Sidebar Sidebar
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

Does the PCIE slot make a difference?

A

aslowdodge

macrumors member
Original poster
Nov 12, 2012
94
9
my 4.1 to 5.1 now has 2x 3.46 and a saphire pulse radeon 580 8gb , 32 gb of ram.
The GPU is on the bottom slot and my samsung 970pro as boot is not in slot 2, but slot 3 until I get spacers. I thought slot 2 was faster than slot 3, but the results were :

Screen Shot 2019-01-22 at 5.18.33 PM.png


My understanding is this is about as fast as it will get without spending another $400. But is moving it to slot 2 going to make a difference?
 
aslowdodge said:
my 4.1 to 5.1 now has 2x 3.46 and a saphire pulse radeon 580 8gb , 32 gb of ram.
The GPU is on the bottom slot and my samsung 970pro as boot is not in slot 2, but slot 3 until I get spacers. I thought slot 2 was faster than slot 3, but the results were :

View attachment 816974

My understanding is this is about as fast as it will get without spending another $400. But is moving it to slot 2 going to make a difference?
Click to expand...

Won't make any difference in your case unless you go for something like Highpoint 7101A
 
With the HighPoint I would see speeds close to double like 3000. Would that make much of a difference that I would see in usage for photoshop and Lightroom?
 
aslowdodge said:
With the HighPoint I would see speeds close to double like 3000. Would that make much of a difference that I would see in usage for photoshop and Lightroom?
Click to expand...

Shouldn’t make much difference. Unless your primary workflow is just keep importing very large pictures.

E.g. you have a 200MB picture file. Finish loading in 0.1s or 0.2s should make much difference to the whole workflow (in PS). Once the picture is loaded. The storage speed doesn’t matter anymore (assume you have enough RAM).

And if you have a 30GB files (for LR). Finish loading in 10s or 20s also shouldn’t make huge difference for the entire workflow. I expect you won’t finish editing a 30GB project in just few minutes. If the whole project cost you 1 hour, then it’s you to decide if that 10s saving is significant.

In most case, once you moved to 1500MB/s NVMe storage. The bottleneck is not there anymore. Further improve the storage speed may provide little to no benefit.

Even importing can be CPU bottlenecked if your storage speed is already fast enough. E.g. I don’t really think you can really finish importing 30GB files into a project in 10s (even with the Highpoint card). If the actual time require is noticeably more than 10s, which means there are something bottlenecking the NVMe, so, further improve the NVMe bandwidth should not makes any difference.

Of course, PS / LR can use the NVMe as scratch disk as well. But again, that can be CPU bottlenecked. Unless you can now clearly identify the storage bandwidth is still the bottleneck (e.g. constant 1500MB/s data transfer in activity monitor). Otherwise, go to 3000MB/s shouldn’t make much difference.
 
100% agree with the above. In real-world scenarios, at ~1500MB/s this is not the bottleneck for PS or LR on MP5,1. Lots of RAM helps tremendously. I'm using NVMe on PX1 as system drive and it helps with immediate startup for many apps, but it's a negligible difference for PS. Really helps with other Adobe apps with plugins at startup (AE, PPro, Media Encoder, Audition, etc) more than it does for PS.

If you have a free PCIe slot, adding another NVMe on PX1 may do more. You can fully keep scratch drive(s) separate from system drive. This is not possible for many people who use these machines. For many, putting 2+ SATA SSDs in a RAID0 is about the only available (cost conscious) solution to do something similar, but I will caution RAID in Mojave is a mess. It may "work" but it is implemented poorly and has issues. Doubt these were resolved in latest update, but great if they are. These RAID issues do not present in High Sierra.

The advantage of the SYBA SI-PEX40129 or High Point SSD7101A that is not as frequently discussed - they allow mounting of additional NVMe blades and do not necessarily require a RAID0 setup. (Each SSD shows on its own.) There would be a bottleneck or maximum data rate through a single PCIe slot, but using the x16 slot (above GPU) and adding additional fast SSD storage for a scratch disk on its own (separate from system SSD) would likely limit that bottleneck. This is probably easiest way to increase performance if additional PCIe slots are not available. If either drop in price, I'll be exploring that option.
 
thanks! So my original plan was

saphire pulse radeon 580 8gb slot 1 (the fastest)
samsung 970pro as boot in PX1 in slot 2,
sonet tempo with 6gb ssd in slot 3
usb3.0 card in slot 4

Seeing the PX1 block the fans on the GPU does it make more sense to move things around to
saphire pulse radeon 580 8gb slot 1 (the fastest)
usb3.0 card in slot 2, (very small would only block 1/3 of one fan instead of 1 and 1/2 fans on gpu)
samsung 970pro as boot in PX1 in slot 3
sonet tempo with 6gb ssd in slot 4

Any issues or sacrifices here?
 
Take a look at your USB3 card. My PX1 is in slot 2 above GPU and when testing with RX580 actually fits better. There is limited "stuff" on the bottom side and airflow prevention is limited to 1 fan and a very narrow portion. It's still mostly open. The USB3 card would actually block more.

Take a look at the thermal pad fix for RX580 that I use when the card is in my system. Removing the two screws really helps with the fit. The PX1 is even less of an issue than the Velocity Duo X2 was.

https://forums.macrumors.com/thread...-for-pcie-slot-2.2134562/page-2#post-26413743
 
  • Like
Reactions: h9826790
bsbeamer said:
Take a look at your USB3 card. My PX1 is in slot 2 above GPU and when testing with RX580 actually fits better. There is limited "stuff" on the bottom side and airflow prevention is limited to 1 fan and a very narrow portion. It's still mostly open. The USB3 card would actually block more.

Take a look at the thermal pad fix for RX580 that I use when the card is in my system. Removing the two screws really helps with the fit. The PX1 is even less of an issue than the Velocity Duo X2 was.

https://forums.macrumors.com/thread...-for-pcie-slot-2.2134562/page-2#post-26413743
Click to expand...

I ordered the thermal pads to do what you did, but just wondered if the tiny usb card being less than 1/3 the size of the PX1 would work better
 
Really depends on your USB3 card. Mine is 4-port and fairly full length wide (measuring from PCIe connector to metal access door). The PX1 is very narrow and even though it runs nearly the length of the RX580 (measuring along PCIe connectors), it does not block the same volume and seems to allow RX580 to breathe more. PX1 is much better than Velocity Solo X2 or Duo X2 above RX580.

I think more important is to use the card with the least amount of "stuff" on the bottom in the slot above RX580. The RX580 fans protrude out from the housing and will knock/hit if there are screws, chips, etc on the bottom of the card in slot 2. Removing the two screws and stacked thermal pads were my fix for working around that issue. It's a VERY tight fit when the RX580 is the installed GPU.

Some people are using cards right above the RX580 that nearly fully block/deflect the fans and appear to do so without issue, so either of these setups shouldn't be too much of a problem. The single fan setup/location on GTX 1080 FE is much better suited for MP5,1 with basically everything, but it is not Mojave-ready.
 
My 4 port usb is about a half inch wider than the PX1, but only about 25% as long as the PX1. I will try it to see if there are clearance issues, but my concern was I bottlenecking any of the cards
 
RX580 should be in x16 (slot 1)

PX1 is only x4
Tempo SSD Pro Plus is x4 (unsure if that is your model)
Most USB3 cards based on FL1100 (like Inateck KT4004) are x1
 
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back