:58 AM Btw, even with the high-k gate in the Intel CPUs, such high speeds will certainly reduce the life of these CPUs as far more electrons are stripped off. When Intel releases a CPU above 4Ghz, then we will know that they have solved this problem.
i have also been OCing since 98. thats actually what got me started in this crazy biz.
for the last 5 yrs intel silcon has been near 100%. (anything after the "pres-hot" (prescot)
Intel has a safety buffer of 25% or better.
3.8GHz (turbo) x 25% = 4.75GHz.
i sell 4.5GHz to be sure.
but again those not into OCing... stock vs stock SB still wins.. :-)
albeit not by much.
Steven Davis February 5th, 2011, 12:13 PM I was going to pull the trigger on a machine and went to my wishlist and saw the sandybridge i7 was out of stock. Well that was fun, now I have to rebuild the whole thing. I'd like to get the machine before March, and don't know if the Sandybridge is worth waiting. I don't mind buying the newest thing if it's worth it over what is already available.
I'll have to do something thinking. I however need a machine with a punch.
Scott Chichelli February 5th, 2011, 12:29 PM well you dont have to wait you can buy the 980x on the X58 platform. its is a more robust platform.
but you will pay more!
FYI it will be more like mid march is my guess.
Intel is shipipng out test chipsets in about 10-14 days followed by full shiping 1-2 weeks later
then the manufacturers have to build and ship to the US.
Scott
Steve Kalle February 5th, 2011, 12:33 PM Steve,
you are mistaking the TLER issue in a parity drive arry with standard raid 0
and the tler issue can and will occur in a parity raid with seagate and any other brand that is not an "enterprise" drive. the enterpirse drives have a different firmware.
the spin downs and tler issues do not occur in raid 0, 1 or 10
Scott
ADK
TLER stands for Time Limited Error Recovery which means a drive reports as bad after 7 seconds. This has nothing to do with Raid 5. I have personally seen a WD drive in Raid 0 without TLER on an Intel ICH8r lockup a PC and cause data loss even though the drive ended up being fine. WD used to allow people to enable TLER in desktop drives but quickly changed that about 2 yrs ago. When they changed the firmware, they also enabled constant spinning down which causes problems with any drive, enterprise and desktop. I tested a similar spin down with my Areca 1680ix (using the Areca software) and Seagate Constellation ES drives and it acted just like regular WD drives. This is a PITA for video editing as the drives constantly spin down.
This non-TLER issue can happen with any Raid array including 1 & 10.
Then, there is also the staggered spinup issue. If WD drives have been on a hardware controller and then moved to the onboard controller which has no staggered spinup, the drives will not work because they are waiting for a staggered signal.
One more issue is their warranty - retailers show 5yrs but most people see 3yrs when doing an RMA thru WD.
FYI, anyone using Raptors, TLER is DISABLED. This drive is what I referred to as causing data loss in a Raid 0 array I had a few years back.
Steven Davis February 5th, 2011, 01:08 PM Well, I am researching this one
Newegg.com - Intel Core i7-950 Bloomfield 3.06GHz 4 x 256KB L2 Cache 8MB L3 Cache LGA 1366 130W Quad-Core Processor BX80601950 (http://www.newegg.com/Product/Product.aspx?Item=19-115-211&SortField=0&SummaryType=0&Pagesize=10&PurchaseMark=&SelectedRating=-1&VideoOnlyMark=False&VendorMark=&IsFeedbackTab=true&Page=2#scrollFullInfo)
Scott Chichelli February 5th, 2011, 01:37 PM yes Steve,
i am well aware of both potential issues.
we have never seen a spin up issue. who in their right mind would go from a serious raid 5,6 array back to onboard?
to give you an idea of how large a base i am coming from 400+ drives a month.. (not hard with minimum 5 per system up to 9 or more on bigger ones)
i am not a big fan of the raptors thruput is not much better than regular. seek times is their only benefit.
and for editing its moot. audio it helps a tad for samples
here is a comparison..
Intel SAS RS2PI080 8 Port controller with 512 DDR2 Ram
8x WD VelociRaptor 600 GB, SATA 6 Gb/s, 32 MB Cache, 10,000 RPM
8 Drive Raid 5 - 745MB/s read 735MB/s Write
8x WD 1TB SATA 6 Gb/s, 64 MB Cache, 7200 RPM
8 Drive Raid 5 - 703.8MB/s read 670MB/s Write
both are some serious thruput...
FYI wd waranty is 5yrs.. you have to be careful where you buy drives. Sams, costco, staples etc have different versions. often the warrnty is only 1 yr. buyer beware.
all oem versions are 5 yr. like what i sell as well as newegg etc.
other retail box versions are 5 as well..
Scott
Scott Chichelli February 5th, 2011, 01:38 PM Well, I am researching this one
Newegg.com - Intel Core i7-950 Bloomfield 3.06GHz 4 x 256KB L2 Cache 8MB L3 Cache LGA 1366 130W Quad-Core Processor BX80601950 (http://www.newegg.com/Product/Product.aspx?Item=19-115-211&SortField=0&SummaryType=0&Pagesize=10&PurchaseMark=&SelectedRating=-1&VideoOnlyMark=False&VendorMark=&IsFeedbackTab=true&Page=2#scrollFullInfo)
the X58/950 combo is a great platform.. you could OC the 950 to 3.8GHz and get basically the SB performance.
or wait
Scott
Steve Kalle February 5th, 2011, 02:06 PM yes Steve,
i am well aware of both potential issues.
we have never seen a spin up issue. who in their right mind would go from a serious raid 5,6 array back to onboard?
to give you an idea of how large a base i am coming from 400+ drives a month.. (not hard with minimum 5 per system up to 9 or more on bigger ones)
i am not a big fan of the raptors thruput is not much better than regular. seek times is their only benefit.
and for editing its moot. audio it helps a tad for samples
here is a comparison..
Intel SAS RS2PI080 8 Port controller with 512 DDR2 Ram
8x WD VelociRaptor 600 GB, SATA 6 Gb/s, 32 MB Cache, 10,000 RPM
8 Drive Raid 5 - 745MB/s read 735MB/s Write
8x WD 1TB SATA 6 Gb/s, 64 MB Cache, 7200 RPM
8 Drive Raid 5 - 703.8MB/s read 670MB/s Write
both are some serious thruput...
FYI wd waranty is 5yrs.. you have to be careful where you buy drives. Sams, costco, staples etc have different versions. often the warrnty is only 1 yr. buyer beware.
all oem versions are 5 yr. like what i sell as well as newegg etc.
other retail box versions are 5 as well..
Scott
WD's diagnostic software can't run on a drive in raid. If you have a problem with a drive, you must hook it up to the onboard sata.
Google the warranty issue to see what I am reading. People who buy a WD Black should have a 5yr warranty but many people have rma'd them and when doing so, WD's site shows only a 3yr warranty.
I don't know how you haven't seen the spin down issue. If you are constantly testing and using the drives, then you won't see it. I have seen it first hand and so have thousands of others.
|
|