We have just received the NA250A-GPU and I have a question about the cards installation layout in my Mac Pro as well as the NA250A-GPU. The following are the details of my system and PCIe cards. Please advise what would be the best configuration of cards installation in both devices in order to achieve the best performance.
- Mac Pro early 2009
- El Capitan OSX 10.11.5
- GT 120 for UI
- GTX 780 for Cuda processing
- Titan X for Cuda processing
- Red Rocket card
- Black magic decklink card
- USB 3 card
Answer (Last update July 1, 2016):
Please see the following for the suggested cards arrangement in both Mac Pro and NA250A-GPU.
Since you are using GTX 780 and TITAN X for CUDA processing, we would suggest you to visit http://www.nvidia.com/object/mac-driver-archive.html and download/ install the latest Mac CUDA driver for the two GPU cards that working with Mac OS X 10.11.5.
|Mac Pro (Early 2009), OS X 10.11.5|
|Slot||PCIe card model|
|4th PCIe slot (x4 signal)||USB 3.0 card|
|3rd PCIe slot (x4 signal)||Blackmagic DeckLink card|
|2nd PCIe slot (x16 signal)||NA250A host card|
|1st PCIe slot (x16 signal)||Nvidia GT 120 graphics card (for GUI)|
|Slot||PCIe card model|
|SLOT2||Nvidia GTX 780 GPU card|
|SLOT3||Nvidia TITAN X GPU card|
|SLOT4||Red Rocket card|
We specialize in Forensic Computers for the Military and law enforcement and is currently looking to purchase the NA255A-XGPU. The followings are the PC specs and requirements:
- Motherboard: ASUS X99-E WS/USB 3.1 Socket 2011-v
- CPU: i7-5820K 3.3GHz
- RAM: 32GB DRR4 2133MHz
- PCIe Slots: 7 x PCIe 3.0/2.0 x16 (single x16 or dual x16/x16 or triple x16/x16/x16 or quad x16/x16/x16/x16 or seven x16/x8/x8/x8/x8/x8/x8)
- Windows 10 Pro 64-Bit
- Motherboard: ASUS Z10PE-D8 WS Socket 2011-v3
- CPU: 2x Intel Xeon E5-2630v4
- RAM: 64-128GB DRR4
- PCIe Slots: 7x PCIe 3.0
- Windows 10 Pro 64-Bit
GPUs to be installed in the NA255A-XGPU are 4 x GTX 980 or 4 x GTX1080.
Please help to verify if the NA255A-XGPU is able to support 4 x GTX 980 or 4 x GTX1080 simultaneously with the computer specs provided. Besides, if it is possible to get an upgrade to 1500W power supply instead of 1200W, or can we purchase without a power supply?
Answer (Last update July 8, 2016) :
For your reference, we had provided a real case with a similar scenario as yours:
A customer installed four Nvidia GTX 980 GPU cards in his NA255A. The host computer he uses is Asus X99-E WS motherboard installed with Intel Core i7-5960X CPU; he installed eight 8GB DDR4 RAM modules on the motherboard, so totally there’re 64GB RAM on his Asus X99-E WS. The customer reports the four GTX 980 cards in NA255A work properly with Asus X99-E WS.
We have confirmed the above setup from the customer is workable. However, you mentioned the CPU you used for the Asus X99-E WS/USB 3.1 motherboard is i7-5820K. But from our success case, the CPU used on Asus X99-E WS is a higher end model of i7-5960X.
Please note that the max number of PCIe lanes for i7-5820K is only 28 lanes, but the max number of PCIe lanes for i7-5960X is 40 lanes.
Furthermore, your RAM for Asus X99-E WS/USB 3.1 motherboard is 32GB DDR4, but, from our success case the RAM is 64GB DDR4. The size of RAM on the motherboard may also have an influence on the number of GPU cards workable with motherboard.
Therefore, four GTX 980 GPU cards in NA255A should also work with your Asus X99-E WS/USB 3.1 motherboard. But most importantly, the CPU model and size of RAM must be considered since the i7-5820K CPU only has 28 lanes and motherboard only has 32GB in size.
As we mentioned, the ASUS X99-E WS can support up to 4 GPUs; however, the Z10PE-D8 WS can only recognize up to 2 GPUs. We suggest to choose the GTX-1080 because this GPU requires less power and offers better performance.
Currently we don’t have the 1500W power supply option available. And due to its elaborate design, we don’t recommend using any different power supply other than the one included in the case.
I’m interested in purchasing the NA211A-GPU PCIe expansion chassis for my 2009 Mac Pro computer. I need it specifically for the DaVinci Resolve software, which relies heavily on GPU Cuda core processing. I plan to use the chassis for a second processing card, as well as a Red Rocket card. Is there any reason why the expansion chassis would not work with Resolve?
Answer (Last update May 23, 2016):
The DaVinci Resolve definitely works with the NA211A-GPU PCIe expansion chassis. However, you would not be able to install both the GPU card and a Red Rocket into the NA211A-GPU at the same time as the NA211A-GPU has only one open type PCIe x8 slot available for the PCIe x16 card. We would suggest that you install the GPU card and a PCIe x4 card (moved from your Mac Pro) in the NA211A-GPU, and keep the Red Rocket card in your Mac Pro for best performance.
Please take note that either one of the following combinations of PCIe cards would work in the NA211A-GPU.
1 x GPU card
1 x low speed PCIe x4 card
1 x Red Rocket
1 x low speed PCIe x4 card
We would recommend the NA250A-GPU (/product/pcie-expansion-enclosure-na250a/) should you need to install two or more PCIe x16 cards at the same time.
How to convert the NA211A to Thunderbolt PCIe expansion enclosure, NA211TB?
Answer (Last Update April 16, 2013):
The following video link demo how to convert the NA211A to Thunderbolt PCIe expansion enclosure, NA211TB.
The description states that PCIe 3.0 x4 slots will be available when upstream port working under PCIe 3.0 x8 mode. Does this mean that if I am running at PCI-e 3.0 16x mode I will not have access to the PCIe x4 connectors?
Answer (Last update April 16, 2013):
If the upstream is running under PCIe 3.0 x16 mode then the two PCIe x4 connectors will not be available. The maximum bandwidth that host interface can support is up to 128 Gbps external PCIe 3.0 x16 between host and NA265A.
Does NA221A-NB (Laptop Package) work with Lenovo T530 Laptops running Windows 7 64-bit?
Answer (Last update Jan 4, 2013):
NA221A-NB works with laptops that come with ExpressCard 34/ 54 slot. For older laptops that support only PCIe 1.0 ×1 spec, the bandwidth is up to 2.5 Gbps (2.5 Gbps from 1.0 × one lane ＝ 2.5 Gbps) when NA221A-NB is connected with the laptop. However, for newer laptops with Intel Sandy Bridge micro-architecture (w/ PCIe 2.0 ×1 spec), the bandwidth will be 5 Gbps (5 Gbps from 2.0 × one lane ＝ 5 Gbps) when you connect NA221A-NB to the laptop.
For our holographic products, I’m interested in using your product for hosting multiple GPUs outside of a host PC box, but I’d also like to use your Host Adapter card to connect a custom external backplane or FPGA to a host PC. Is that possible?
I’m essentially looking for the fastest way I get data off of a GPU and into an externally located FPGA. It seems to me that your Host Adapter would do the trick. However for the particular application I have in mind, the NA255A-GPU box is not required. I would rather connect your Host Adapter card directly to my FPGA via the external PCI-e cable OR I could create a custom backplane that my FPGA card sits in. In this case, I would just need to know how the Host Adapter interacts with the upstream port. Or another possibility would be whether you manufacture smaller backplanes than the one in the NA255A-GPU chassis.
Answer (Last update Jan 3, 2013):
Our pin assignment for the upstream port and the Host Adapter are following the standard specs from the PCI associate as shown onhttp://www.pcisig.com/home.
I believe you are referring to the application as the following image and the example at (http://www.dinigroup.com/new/DN2076k10.php). We also have a smaller 3 slots board that is available in PCIe 2.0 version only.
I’m looking to use NA255A-GPU for heavy multi-GPU computational purposes. Is the Host Adapter PCIe card that plugs into the host PC a simple “data pass through” layer? Does it add any additional latency when transferring large amounts of data between GPUs and host system memory? Is the upstream port slot inside of the NA255A box a special slot? Or is it a standard PCI-E slot? If I were to connect the Host Adapter card to a custom backplane not manufactured by you guys, say a backplane with a custom FPGA, does that custom backplane and FPGA still get recognized by the Host PC’s bios? I only ask because:
1. The entire system is very attractive for what we do here at Zebra. We do a lot of GPU intensive computations and your product’s ability to give us 4 GPUs in a single box is very attractive.
2. We have custom FPGAs and backplanes with both our static and dynamic holographic display products. We currently connect to these using much lower bandwidth protocols (ex: ethernet and USB). I’m curious if it would be possible to use your Host Adapter card to connect directly to our custom FGPA cards or backplanes so they are directly accessible to the host PC.
Answer (Last update Jan 3, 2013) :
The NA255A is a standard multiple PCIe slots expansion enclosure. It only require a free PCIe x16 slot from the host PC and doesn’t need any software installation for the enclosure itself. Of course, if the third-party’s PCIe cards that to be installed into NA255A require driver installation then the driver installation procedure is just the same as you plug the PCIe card into your host PC. So, the NA255A is a “data pass through” PCIe slot expansion. Just imaging the NA255A let your single PCIe slot of host PC to be multiple PCIe slots.
The upstream slot is pre-installed with a target PCIe adapter in order to connect with the host adapter board of the host PC. If the FPGA is also a standard PCIe interface that can be installed into a standard PCIe slot of the host PC then it can also be installed into any expansion slots of NA255A as well but not in the slot of upstream port.
What is the bandwidth provided when TurboBox NA210A is connected with a laptop or a desktop PC?
Answer 1 (Last update Dec 2, 2010)
The TurboBox NA210A is PCIe 2.0 spec, providing throughput 5 Gbps per lane.
For desktop PC users, since both host card and target card are PCIe 2.0 ×4 spec, when TurboBox NA210A is connected with desktop PC, the bandwidth provided is 20 Gbps (5 Gbps × 4 lanes ＝ 20 Gbps).
For laptop users, since current technology of laptop supports only PCIe 1.0 ×1 spec, when TurboBox NA210A is connected with laptop, the bandwidth is 2.5 Gbps (2.5 Gbps from PCIe 1.0 × one lane ＝ 2.5 Gbps). However, after the release of laptop with PCIe 2.0 ×1 spec (Intel Sandy Bridge microarchitecture) in the first quarter of 2011, when TurboBox NA210A is connected with laptop, the bandwidth will be 5 Gbps (5 Gbps from PCIe 2.0 × one lane ＝ 5 Gbps).
If three third-party PCIe cards are plugged on TurboBox NA210A, will the bandwidth provided between TurboBox NA210A and laptop/desktop PC be enough for the three PCIe cards?
Answer 2 (Last update Nov 17, 2010)
Yes, the bandwidth for those three PCIe cards will be enough. An example of using an ATTO 8 port external 6Gb/s SAS/SATA RAID controller card and two Avid Pro Tools HD3 cards is given in the following:
For desktop PC users, when eight SATA hard disk drives are used with ATTO 8 port external 6Gb/s SAS/SATA RAID controller card, the RAID controller card consumes only 500 to 600 MB/s bandwidth which is approximately 5 to 6 Gbps. As for the other two Avid Pro Tools HD3 cards, their bandwidth is around 2 Gbps; therefore, the bandwidth for the three PCIe cards between TurboBox NA210A and desktop PC is about 7 Gbps, which accounts only one-third of the total bandwidth 20 Gbps that is provided.
For laptop users, since current technology of laptop supports only PCIe 1.0 ×1 spec, when TurboBox NA210A is connected with laptop, the bandwidth is 2.5 Gbps (2.5 Gbps from PCIe 1.0 × one lane ＝ 2.5 Gbps). When an ATTO 8 port external 6Gb/s SAS/SATA RAID controller card and two Avid Pro Tools HD3 cards are used on TurboBox NA210A, they consume about 7 Gbps bandwidth. In this circumstance, an ATTO 8 port external 6Gb/s SAS/SATA RAID controller card and two Avid Pro Tools HD3 cards can still be used; just the performance of the three PCIe cards would be limited due to the PCIe 1.0 ×1 spec on laptop; however, PCIe 2.0 ×1 spec (Intel Sandy Bridge microarchitecture) on laptop will be released in the first quarter of 2011, and the bandwidth will be 5 Gbps then (5 Gbps from PCIe 2.0 × one lane ＝ 5 Gbps). At that time, the performance for the three PCIe cards will be improved.
How do I know the third party PCIe cards plugged on TurboBox NA210A are recognized by computer?
Answer 3 (Last update 2010)
For users of Windows operating system, please check “Device Manager” from Computer Management. As for users of Mac OS X, please check “About This Mac” under the Apple icon to see if the third party PCIe cards are recognized by computer.
For detailed information about the verification of third party PCIe cards, please check the user manual and see the ninth section “Verify Installation” at the third page of the user manual.
After I connect the TurboBox NA210A to laptop/desktop PC, the TurboBox can not be recognized by computer. What can I do to solve this problem?
Answer 4 (Last update 2010)
If the TurboBox NA210A is not recognized by computer, it can be BIOS’ problem. Please update your laptop/desktop PC to the latest BIOS from the Web site of your laptop/desktop PC manufacturer. After updating the laptop/desktop PC to the latest BIOS, the TurboBox can be recognized by computer.
Wondering if your Model# NA210A can support a KONA 3 and/or Red Rocket card(s)?
I want to use this with a MacBook Pro.
Answer 5 (Last update Jan 24, 2011)
Only MacBook Pro(Unibody Model ID 5,1) has problem on its ExpressCard interface compatibility; others models of MacBook Pro are working well with NA210A.
The spec. of Red Rocket card(s) is no problem working with NA210A but the spec. of KONA seems to ask for PCIe x4 bandwidth. Since Laptop is only provide PCIe x1 bandwidth so when KONA 3 works with a Laptop then I suppose it cannot reach full performance. SD will work but I cannot make sure on HD.
My system is: Macpro (8-core) 2×2.8 Quad intel Xeon. I am using Protools Accel cards a total of three PCie cards are already installed and a video card a total of 4 slots. I need to use the video card inside the expansion box so i can use the slot for the host card. I will be using a UAD card inside the expansion box and one video card. Would this be possible?
Answer 6 (Last update Jan 26, 2011)
Yes, NA210A is no problem working both Video and Audio cards and the bandwidth from DT will be fairly enough for handle those two cards together. UAD’s engineer has tested NA210A with their 2pcs of UAD-2 Duo and the result is workin well. UAS will announce this news after they finish their big project of new products.
I’m running a mac pro 2.66ghz quad core intel nehalem processor with 16gb of 1066mhz ddr3 ram. The mac OS is 10.6.6. I want to run the NA210A-L (DT) loaded with two UAD 2 quad cards and a Sonnet tango express card, or quite possibly, three of the UAD2 quad cards.
Answer 7 (Last update Jan 26, 2011)
Yes, it is workable. UAD’s engineer has tested NA210A-L(DT) with 3* UAD2 Duo cards and the result will be announced on UAD’s website soon.
What is the bandwidth provided when NA220A is connected with a laptop or a desktop PC?
Answer 1 (Last update Oct 4, 2010)
The NA220A comes with PCIe 2.0, providing throughput of 5 Gbps per lane. For desktop PC users, since both host card and target card are PCIe 2.0 x4 spec, when NA220A is connected with desktop PC, the bandwidth provided is 20 Gbps (5 Gbps x 4 lanes ＝ 20 Gbps).
The older version of laptops only support PCIe 1.0 x1 spec, when NA220A is connected with laptop, the bandwidth is 2.5 Gbps (2.5 Gbps from PCIe 1.0 x one lane ＝ 2.5 Gbps). However, after the release of laptop with PCIe 2.0 x1 spec (Intel Sandy Bridge micro architecture) in the first quarter of 2011, when NA220A is connected with laptop, the bandwidth will be 5 Gbps (5 Gbps from PCIe 2.0 x one lane ＝ 5 Gbps).
Does the NA220A support the: UAD-2 QUAD DSP Accelerator?
Answer 2 (Last update Mar 9, 2011)
Yes, NA220A works with UAD-2 QUAD. Moreover, UAD-2 QUAD only consumes 10W , and the NA220A is providing 80W output power which is sufficient.