new Dell Precision 5510 (Twin of XPS15)

The understanding I got when I investigated TB3-attached GPUs a few months ago was, in this case, the discrete GPU would still be used to drive apps on the integrated display. Displays attached directly to the external GPU would be driven by the external GPU (and in a "docked" configuration this might be fine if you don't use the laptop screen). I think the reason for this is just the way the drivers are set up, and probably in the future we could get some option that would allow a user to choose which GPU was used for Optimus, but that's not how things work right now.

One more thing to remember is that bandwidth over TB3 is 40 Gbps, but bandwidth directly over PCIe 4.0 is 256 Gbps (16 lanes at 16 Gbps each), so an external-attached GPU won't be able to perform as well as an internal one.

 

Yep, that's an important point. Since we're talking about the 5510 here, the question then is: Given the bandwidth limitations, will that external GPU perform significantly faster than the built-in M1000M, which is no slouch, either? Obviously, if the answer is negative, then there's no point in using these.

 

non-ECC bus: 64 bit; dual-channel mode: 128bit. ECC bus: 72 bit; dual-channel mode: 144 bit.

That's 8 bits more, because ECC memory needs to store 1 extra bit for every 8 bits. This extra bit is needed to check if the 8 bits aren't corrupted. So, the busses on the motherboard, between the CPU socket and the DIMM slots need to be wider in order to transfer the ECC bits. ECC DIMMs will work on non-ECC motherboards; then the ECC is simply turned-off. The bandwidth remains the same.

 

Yes, way too expensive, compared to the Dell 5510. On the other hand: Dell screwed up the keyboard. But the MBPr 15 also lacks specific Home, Page Up, Page Down, End keys.

 

I as looking at the zbook studio as well, but I think those arrow keys make Dell's look full size by comparison. Seems with all these keyboards you trade one feature for another.

Yes, which is why ECC DIMMS have odd number of chips. As to bus size, how is that implemented if the DIMMs have the same number of pins? Especially if current Xeon can use both ECC and non ECC?

It should be pointed out that these errors are quite rare to begin with and probably (likely) would not adversely affect the majority of users. And if it did persist it's likely more due to faulty or failing memory chips...

 

There could be pins that are present but unused in non-ECC modules...

 

If the HP keys are the same size as previous ZBooks, then they're basically the same size as the Dell ones...but the left and right arrows are double-size (top to bottom) vs. the Dell ones.

 

The Dell ones are 2/3 size. The ones on the zbook studiohave full size left and right arrow, but 1/2 size up/down arrow.

 

Given how those 3rd party external graphics adapters operating through the PCI-e x1 connections provide a decent show (with e.g. only 10-20% penalty in their performance), it would seem unexpected that Thunderbolt 3 would be the bottleneck here anytime soon.

 

I saw a tech preview of ASUS solution on one of the CES vids on YouTube and looked like really good performance numbers. Not clear if ASUS solution is proprietary. But having an upgradeable desktop grade solution when docked is pretty cool indeed.