Hi guys,
I’ve just ordered the final parts for my new (and fairly ludicrous) 5,1 build, which will include a Cubic Desktop Xpander (which expands one PCIe x16 slot into four). And I’m wanting to pick people’s brains on the ideal layout of the PCIe cards in order to maximise bandwidth across the lanes when working in Davinci Resolve.
The full build will include the following parts:
My primary objectives would be to:
Basically this amounts to whether it’s better to keep the Cache Drive internally on a x16 Slot to maximise write speed and whether that would cost much GPU performance (having both GPUs in the Xpander) over keeping one of the Radeon VIIs internally on Slot 1?
The USB3.1 card will basically only be used to transfer data to the internal 6x 3.5” SATA RAID, so that shouldn’t be using up any bandwidth when I’m actually working (which makes me think it’s a good candidate for living in the Cubix).
I don’t know how much bandwidth the Boot Drive is likely to be using when I’m working (since all of the Media will be read from the 6x 3.5” SATA RAID, and all of the writing will happen to the NVME RAID Cache Drive. But I’m guessing it could potentially affect throughput to the Cache Drive or GPUs (if they’re in the Xpander at the same time). But would having it feed into the x16 slot (via the Xpander) allow it to run at higher speeds (compared to its 1500MB/s limit in the internal x4 slots) even with the other demanding components in the Cubix at the same time?
I’d appreciate any thoughts or suggestions anyone could offer, as expanding PCIe slots is a new game to me.
Here are the four basic layouts I’m currently pondering the efficacy of:
Cheers
I’ve just ordered the final parts for my new (and fairly ludicrous) 5,1 build, which will include a Cubic Desktop Xpander (which expands one PCIe x16 slot into four). And I’m wanting to pick people’s brains on the ideal layout of the PCIe cards in order to maximise bandwidth across the lanes when working in Davinci Resolve.
The full build will include the following parts:
- 12-core 3.46Ghz CPU
- 96GB 1333Mhz RAM
- 2x Radeon VII GPUs
- Media Drive: 6x 6TB Seagate Ironwolf 3.5” SATA Drives in RAID 10 (internally)
- Boot Drive: 512GB Samsung 970 Pro NVME (on 4x PCIe 3.0 card)
- Cache Drive: 2x 2TB Samsung 960 Pro NVME RAID 0 (on 8x PCIe 3.0 Card)
- Blackmagic Decklink Mini Monitor 4k
- Sonnet Allegro Pro USB 3.1 4-port USB Type-A Card
My primary objectives would be to:
- Maximise the processing throughput of the two Radeon VIIs
- Maximise write speeds to the 2x 2TB NVME Cache Drive
- Maximise the overall speed of the 512GB NVME Boot Drive
Basically this amounts to whether it’s better to keep the Cache Drive internally on a x16 Slot to maximise write speed and whether that would cost much GPU performance (having both GPUs in the Xpander) over keeping one of the Radeon VIIs internally on Slot 1?
The USB3.1 card will basically only be used to transfer data to the internal 6x 3.5” SATA RAID, so that shouldn’t be using up any bandwidth when I’m actually working (which makes me think it’s a good candidate for living in the Cubix).
I don’t know how much bandwidth the Boot Drive is likely to be using when I’m working (since all of the Media will be read from the 6x 3.5” SATA RAID, and all of the writing will happen to the NVME RAID Cache Drive. But I’m guessing it could potentially affect throughput to the Cache Drive or GPUs (if they’re in the Xpander at the same time). But would having it feed into the x16 slot (via the Xpander) allow it to run at higher speeds (compared to its 1500MB/s limit in the internal x4 slots) even with the other demanding components in the Cubix at the same time?
I’d appreciate any thoughts or suggestions anyone could offer, as expanding PCIe slots is a new game to me.
Here are the four basic layouts I’m currently pondering the efficacy of:
Cheers