*** Official Clevo P65xSA/SE/SG / Sager NP8650/51/52 Owner´s Lounge ***

I go for the Crucial M500 M.2 240GB, i think. I hope this HDD (500GB) from my old laptop wil fit in.
And maybe later i will exchange the HDD for a SSD

 

Which is where I was when I ordered my Mythlogic Dia 1614. Do I just get ...
A) (2) 512GB Crucial m550 M.2?
or ...
B) (1) 512GB Crucial m550 M.2 + (1) 512GB XP941 M.2-PCIe?

In the end, I actually went for C ...
C) (1) 512GB Crucial m550 M.2, leaving the one (1) M.2-PCIe unoccupied

Why? I decided that I'd wait for more SATA PCIe options to come out, and for NVMe to mature**

**SIDE NOTE: I largely run Linux (yes, for work since '98) and the great thing is that various patches and updates come out for kernels to address more capabilities, especially when it comes to storage where a lot of "firsts" are in Linux because of industry needs. I've also dealt with a lot of NT internals, especially for boot and storage, since '93.

That's the "burst" throughput. The question is ... when do you actually get that?

In reality, most people are reading randomly most. The benefit of NAND over platter is that NAND random reads can easily break 100MBps, or at least 25MBps in the worst case, whereas platter is struggling to sustain just a few MBps because of all of the seeks. So in this day'n age one should really be using at least some NAND.

The Hybrid platters, aka SSHDs, used to provide 4 and then 8GB of SLC, and now have 8 or even 16GB of MLC. My favorite will still be the Seagate 750GB SSHD with 8GB SLC + 750GB 7200 spindle. When those could be had for $70, they were the best, single drive.

Furthermore, as many studies have shown, most of the time, it's the same, static binaries that get used over and over -- boot, OS, a few apps, etc... Just caching in NAND those 20GB of files improves performance 3-10x over platters. Hence why there is Intel Smart Response Technology (SRT), Device Mapper Cache (dm-cache), etc...**

In other words ... if you are really "cost conscience," try this ...
- Get a platter, and
- Get a very cheap (even refurbished) 30-60GB NAND 2.5" device

Setup Intel SRT (select chipsets) to use the NAND device and use the system for a few days. If you see major performance improvements in booting and running your common apps, then that might be enough. Prior to Intel SRT options, I used to tune SuperFetch in Windows, although it's not quite the same.**

However, it is 2014 now.

mSATA and M.2 devices of 256 and 512GB are getting inexpensive enough that you could just run on them, so you don't take up a 2.5" slot. Alternatively, a 128GB M.2 device is enough to contain Windows (C: drive), and you can put all programs on another platter (D: drive).

So you can save your money and wait for some new NAND PCIe devices to come out, bringing the prices down. You'll still have that M.2/PCIe slot open. That's what I did, although I did opt for the biggest 512GB m550 M.2 option to start.

-- bjs

**NOTE: Microsoft really screwed-the-pooch on ReadyBoost, and should have focused more on its SuperFetch components. Intel SRT is a much, much better solution, but is only supported on select chipsets (even though it's virtually a pure software solution).

 

Cisco78 -- what Dabeer is saying is that using mSATA with M.2 is actually a use of conflicting terms.

mSATA is a Mini-PCIe x1 slot (data and power) that has been modified to allow the SATA data traces (4 single + 3 ground).
M.2 is a different slot with power, SATA data traces and, optional, multi-channel PCIe traces.

So one would say M.2 SATA, not M.2 mSATA. It seems anal, but it's actually not.

Unless the 120/240/480GB m500 is significantly cheaper than the 128/256/512GB m550, I would opt for the Crucial m550 instead.

I also, strongly prefer MLC over TLC, and went on that dissection a bit ago. Simple physics is that the MLC will outlast the TLC 10x over, regardless of the logic that merely mitigates the end user seeing the constant errors that occur with any NAND device. So that means the 940 Evo, Ultra II, etc... are out of consideration for me as I'm looking at the 950 Pro, Extreme Pro, etc... instead.

I'm still debating whether to add a 2.5" 1TB m550 to the system to complement the M.2 512GB m550, or possibly go with a 950 Pro or Extreme Pro instead. Sometimes going with one vendor is not always ideal from a risk mitigation standpoint. I really need to look at the warranty terms, especially on the 950 Pro and Extreme Pro, and see if they also cover reduced performance, which will definitely occur over time.

From the looks of it, you can fit the following combinations in the P650xSE ...
- (2) 2.5" x 7mm
- (1) 2.5" x 9.5mm
- (1) 2.5" x 12mm
And possibly even ...
- (1) 2.5" x 15mm

And the following, additional combinations in the P65xSG ...
- (1) 2.5" x 9.5mm + (1) 2.5" x 7mm

If one considers the new crop of single platter, 2.5" x 5mm drives, let alone NAND options, there are probably even more options. E.g., if someone uses a 5mm NAND or platter device, depending on where the SATA card edges are located, it might be possible to use it with a 9.5mm high platter in even the P65xSE.

If the 2.5" drive is only 7mm high, you can probably use both. Don't discount the use of platter.

E.g., I still use a platter for swap file (Windows); swap, /tmp and /var (Linux), even when I disable swap and use tmpfs for /tmp (Linux). It's also good for an added, extra backup device.

 

Hmm is there anything wrong w/ current mSata drives? Figured it would be a good slot for the OS drive with a Samsung 850 pro 1tb, is there an advantage that 2.5" has over it?

M2 is a bit pricey for me still since 500gb is nearly the same cost as the 1tb pro series.

Actually I guess I might have misunderstood that, Dabeers were you just correcting him saying its not msata but M2 ( as in the difference ? )

This made more sense to me

Appreciate it B.j.

 

Maybe usefull for you, i found the configuration page for the 67x Clevo / P705 XMG

mySN.de | SCHENKER XMG | Schenker Technologies - XMG P705 PRO Gaming Notebook 49,3cm (17.3")

 

Sager NP8652 spotted: Sager Notebook :: Series Notebook :: NP8652

 

No matter how many times I see it, that is still one sexy chasis. Wish they had kept that w/ the p751zm

 

For me, I decided to get a 1TB 7200pm HDD and an m.2 128GB SSD. m.2's are just too new right now and with the temp issues I saw in the Plextor and Samsung, I decided to just get the cheap and small m.2 one for now (only $135) and wait for either a big sale for a 1 TB SSD 2.5" or when m.2's get better cooling.

 

Thank you for the clarification! Laptop newb here.

And whats the difference between the NP8651 & 2?

 

i am planning to buy a 67x maybe from xmg, as they where in my country

i am thinking of the 4710 with the gtx980, some ram, i already bought a samsung 950 1tb pro...
my old laptop is a alienware 17 r3... 1.5 times bigger and around 60% heavier...

 

The NP8652 is supposed to be 3mm thicker and has a 980M instead of a 970M, although the current page for NP8652 appears to just be a placeholder (after all, it's just a carbon copy of the NP8651 page), but hopefully this means Sager will be announcing it soon!

 

What a tease!

 

The ETA of the p650SG has been moved to december 9TH (PCspecialist's answer to my enquiry), let's hope that's just two preventive weeks ^^

 

Is there any place in the US that sells or will sell the P651 variants? The P650 looks fine, but I'm not a fan of the chrome on the back and I'd rather have a more fingerprint resistant sandblasted finish.

 

Probably a good choice, because according to the Clevo specs, you can't use both a PCIe drive and a SATA drive simultaneously. If you occupy the PCIe M.2 slot with a PCIe drive, you lose the ability to simultaneously use the SATA M.2 slot. You can, however, use two SATA drives simultaneously.

You'll probably know more than I about the reasons why this might be the case, or why it might not even be correct, of course

I agree, not that my opinion should matter for much

Correct. The problem here is that we're talking about two different things - the physical form factor, and the logical protocol used.

SATA can refer to the physical form factor or the logical protocol. In the case of 2.5" drives, it's both.

M.2 and mSATA are physical form factors. M.2 can support devices that communicate with either SATA or PCIe. Devices that are mSATA communicate using SATA. M.2 and mSATA are not interchangeable or compatible.

SATA (the logical protocol) has a limit of 600MB/s. PCIe v2 x2 (using 2 channels) has a limit of 1GB/s. PCIe v2 x4 (4 channels) has a limit of 2GB/s.

For M.2, PCIe drives are currently more expensive, and run hotter, than SATA drives. For normal usage, you probably won't really notice the difference in speed between a PCIe drive and a SATA drive, at least not the way you'd easily notice the difference between a SATA SSD and a SATA HDD.

Just make sure you get the 2.5" SATA version, and NOT the mSATA version - and while you're at it, get the 850 Pro instead

The P650SE/SG does not have an mSATA slot. So I guess the advantage would be that 2.5" you can use, mSATA you cannot

 

I did some digging. It appears they added it to their interwebsite along with the screenies in that video. Is 77 C a good score? seems ok to me but im not expert?

Eurocom Benchmarks

 

I know, right? ... *sigh*

 

oh cmon guys, the p751zm aint that bad looking rather....bat...looking....

 

that's what I'm asking for, it seems like clevo added support for PCIe booting, otherwise it will have to wait until HM97

 

Absolutely nothing at all. mSATA is a specification the ATA Consortium came up with after several OEM were creating non-standard Mini-PCIe (a PCIe x1 pin-out) pin-outs for SATA.

It was inevitable this would happen. In fact, there were several reasons why OEMs went there, and the ATA Consortium followed.

1) The lack of any supported storage stack (firmware, common OS boot loaders, common OS support, etc...) for direct use of NAND devices. Ergo, Intel's ONFI consortium had not address it yet. I.e., in the embedded space, we use uboot and setup the devices manually. This is fine for Linux, CE-based Windows, etc... It's not going to work for NT, common Linux distros installers (even if it can be retro-fixed for an existing install), etc...

2) The cost of adding the logic to support an existing, storage standard. E.g., like a PCIe-to-SATA controller on-card to make use of existing AHCI/SATA support (and existing NAND SATA support logic from 2.5" products).

3) The reality that chipsets usually had unused (unconnected) SATA ports in embedded, portable, etc... system designs. I.e., it was cheap to route the pin-outs through the Mini-PCIe connector.

It's all the same 6Gbps SATA logic, stack, etc... almost always going to the same chipset controller. So there's zero (0) difference other than size.

The P650x only has 2.5" SATA bays and M.2 slots, no mSATA slots. mSATA and M.2 slots are physically and electrically different.

mSATA uses Mini-PCI Express (Mini-PCIe) form-factor, which is 30mm wide with a 0.8mm pin-out. It adds SATA data lines to the existing, single channel PCI Express (PCIe x1) design.

M.2 is a new form-factor, which can be of a variable width and configuration from 12-30mm with a 0.5mm pin-out. It has several configurations, but the most common is 22mm with SATA and, optionally, PCIe x4 data plus power, typically in a 60-80mm length. Another configuration is a 22nm with SATA, PCIe x2, multiple-USB and PCM, typically in a 30mm length.

The main "advantage" of M.2 is not only the various configurations, but the removal of various logic on the card. I.e., Mini-PCIe (later used for mSATA) required a PCIe bridge or other end-device to support the technology used in the card, if it didn't speak PCIe directly.

M.2 introduces many different configuration pin-outs so existing peripheral, I/O and other technologies can be run to the card, and a bridge or other support to talk to the PCIe x1 channel not required.

One could say mSATA spawned the ideas behind M.2. There is even a reservation for a future memory specification that will likely bring costs down over the current M.2-PCIe configurations, because a bridge/controller to PCIe won't be required.

 

@bjs

have a look, xp941 doesn't boot on anything before z97

http://www.pugetsystems.com/labs/articles/Samsung-XP941-Plextor-PX-G256-M6e-M-2-Qualification-575/

 

I might have pulled the trigger on the P650SG, might have, but with its ETA of DEC 9, that's a LOL and a quarter.

Alfred is not picking up the phone, I've been calling non-stop

 

Even though I already replied to this, I want to add to BJ's answers above:

The only thing wrong with them is that they cannot be used in the P650SE.

... and the fact the mSATA cannot be used in the P650SE.

Sorry for sounding like a broken record on this point, but I want to drive the point home - I want to make sure everyone understands, and nobody wastes money by buying the wrong form factor.

 

just saw in the link that you linked that the gpu (970m) is throttling. its in 810mhz when it should be 910 ish. that aint nice :/

 

Which is why I was asking. Most specifically ...

Article: 'In fact, the only motherboards we could get to boot to it were ones with either a Z97 or H97 chipset. Interestingly, these are also currently the only chipsets that include native M.2 support ... Unless you have a Z97 or H97 motherboard, don't expect to use a Samsung XP941 as anything other than a storage drive.'

Now one thing to keep in mind here is this ... who creates the Storage Driver for Windows? It's the same thing with Intel Smart Response Technology (SRT), Intel purposely limited it to select chipsets like the X79, and 87+, even though it had nothing to do with the capabilities of the chipset.

Assuming uEFI, Intel provides the storage support so Microsoft's Bootmgr.efi/BOOTMGR can load the driver with the kernel for anything that Microsoft does not include.**

So that said ...

It's just PCIe pin-out. That's why it still works in Windows, as even PugetSystems discovered. Given that Windows uses signed drivers, let alone Intel does not release the source code for its Windows drivers, one is SoL there, if one doesn't have the chipsets Intel enables the solution for, when it comes to NT boot.

But under Linux, I'd really be surprised if it didn't work -- unless it doesn't work for even Z97. I.e., some GRUB2 driver required for the technology. That would be the only weak point, but then again, uEFI is pretty much uEFI, and if it's a presentation of local storage under uEFI, it should work in GRUB2.**

-- bjs

**NOTE: Nearly all Linux systems load a modular, initial RAM filesystem (initramfs), a legacy name for it is initial root disk (initrd), which contains everything the system needs to not just boot, but even run an entire system, including any and all drivers, complex volume organization, etc... (hence they legacy name "root disk"). It is a read-only image that runs from memory (the system then "pivots" from the read-only initramfs to the actual, root filesystem at the end). Anyone who has been dealing with Enterprises for awhile, especially enterprise storage, really doesn't like to work with NT, especially never pre-NT6 (you can only have one, single storage driver, ntbootdd.sys, for NTLDR in NT3-5). That's why most everyone runs, if they can, their Windows Server instances atop of VMware ESXi, which uses (even if dated from late 2006 technology) Red Hat Enterprise Linux 5 at it core. I really cannot stand dealing with boot-time issues when I have Multipath I/O at work under Windows, although at least NT6.1+ (Server 2008+) adopts a semi-modular solution like Linux has for any and all storage possible.

In other words ... if the Linux kernel can load a driver at run-time for it, it can very much boot from it! NT is not the same.

 

If that was the case, why didn't you include my very next 2 statements?

I bolded for a reason.

The P65x has no mSATA slots.

But there is still *0* performance difference between 2.5" SATA, mSATA and M.2-SATA.

 

An oversight on my part - sorry about that. But you are correct - the P650 has no mSATA slots, and mSATA and M.2 are both physically and electrically different, and so mSATA cannot be used in the P650

The only advantage a 2.5" drive might have is in heat management - the case acts as a heatsink or at least a heat spreader that the M.2 drives won't have.

 

Huh? Where did you get this information? That doesn't make any sense to me.

Why would the SATA pin-out in one M.2 slot be dependent on another M.2 slot at all?

Everything I've read says the P65x has three (3) M.2 Slots ...

2280: SATA + PCIex4
2280: SATA + none
2230: SATA + PCIex2 + USB + Audio + etc...

There's no reason using the PCIex4 channels in one 2280 slot should affect the SATA channel in the other. It makes no sense to me at all. The two SATA ports are complete separate. Serial ATA has always been point-to-point, bus/controller-to-ASIC/IDE (Integrated Drive Electronics).

The 2230 port is used for MAC-PHY (PCIe) and optional CDC-ACM (USB) -- i.e., WiFi and 3G/4G/LTE, respectively.

Virtually all WiFi chipsets -- most specifically, the Media Access Control and PHYsical IC(s) -- today have a PCIe x1 interface on-package, just like with Mini-PCIe. And virtually all 3G/4G/LTE solutions use USB -- most specifically, Communications Device Class Abstract Control Model -- which gives it a Serial-UART type command set.

SIDE NOTE: WiFi is synchronous, baseband, frame-based (like "block" storage) while 3G/4G/LTE are asynchronous, broadband, byte-based (like "character" modems).

I'm not going to even touch that other than to point to the oracle of Wikipedia to the "hows and whys" NVMx is being designed ...
- NVM Express - Wikipedia, the free encyclopedia

It's not just the throughput, far from it. SATA's stack (AHCI) is grossly inefficient for NAND. The fact that SATA has a physical signaling limitation of 6Gbps, which really only matters when you're doing single, large transfers, just exposes one limitation. In reality, one is never going to get sustained transfers like that, but AHCI/SATA has far, far many other issues. Most people won't see this until we actually get a new NAND interface and full NVMx stacks in our hardware, firmware and OSes.

In-a-nutshell ... NAND reads are like a really, really slow DRAM, which is still 2 orders of magnitude (100x) faster than a spindle seek. So it needs to be treated like such. That is where NAND will really flex its true potential, with a stack like NVMx. In the near future, we'll see some M.2-PCIe NAND devices provide a NVMx stack, all while a new, standard interface is developed for M.2 that all NAND devices use, along with its NVMx stack.

SIDE NOTE: NAND writes are painfully much slower, but can be buffered by on-device DRAM, so they seem immediate. Unlike a "read," where one has to "wait" on the extraction from the device, a "write" can be shunted down a pipe and largely "forgotten about." I.e., let the buffering in the device deal with it, at least until they fill, and then the device has to send back and overflow/flow control signal.

At least until Intel puts a native, NAND controller on-chipset, with full NVMx support. Then you'll likely eventually see legacy SATA NAND devices, still using the legacy AHCI support, become more expensive than the new, native NAND interface. M.2 has a reservation for it.

This has happened with many technologies over years, even SATA itself.

I.e., originally SATA was implemented using ATA-133 to SATA-1.5Gbps PHY bridges. SATA itself is actually still the same, ATA-based stack from the old parallel days. But until native SATA-3Gbps controllers and end-devices were common in ICs, SATA was more. Now ATA is more costly.

Right now, DisplayPort is undergoing the same, although "Big Media" doesn't like it. Only implementing DisplayPort (and that's pure DisplayPort, not DP++) on monitors drastically cuts down on component cost. It's already taken over inside of notebooks, as the legacy FPD-Link (commonly referred to as LVDS, but LVDS is a generic term) from the mid '90s (yes, that old) has finally been shed. it has made it much easier and cheaper on both system design and LCD costs for notebooks. Even DVI-D (yes, even ignoring DVI-A/DVI-I) is not exactly "full digital," only DisplayPort (at least not DP++). VESA also figured "Big Media" will still try to keep HDMI around, so there are inexpensive "Active" DisplayPort-HDMI converters out there (under $25), so even "pure" DP ports (non-DP++) will work.

So don't be surprised when you start seeing DisplayPort only monitors in the near future. Systems will eventually get there, although most will still probably have DP++ ports so HDMI and even mini-DB15 VGA can be used.

SIDE NOTE: You can blame Intel for "dragging its feet" on DisplayPort. AMD has been putting DisplayPort on everything for many years now, because it's so cheap to do so.

 

Which begs the question ...

Does the XP941 run "hot" ...
- Because of the actual NAND ICs increasing in performance?
Or ...
- Is it actually the PCIex4 bridge IC and other support logic used on-board?

That's my question! Why?

Because NAND IC thermal designs are actually well-mitigated.
PCIex4 (or x8 for that matter) bridge IC and logic (if on-die with other controller) are typically used in add-on cards for NIC and Storage ... with heat sinks on them.

 

It's suggested by the specs in the OP here and at Tech|Inferno, and confirmed by the manual, which, in the Storage section of Appendix D (Specifications), states:

Me neither. I was hoping you could explain it to me

It does. Slot 1 is 2230, and has key A, which makes it for PCIe x2, WiFi, USB, etc. Slot 2 is 2280 and has key M, which makes it for PCIe x4 or SATA. Slot 3 is 3042* or 2280 and has key B, which makes it for SATA or 3G/LTE.

* seems to be, based on the info in the manual combined with the pics of the motherboard

I'm glad I'm not the only one to which it makes no sense

Which is why I said currently

 

Oh, BTW, not to go too off-topic, but I don't think Intel will be "dragging its feet" on a native NAND interface and NVMx much longer. If they do, they will only hurt themselves, not just the consumer.

ARM's in-order Cortex A53 and nVidia's in-order "Denver," let alone ARM's out-of-order Cortex A57, is really starting to not only kick the (censored) out of Atom, but the last can actually challenge full 64-bit PC processors.

So ... given how most embedded ARM System-on-a-Chip (SoC) designs typically have NAND directly connected via pin-out, Intel will be really in trouble if they don't get a new interface and stack for PCs.

E.g., I'm really looking forward to what AMD's Opteron A series does to Intel in the server market. Especially if AMD pushes forward a NAND design if Intel drags its feet and plays games with marketing/sales-driven, software support crippling.

I know most people in the Windows world don't care, because Windows RT cannot run Windows apps (which are heavily tied to Win32/x86) and Office RT's compatibility is a joke.

But in the Linux server world, and even what we're seeing from companies like Valve (Steam) in gaming appliances (porting kits out of Win32/x86 -- even though Linux took over most game development workstations back by '05), once ARM starts getting a full 2x or even 3x advantage (it already has about 50% over Atom), and besting full PC processors (50%+ at 1/2rd the power), things are going to get very ugly for Intel ... and possibly Microsoft too if they cannot get people off of Win32/x86 compatibility.

Which could make the notebook market very different by 2018. People run apps, not OSes, and actually prefer to deal with appliances and tablets, especially if they give them all sorts of capabilities and, better yet, compatibilities, in addition to cost and power savings with far better performance. Right now tablets have better sales, but the primary computing device for most people is still a PC.

There's already rumblings of this around the XBox One and Sony PS4. I.e., it's not the AMD 8000-series dual-GPU that's the bottleneck, but the 8-core AMD Processor 15h (Jaguar) x86 design. Even Intel's fabrication advantage is not going to offer much more performance in an x86, let alone definitely not at that power envelope. GPUs have outgrown x86.

And NAND is begging for a native interface to the processor ... especially against pin-outs on the SoC in non-x86.

 

They clearly have a translation/logic issue.

Two M.2 2280 SSDs with SATA (Serial)
Interface (RAID 0/1) (Factory Option)
OR
One M.2 2280 SSD with PCIe Gen2 x2/
x4 Interface (Factory Option)

They should clarify it. Furthermore, note it's under the storage portion.

As far as ...

Slot 1: for M.2 2230 WLAN Combo Module
Card with PCIe & USB Interfaces
Slot 2: for M.2 2280 SSD Card with SATA /
PCIe Gen2 x2/x4 Interface
Slot 3: for M.2 2280 SSD Card with SATA
Interface
OR
LTE/HSPA+ M.2 3042 3G or 4G Module
Card with USB Interface

Basically that means ...

- Slot 1 (2230/Key A) has both PCIex2 and USB to it, so it supports a "Combo" module, including 3G/4G/LTE (USB CDC/ACM) on-card with the WLAN (PCIe-to-MAC/PHY), if it fits in 22x30mm. However, it's unlikely to fit, let alone the EMF/EMI issues in that small of a card hasn't been addressed.**

SIDE NOTE: I was previously wrong about it supporting SATA. It's Key A, not Key B. Although both include I2C, which is usually associated with most direct RFKill (turn off radio) hardware/software support.

- Slot 2 (2280/Key M) has both PCIex4 and SATA to it, so it can support either M.2-PCIe or M.2-SATA storage devices in 22x80mm.

- Slot 3 (2280 or 3042/Key B) has SATA and USB (no PCIe, even though Key A has the pin-out), so it can support a M.2 SATA storage device in 22x80mm or a 3G/4G/LTE (USB CDC/ACM) card of 30x42mm..

Today, most Mini-PCIe "Combo" modules with both WiFi and 3G/4G/LTE are 30x42mm. So what Slot 3 does is offer a way for vendors to adapt those existing designs. But I'm kinda curious why they didn't do that with Slot 1? Why didn't they create a "2242 or 3042" Slot 1?

Is the USB in Slot 1 / Key A being used by Bluetooth in a Combo card? Unless it's an EMF/EMI consideration, or a post-design issue (M.2 spec changes or other layout issues), I really don't understand why they wouldn't put the 3042 option in Slot 1. It only adds 8x12mm, let alone you don't have to make Slot 3 wider.

**SIDE NOTE: Back around '00, we couldn't get MAC and PHY all in one IC due to the sheer EMF generated by the PHY that would cause EMI in the MAC. Early WiFi cards had physical shielding -- literally a "wall" of metal between them.

 

I posted my research about this in a previous post, which pretty much agrees with what you posted about the slots. What I also found is that WiFi/Bluetooth M.2 cards are 2230 and are keyed for A and/or E, whereas 3G/LTE M.2 cards are 3042 and are keyed for B. Thus there's really no point in having Slot 1 support 3042 because of the keying (unless they made it a keyless slot? If that's even possible?) Or maybe they just wanted to keep WiFi/Bluetooth and 3G/LTE on different slots/cards? Or maybe the M.2 spec is too limited compared to Mini-PCIe to allow both Wifi/Bluetooth/3G/LTE all on the same combo card?

I'm not an electrical engineer or computer (hardware) engineer, so I'm basically guessing based on logic and my limited understanding of what I've been reading.

 

I decided to get the Sager NP8651 after finding it two days ago and reading the last 79 pages of this forum. To save money I was going to order without an M.2 drive and no OS (I can get cheap Win8.1 through school). I plan to get an M.2 drive aftermarket and want to run win8.1 on it and I need to know if this is compatible: Amazon.com: Transcend 32 GB SATA III 6Gb/s MTS800 80 mm M.2 SSD Solid State Drive TS32GMTS800: Computers & Accessories

It has the "B" and "M" key and will it work as the main os drive? If the one I listed will not work, can you link to a cheap 64g M.2 drive. Thanks!

 

As of today the laptop is on its way over to me.

Sent from my Nexus 5 using Tapatalk

 

I'd go ahead and get the $46 64GB version of the same drive, but it's up to you. Yes, that drive should be compatible.

 

but it is still a chipset problem, what would be a possible way to go around that? is that the reason why macs can boot PCIe without a problem? simply because their OS boot loader is different?

 

Congrats man!:thumbsup:
Tell us what you think.

 

So after reading through all the pages I finally decided pulled the trigger on the P650SE. I can't wait to get my hands on this skinny new toy. This will be my first laptop I ever bought for myself. Now here's come the hardest part of any online order the wait.
Let the waiting games begin.

 

December 9th for the SG... Please tell me this is a mistake... I can't wait any longer!

 

Ummm ... those are a lot of "assumptions." I.e., Keys are just to prevent people putting in the wrong cards with the wrong pin-outs, and have nothing to do with the various options capabilities of the cards themselves.

E.g.,
- Today's WiFi (MAC/PHY) baseband ASICs have a PCIex1 interface
- Today's 3G/4G/LTE (CDC-ACM) broadband modems have a USB interface
- Etc...

So as long as the designer creates the "combination" on a card, and the type of M.2 slot has the features needed, there is no reason why different combinations cannot exist. The keys are what the pin-outs provide in logic, not the type of cards that are capable.

Ahhh ... no. Size has nothing to do with keying.

Again, size has nothing to do with keying. The keying is the pin-out of the logic connections.

The "keys" are only to prevent people from putting the wrong card with the wrong pin-out into the wrong slot. They are an 100% consumer-servicing construct, and virtually 0% technical reason. It is commonly used in slots to prevent consumers from inserting cards into slots with incorrect voltages, pinouts and other things -- typically voltage and ground (as data signals usually don't cause boards to short).

E.g., if you think DDR, DDR2 and DDR3 are keyed differently for signaling, that wouldn't quite be the primary reason. They are keyed differently for voltage, first and foremost.**

There's no such thing as "keyless." There are "multi-keyed" slots/cards, sometimes called "universal" if they support all keys.

E.g., PCI[-X] cards can be either 3.3V, 5V or 3.3V w/5V tolerance. PCI[-X] slots are usually only 3.3V or 5V.

**SIDE NOTE: Intel had a particularly nasty oversight during AGP where they "reused" some early 3.3V keying for late 1.5V slots, and called them "universal." E.g., plug in an original AGPx1/x2 card that only did 3.3V into an AGPx8 board with an "universal" slot and it would typically fry it. I know, I did it myself, and trust me, I let my friends at Intel know about it. They also screwed up the original AGP spec as it was 1mm off in mechanical form-factor, which two (2) OEMs ran into issues with. These were largely due to the fact that AGP wasn't a "standard," but actually "trade secret" that was a segmented PCI bus with some software for DiME (very long story). But, in a nutshell, AGP would never flown under the PCI SIG standards.

That all said ...

Today ... there are plenty of Mini-PCIe slots of 30x42mm that have both 802.11 WLAN baseband ("WiFi") and 3G/4G/LTE broadband. So why wouldn't there also be a M.2 card of 30x42mm (3042) that also provides the same? That's what I don't get.**

So ... why didn't Clevo create a slot that could take either a 30x42mm or 22x30mm? Why did they use one of the 2280 slots? Unless I'm really not reading things correctly, Type A, B and E all provide the necessary PCIe, USB and (for direct RFKill) I2C. SATA, which is only in type B, is not needed for either.**

Now I can see the 22x30mm being almost too small, for current EMF/EMI designs to have both baseband (WiFi) and broadband (3G/4G/LTE), but 30x42mm with such is common today -- again, today's Mini-PCIe 30x42mm cards with both already have to provide a PCIe to USB bridge on-board. So either A, B or E all provide both PCIe, USB and (for direct RFKill "hard switch") I2C. So why not combine the 22x30 and 30x40mm need, and leave the 22x80mm slots alone?

"Too limited"??? I don't know what you mean.

M.2's 0.5mm pitch provides more traces than Mini-PCIe's 0.8mm pitch. So one can fit in more PCIe channels, more things like SATA, USB, I2C, etc...

Actually, you don't need to be one. However, you do need to be familiar with how common baseband 802.11 WLAN ("WiFi"), broadband CDC-ACM UART modems ("3G/4G/LTE"), etc... are typically connected via PCIe, USB, etc... respectively.**

**SIDE NOTE: It's probably because I'm a Linux wennie, but a lot of these details are commonly exposed, even as all drivers are autodetected and setup. Standards and details tend to be well known as a result. Although Windows gurus also end up learning them too, as they learn the underlying, common technology, instead of just vendors and their product drivers.

People have been jumping to a lot of assumptions on what the "keys" do and what the "combinations" are designed for. That the problem. They only define the pin-outs provided, not the actual cards.

Again, today ... for Mini-PCIe already ... we have 30x42 (3042) cards with both baseband and broadband wireless. M.2 changes nothing other than adding more direct pin-outs so they don't have to bridged from the PCIe.


**SIDE NOTE: Now the only thing I can think of is a little bit of non-sense that some carriers like to do. E.g., AT&T stupidly likes to put old, out-of-date drivers in a storage area of its broadband cards. So ... maybe that's why they expect a Type B ... with SATA. Why? For nothing more than the drivers. It's not used for broadband (or baseband for that matter) wireless at all. But the company with the Death Star logo typically has its way ... and that's why they also get their own slots too. And I assume non-SIM carriers might have issues with an unified baseband+broadband.

That still kills me, especially here in North America. There's no reason why we cannot have an "unified" broadband card, even a global one. It can be locked down in the firmware and even selectively authorized at the SIM-level for only specific channels. Many of us involved with the IEEE have been round and round with the FCC on this, and even the FCC eggheads know that someone who knows what they are doing can still all sorts of stuff with the radio that they're not supposed to do ... regardless of OS.

 

Nope. The Key merely provides a way for the slot to tell the consumer what logic pin-outs are available -- i.e., so they don't insert the wrong card and Voltage or Ground load doesn't short a data one.

The type of card is actually left to the designer.

E.g., Anything with USB can support all sorts of stuff, including the typical broadband UART modem ("3G/4G/LTE") via the CDC-ACM device type/command set. That would be type A, B or E.

As I hinted before, I'm starting to assume several carriers will only produce a card that fits in a Type B. Why? Because they want a SATA interface pin-out so they can store drivers and software on the device without having to add a bridge to SATA (either via PCIe, or even via USB). But there's no reason why type A, B or E couldn't support a "combo" baseband + broadband wireless, because they have PCIe and USB. I'm seeing it's more of a political reason.

E.g., AT&T even pushes their software and drivers into the nVidia Shield Tablet, even though it's unlocked and works with T-mobile here in the US. That little SATA logic makes all the difference to them, hence they Type B connector "requirement" (that is not technical) in a PC solution.

 

Apple controls the hardware + OS. That is everything! Sun used to have similar in the server world. Controlling both the hardware and OS means you can do all sorts of things at the firmware-level, device numeration, pre-OS intelligence and setup, etc...

The PC world is very open-ended, with a lot of specifications that a lot of OEMs, IHVs, etc... implement partially or incorrectly at times. That is what always causes issues. Microsoft's boot loaders also have been the absolutely most inflexible and simplistic of just about any PC OS, and don't coupe with variations well. But at least with uEFI-GPT, Microsoft has finally setup a dedicated Microsoft Reserved (MSR) partition, instead of using undocumented, hidden sectors to store things (that can be overwritten, even by its own OSes and tools). And while Linux has always been flexible for boot, the sheer aspect of device enumeration and standards has been a 2 decade-long argument that has finally reached an agreement to "unify", for the most part, as of January.**

Now as far as PCIe ... that's a whole different story. I'll try to use an analogy ...

Saying "PCIe" is like saying "transportation."
Saying "SATA" (which involves AHCI) is like saying "drive on the Interstate System."

If I tell you (assuming you're an American) to "drive on the Interstate system," other than asking for a destination, you're probably set. Maybe you'll ask for directions. But the Interstate System has been setup to give a controlled flow of data that is not the most efficient, but probably the most common, efficient system for getting a lot of varying traffic close to about anywhere.

If I tell you "transportation," you'll say W-T-F? And that's really the thing ... what did I just say?

Virtually all SATA ports eventually connect to PCI channels. You just don't see it. Even at the OS level, you already have a block device driver, and even a programmer using the "raw" device is already accessing blocks (if not using the filesystem atop of the AHCI block support). The system calls know how to push the data (usually 4KiB blocks) down through the AHCI stack, including the 32-bit data link layer and from memory mapped I/O in RAM through the PCIe I/O interconnect's bridge to SATA and up through the specific SATA port to the SATA device (and back up the stack on its Integrated Drive Electronics).

PCIe is literally just the I/O interconnect. There's no single standard for storage. There's no single standard for communication. There aren't even any standard, end device drivers in an OS for a "PCIe device" other than how to move bits across it, raise interrupts, etc... It's just an interconnect that lets the I/O talk to ports, modify registers, do memory mapped I/O (aka Direct Memory Access -- I/O device can copy to/from RAM), etc... There's nothing to handle anything for general block/frame (XXX bytes at a time), character/byte (single byte at a time), etc... type devices. Those have to be written, as well as have the system calls in the OS for various functions, etc...

So ... someone has to provide all those solutions. Anyone connecting NAND to PCIe needs to have a PCIe bridge, some sort of hardware/firmware, and write all the uEFI support, boot loader support, OS driver and other support, etc... It's not easy, not at all. It's extremely involved. There are designs out there. In fact, before mSATA, there were some for Mini-PCIe as well. Again, mSATA only added SATA pin-outs to Mini-PCIe to avoid having to do such, and leverage existing SATA-based NAND designs, support logic, etc...

E.g., Like for its PCIe XP941, Samsung builds it's own Boeing 787s, plus it's own buses and corresponding a bus system, that covers every airport and plausible destination, even if it doesn't have to build the airports and bus terminals that people will still use (PCIe). Another vendor could use drones, and not even be able to move people. Etc... There's no standard and no real "requirements" other than to move things, when it comes to PCIe.

The ONFI was supposed to provide a "native" NAND interface years ago. They haven't exactly yet. NVM Express (NVMx) is eventually going to become the AHCI for NAND devices, like AHCI is for SATA today. Until then, vendors creating PCIe devices are all creating their own, one-off designs. Even the lack of a physical interface and interconnect for NAND that everyone agrees upon is part of the problem. M.2 has a reservation keying for a future EEPROM interconnect. So that new "card" type, with NVMx, will finally "standardize" a new solution, and bring costs down.

Understand that in the embedded world, we typically connect NOR and NAND directly to the controller, or via some bridging that, ultimately, is addressed almost just like DRAM. Of course, that's not always the case, but typically the device is very, very close to the processor, and not over a complex I/O bus and through various, other bottlenecks.

I.e., NAND can even be mapped into address space with ranges that don't overlap with DRAM (or SRAM for that matter). Of course, we also have to manage the entire device, cell rotation, etc... Typically uboot is the boot loader and we have all sorts of options to organize the blocks, as well as specific file systems to manage the device (e.g., Linux's Journaling Flash File System releases, JFFSx). It's still inode-based, like Ext and XFS for compatibility with existing GNU/POSIX system calls, but looks nothing like them.

That's not an option in the NT world. Hence why we have this whole legacy stack and all sorts of other things. But NT will get there, and it will be Intel drivers, firmware and things like NVMx along with a new interface design. Until then, every vendor's PCIe option will be different.


**NOTE: SIDE NOTE: In the Linux world, projects built around the newer systemd facilities are trying to set a lot of new standards -- from enumeration to real-time interaction -- that takes advantages of a lot of existing standards that went underutilized in prior releases. E.g., modern Linux systems have extensive message passing and all sorts of orchestration, but typically not in a centralized system manager -- which systemd, with its various, modular components, provides. This is causing both a lot of consternation and, even worse, Fear, Uncertainty and Doubt (FUD), largely because the common approach in the GNU/POSIX worlds has been to hack scripts to change things. But one thing is for sure, things built around systemd attempts to finally "standardize" a lot of things, including solving a lot of long-time PC device enumeration and naming in a way that is ... to stretch the argument, very Sun-like. A major driver is the cloud generation, which is a very, very long story, but with a lot of broad, industry buy-in (from CoreOS to some of the largest, major Debian userbases, and not just Red Hat and SuSE).

 

You write way too much, and you can edit your post if you want. So no triple post needed . NAND is more than satisfying for gamers / PC enthusiast, not everyone can afford a 1To MLC like a 850 pro especially students that are numerous in the NBR community.

Now for the explanation part regarding computer science it's always great to have someone sharing its knowledge but you should sum up your toughts especially on a forum where people are waiting for news/updates on their 1.6k pre-ordered machine. Thank you for the explanations of course . Benchmarks show that a "pci-e" SSD is way faster, that's the point discussed.

 

another option would be to use the

Spoiler

function, that way theres no need to shorten your posts and anyone willing to have the "full knowledge" can still read up, otherwise the posts are more compact and easier to scroll over by other users just wanting to check on news concerning the P65x

cheers

 

Which vendors have the 3k matte screen available? 970m is fine.

 

As always.
And I knew it was only a matter of time before that came up.

No, that's the mega-oversimplification that keeps being thrown out, which causes people to wonder why they have issues booting, don't see much difference (not the same jump from platter to mSATA), etc... I've tried to explain some of those details, so people can understand what they are actually getting, why they pay more, etc...

It's actually funny because M.2-PCIe NAND devices in 2014 are really the "throwback argument" to 2009 when the first Mini-PCIe NAND devices appeared, prior to mSATA. mSATA was the commodity solution, which hurt performance but opened up availability. Until we have a new standard interface and the proposed NVMx stack, it's still going to be more expensive, and not always better, totally dependent on the IHV and their stack.

Spoiler

And that's the suggestion I was waiting for. I was already considering it and will likely use it going forward on any tech history, EE stuff, etc...

 

How long did it take between placing the order and having it update to 'shipping' status?

 

Hello Folks! I have been following various threads on this forum on Clevo models, but more closely this thread as I have zeroed-in on P650SE. I will be ordering mine soon!

 

hey, just a small question:
if you have to choose your ram for that clevo, and you need 32gb, what would you buy?