Precision 7530 & Precision 7730 owner's thread

Checked it. Same behaviour. So it's definitely a laptop issue. Motherboard exchange it is... :-/

 

I tried 18361 on my M6700 late last week. I immediately ran into an issue where, if I closed the laptop screen/lid, and then opened it again, about 75% of the time it would show a black screen — with the backlight on — instead of the lock screen or whatever I was previously working on. (The laptop is configured to "do nothing" when the lid is shut.) I could recover by putting the system to sleep and waking it again, or using Win+Ctrl+Shift+B to reset the GPU driver. I reverted back to 17763 and the problem went away. I'm not sure if it is the Intel GPU driver or what. I did try reinstalling the GPU driver before reverting. Have you experienced this on either of your systems?

(I always can't help myself and attempt an upgrade when a build that appears to be RTM shows up... and then I always end up running into some stupid issue and regretting it.)

 

Just a file note regarding a successful RAM upgrade for anyone contemplating similar. Have a 7530 delivered with 32GB reported as 2 sticks of Samsung M471A2K43CB1-CTD, upgraded to 64GB with Crucial kit CT2K16G4SFD8266 (2x16GB), works fine.

 

Such a person is typing on my keyboard - but how good it feels when it does work and you do not need to restore the system

7730 connected to TB18DC with two external displays and the internal display: it just works.

Closed and opened the lid 5 times (should be good for 75% testing) with "do nothing" option: the internal display gets disabled and enabled without problems.

But, I do not have your combination of iGPU and dGPU and my displays have a lower resolution (1920x1080 + 1920x1200 + 2560x1440) than yours.

The m6800 display stayed black on the second try (Win+Ctrl+Shift+B worked as well). The next 5 tries worked. Just internal display, "do nothing" option, power plugged. So I see your "sometimes issue" there.

I had problems with my m6800 with AMD FirePro m6100 and Intel GPU to cause black screen issues from Win 8.1 to 10 1509 and to the next 10 build. With all the hassle to just keep a working Intel/AMD driver combination and Win 10 thinking to know it better... Switching from AMD to Nvidia GPU in the m6800 worked much better. I hear/read no issues for the current Intel and AMD GPU combination so they might have got it smoothed out.

This time with 7730 and only Intel GPU (just need CPU + RAM + storage for dev and VM stuff) I (still) see no problems.

 

I had only tried it with my M6700 and not my 7530 (stopped considering it after I ran into an issue). So in my case the problem was experienced with a single 1080p display.

Anyway, well this gives me hope that it isn't just some peculiar issue with my setup; if it happens to other users (even if less often) then I'm sure Microsoft's feedback channels will blow up as this rolls out and it will get fixed. I'll be waiting a couple of months before I try it again...

 

I just ordered a 7730 - i9, P3200, 64GB, 4K, 256GB m.2.

I plan to order four 2TB m.2 drives. I'm hope I can just remove the 256GB drive, clone it externally to a new stick, and keep it for backup.

Question is: Should I leave the new m.2s as separate drives, combine them into a Raid 0, or have redundancy with a Raid 5? How bad of a speed hit would I be taking by using Raid?

My usage is just web browsing and (a lot of) downloading - this 7730 is overkill for my needs. My current laptop has 1TB 2.5" SSD boot drive and a 2TB 5400rpm storage drive. Both drives are usually 90% full. And I typically download and copy 100GB from the SSD to the mechanical drive every day. I'm aware SSD drives don't like being full, but I'm a hoarder...

I have a Synology 4-bay Raid appliance already. I like knowing that if a drive goes bad, I can simply replace it and rebuild my files. But, how would that work in a laptop? If one of my SSDs goes bad, how do I boot the computer to rebuild the array with a new m.2 stick?

I've narrowed down my m.2 choices to the Intel 660p or the WD Blue. The 660p slows down when the cache is full (I don't move 100GB all at one time) or the drive itself is too full, and it has a terrible lifespan as regards TB written. The WD has a good lifespan, but it is a SATA m.2, not NVME, and therefore much slower.

Any input would be appreciated. I want to make the drives last at least four years. Somehow, I feel like a Raid array would help by spreading out the writing evenly across each drive. But, if that's a correct assumption, would the speed be affected too heavily? At least in a Raid setup, I wouldn't be moving 100GB from one drive to another anymore.

 

Theoretically, RAID 0 or RAID 5 would be faster, not slower. If you have 100 GB to write to an array, 50 GB will have to be written to each drive and it can happen with both drives simultaneously, which means the write will complete in half the time. Reading is similar. However, these systems there seems to be a bandwidth bottleneck and it can't actually saturate the pipe for two fast NVMe drives. (It's actually an Intel chipset issue with how they allocate the PCIe lanes, there is more details about it a few pages back, but it applies to most current-gen systems and not just these Precisions.) So performance will be roughly the same whether you choose to use RAID or not, it's just about the convenience of having a larger array that appears to be a single drive. SSDs are much less susciptible to failure than HDDs but you should have your important data backed up elsewhere regardless, RAID or not (RAID is not a backup).

Up to you to choose a drive, look at reviews and make a decision that fits your budget. I've found the Samsung EVO/Pro line to be quite reliable so I always get one of those when given the choice. If you look at the TB-written specs for these disks and then imagine how many GB's per day you would have to throw around to reach that within a few years, I don't think that you'll have any trouble having a modern SSD last a few years. Example: 1TB Samsung 860 EVO (SATA) and 970 EVO (NVMe) both have 600 TBW endurance, that means writing 420 GB (almost half of the total capacity?) every day in order to hit it in four years. Stress tests have shown that these drives usually outlast their TBW estimate.

I chose an 860 EVO for secondary storage in my 7530, it is cheaper as a SATA drive but the speed (less important for secondary storage) is still great, miles above a traditional spinning disk.

 

I have 1 x 860 Pro 2 TB and 3 x 970 Evo 2 TB for 8 TB total. Mine came with a 512 GB Toshiba drive (Class 40). The read or write speed of any one NVMe drive will max out the CPU-to-chipset bandwidth - so it doesn't really matter how you stripe, raid or whatever (I literally had the same thoughts as you did, and then I benchmarked all these configurations and observed how little difference it made).

I very well might convert this to a 4-drive stripe because the chances of one of these drives going bad are really low to begin with, and it simplifies my backup strategy - I just have a single 8 TB drive to deal with, and all writes will be distributed across four drives. I am just using Storage Spaces stripes, and not using the built-in RAID, so I'm not sure if this has built-in RAID 5 or not. If the built in RAID allows for RAID 5, then one drive could fail and it would still work - with RAID 5, it doesn't really matter which drive fails, the rest will tolerate and run a little slower until you replace the bad drive, and then the rest of the drives will recreate the contents of the replaced drive

Personally, I think the 4 x WD Blue drives might be a good bet for this. $250 ea. for 2 TB is not bad so you'd spend $1K for a laptop storage array that is good for 2 Petabytes written (500 TBW each). Read and Write speeds would be like 500MB/s x 4 = 2GB/s for reads and writes for the whole pool, and it's not going to get vastly better than that no matter what kind of drives you had - my 970 EVO's are honestly probably a little overkill for the hardware they are running on.

Edit: It does appear RAID 5 is an option from the BIOS, when I checked just now. Oddly, one of my 970 EVOs wasn't listed, but there could be various reasons for that.

 

Just a followup - I think I've found my "peak configuration" for this machine, after much trial and error.

As someone mentioned earlier in this forum, many a page ago, Intel RAID is limited to 3, and only 3, devices. What really works is to get a higher-end USB key - mine is a Samsung 128GB - and use Rufus to make a Windows To Go image. You can download an ISO from the Windows Media Creator tool, and Rufus will take a while but make a functional image, especially now that Microsoft rolled back a driver that was causing Win To Go blue screens. Remember to "Troubleshoot Activation" after booting from the USB so that it upgrades to Win 10 Pro for Workstations.

Then, drop into Windows Storage Spaces and a spawn a new 8TB stripe, in this case ReFS as I deal a lot with Virtual Machines. Then, I install GeForce Experience to get the latest drivers - or, I believe Aaron44126 described an alternate way to get the latest video drivers. The rest is pretty straightforward.

 

So, I've bought and installed this 32 GB 3 GHz CL16 kit. Works perfectly, although Task Manager reports the base JEDEC speed. HWINFO and CPU-Z both detect the XMP profiles installed and correctly report the single data-rate speed of 1501 MHz: