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

Well I just purchased a 500gb 840 evo and some gskill 16gb cas9 ram all for 330.00 using newegg 5% off promo and Black Friday deals

Shhhh don't tell the wife...

 

Yeah, I'm not telling mine about the ~$300 in RAM and SSD I got, either

Really tempted by the 27" PLS monitor for $220... but I think she might notice when the credit card payment rises even further

 

Pff, what a price.. here in the netherlands u pay € 210 for SSD and € 160 for that ram..

Verstuurd vanaf mijn HTC One met Tapatalk

 

Ok so ready for this?

So as we all know, Haswell iGPU only outputs DisplayPort signals.

On this motherboard there is a dual-mode 40 pin connector (so all the LCD cables have the same end that goes into the motherboard, the motherboard side is 40-pin)

Now that connector has 2 modes, eDP and LVDS, and that is controlled by some resistors on the motherboard that tell it what kind of screen to enable. There is also a realtek RTD2136 eDP-to-LVDS converter that is used in the case of LVDS usage.

So according to the spec there are:

8 different eDP screens that work in this model, in FHD,WQHD+,QFHD resolutions, from Sharp, Samsung, Panasonic, AUO, LG
4 different eDP cables that go to specific brands of screen, Sharp, Panasonic, Samsung and AUO/LG

2 different LVDS screens that work in this model, in FHD, from Chimei and AUO
1 kind of LVDS cable that works in the FHD LVDS models.

So you are correct in the fact that the motherboard indeed only has one connector, however the other-end of said connector (monitor side) and what kind of signal is on the motherboard port varies depending on what kind of screen you get originally.

We're not saying things won't work, just be careful as there are a bunch of different sellers of things all over the world and so what might work for one seller might not work for another, and all the above is based on the spec which could be all wrong too lol!

So thats all we are trying to say, is that its not quite as straight forward as it might seem. So just make sure you have what you think you have before you start ripping things apart

 

About to pull the trigger and buy one these, but looking for some advice on which screen to get...

My reseller won't be able to get the Sharp 4k IGZO for a few months so thats ruled out, otherwise here are my choices:

1080p IPS matte LED LG LP156WF4-SLB5,
3K matte LCD 15.5" WQHD+ / EDP (IPS PRO ) PANASONIC VVX16T020G00 LED 3.0MM,
4K glossy PLS SAMSUNG LTN156FL02-L01 (LED) 2.6MM.

Anyone know which of these is best in terms of colour gamut / overall quality etc?

 

mine is showing up in device manager as AUO36ED which I'm assuming is the edp version. Fingers crossed.

 

I think I'm going to have to order a TN model from lpc-digital so I can bank off of HT's model that worked to make sure my screen is the correct one lol since I already ordered the same screen as Dabeer and HT.

 

@mythlogic or other engineers here:

I'm pretty confident that I'd be able to frankenstein a touchscreen panel to fit the lid assembly of this Clevo. However, my question is, does the motherboard/firmware potentially support a touchscreen? What kind of limitations would I run up against?

PS: Yes, yes. I must be some sort of n00b for wanting a touchscreen, but having a touchscreen greatly augments my development workflows, and I am willing to spend big $$$ to have one in a laptop with a nice GPU. As of now, the only 15" touchscreen laptop available with a GPU that is 860m or better is the Lenovo Y50, and the screen quality on its screen options are garbage.

 

Brother, can you link me to this? I'm in the same market...

I saw a good price on 1 TB pro, but the RAMs were looking full price

Was your 5% off a personal coupon?

edit4: ok found a 5% coupon in my inbox

edit5: need some advice peeps

Can I buy 2 16GB kits to make it 32GB that works seamlessly? The prices are goood. Also, is this 1600 CAS9 better than 2133 CAS11? I read that they kind of equalize...

(sorry, I'm a little excited and it is 3:42 AM, so might be asking something inane )

 

2133 is definitely faster, but u can mix the two 16 GB kit without issues (most of the time at least, there are always exceptions to the rule, of course, but rather seldom)

 

Wow. Good info, there. Thanks! Wish I'd known before ordering the LTN156HL01, but I have high expectations due to HTWingNut's success.

One further question, though... if the system vendor can change the signal from eDP to LVDS when building their version of the product, wouldn't an experienced end-user be able to make a similar change to get it back to eDP? Or is some specialized tool like an EPROM programmer (and knowledge of what to do with it) required?

Thanks again.

 

Is there any information of a similar model for the 13-14" category? I know for a fact the P650SE has a 17" version, but I'd like to replace my W230SS which already looks old in the face of the P650SE...

 

That's insane. Wow. Crazy stuff. Thanks for the info!

But in any case if you see your current LCD and one you are considering are both 30 pin eDP LCD's, the cable is sure to be 30 pin eDP connector at the LCD side is all I'm saying.

 

Yes thanks alaskajoel. I assume straight HDMI is the same story? Anybody know what version of HDMI it uses?

Like iaTa said it looks like switchable graphics might not be an issue once the Rift finally comes out. It's just something I saw. At any rate I don't think there are any 15.6" thin/lightish machines with a 980M that don't have Optimus, and I'd rather not switch to something big and bulky based on an unreleased product. Worst case I guess I can trade up at a later date.

Really nice seeing Mythlogic (and other vendors on here too) with all the info. Makes me glad I ordered from them. A little bummed it might be longer than I first thought, but I've been gaming on a Dell XPS M1730 with dual 8800Ms for 7 years now. I can wait another month or two I suppose.

 

Oculus Rift will require a killer rig anyhow. The 970m or 980m even won't be able to push 120FPS minimum at 1080p with any amount of reasonable detail which is really what is required. 4K is truly what's needed to make it look the best, but we're a long ways out from that.

 

Read latency of the DRAMs are basically the similar. As always, depending on your usage, the difference between 1600/CAS9, 1866/CAS10 and 2133/CAS11 has to do with how you use the DRAM.

if you are reading random areas of memory a lot (like most), latency matters most (especially the last number "D" in the "A-B-C-D" timing) so increased, synchronous clock doesn't impact much. If you're bursting reads and writes, then synchronous clock helps. In any case, low timing at high clock is the ultimate, so 2133/CAS9 (for which there is no JEDEC spec) is definitely faster than 1600/CAS9 because the timing is likely lower (again, the "D" timing in "A-B-C-D" is really the biggest factor). But going for a DIMM with a better clock but much worse timing (2133/CAS13 v. 1600/CAS9 -- again, try to get the full timing to be sure) may actually result in worse performance.

Remember, DRAM cells take many dozens of cycles to actually read/write memory to their cells. The way we mitigate these very slow cells (e.g., 10-16nm ~ 66-100MHz access times) is to burst the read/write transactions so the request, made dozens of cycles earlier from entire blocks of cells, are returned synchronously in a burst. But there are still the physics of the underlying DRAM cells, and there are many arguments now that we're reaching a "point of diminishing returns" with yet higher and higher clocks when some of the fastest, commodity DRAM cells are still barely 10ns (100MHz).

There are also other tunings at the OS level that can impact performance with memory, although consumer versions of Windows usually cannot benefit from most of these. That's where I spend most of my time with servers, tuning those values.

 

Agreed even a 980M might not be enough, but the final version of the Rift isn't out yet so I can always hope it will manage to be somewhat playable. They keep talking about making the consumer version "affordable" but if you're going to need a desktop PC with dual $500 video cards to play games on it then what's the point?

Anyway, after further research it seems that at the very least, just because your external display can't go directly to the dGPU it doesn't necessarily mean you can't run games on a higher-than-60Hz refresh rate (which is what was the cause of my concern), so you can disregard my question and method regarding DP/HDMI not going through the iGPU as some measure of Oculus Rift compatibility, though I imagine it might make things easier.

 

The guys at MySN confirmed they've shipped out the XMG P505 Pro to me yesterday. Should be with me tomorrow! For those of you who don't know, it's based off the Clevo P651SE. Happy to share information, photos, tests, of course! Let me know any requests.

Specs as configured were: i7-4710, IPS 1080p [AUO B156HAN01.2], 970m, 8gb ram, crucial m550 256gb m.2., Windows 8.1, 180w PSU, Intel Wireless AC-7265.

It was my intention to add another 8GB Ram from my current laptop, and eventually buy a large SSD.

 

More info about this issue here: https://forums.oculus.com/viewtopic.php?f=34&t=13541

It's the lack of direct-to-rift mode via Optimus which is the concern as this allows for reduced latency ( more info) which as I'm sure you know is key.

They are looking into it though: http://www.reddit.com/r/oculus/comments/2fgs9v/what_exactly_does_direct_to_rift_mode_do/

So it looks like the only safe option at the moment is to go for the Clevo P750ZM which is the only 15" machine I know of with a 980M that does not have Optimus (or an external port connected directly to the dGPU).

The other possibility, albeit 17" is the Aorus X7 Pro with 970M SLI. The benefit of this is NVIDIA are planning to implement one GPU per eye for the Rift which will hopefully remove any SLI micro stutter.

Am I right in thinking that some MSI machines have external ports connected directly to the dGPU too?

We don't know for sure CV1 will be 120Hz (probably though) and even if it was you don't need super duper ultra graphics in the Rift. It's more about the design and interaction with the 3D experience at this point.

A 980M will be a great match for CV1.

 

120 FPS means the Rift's panels are 60 Hz. The game has to render two different scenes, so rendering on the Rift at 60 FPS is a bit more demanding than rendering the game at 120 FPS with only one camera.

 

The Rift only has one panel. The DK2 panel runs at 75Hz. CV1 will probably be between 96Hz and 120Hz.

But yes stereo rendering is quite a bit more demanding.

 

I have been enjoying my P650SE so far. Runs great on battery too.

So, how do I go about overclocking this baby? With my old MSI I had to rely on a button that linked to software that would enable it. I see no such button on the P650SE. Is it as simple as just running and configuring MSI Afterburner?

 

no you can't, the physical I/O pins for LVDS and eDP are drastically different, unless you get another mobo you are stuck with whatever connector you have. you may know what connector you have based on the screen you have, but it's not guaranteed to work unless you know the specs of your connector matches with screen.

for LVDS there are pins all over from 20-50 pins.
for eDP there are 30 and 40 pin eDPs I've seen so far

the sharp 4k panel uses a 40 pin 4 lane I-PEX eDP connector.
the samsung 48hz 4k panel also uses a 40 pins 4 lane eDP, but back in the lenovo y50 screen replacement thread someone tried to swap out the 4k panel with a sharp one but the connector didn't fit exact.

what mythologic is saying that with so many connector interfaces out there, don't simply just think that you can get one and swap out for new one yourself later. if you have a 30pin eDP, there's no way you are upgrading to 4k, if you have 40 pin eDP, chances are you may not be able to switch to another 4k panel.

in HT's case, that model he had had the same panel as the one in y50 and I recommended him the PLS panel because it worked for my y50, but that's not a case for everyone.

eDP vs LVDS

 

That's not quite what I understood Mythlogic to be saying...

What Myth said, which matches what Prema has previously said, is that the motherboard has a 40-pin eDP-style connector, no matter what.

What Myth said that is new, and has not been confirmed by Prema, is that there is a chip on the motherboard to convert the eDP signal to an LVDS signal. How this chip is enabled by the system builder, and whether or not it can be controlled by the end user, has yet to be determined.

When the chip is active, a custom cable to convert the 40-pin eDP connector on the motherboard side to an LVDS cable on the panel side must be used.

When the chip is not active, and a FHD eDP panel is used, they use a custom cable to convert the 40-pin eDP connector on the motherboard side to a 30-pin eDP cable on the panel side.

When it's a 4k panel, they use a custom cable that's 40-pin eDP on both sides.

Either way, no matter what, it's a 40-pin eDP-style connector on the motherboard.

 

I honestly don't think it's possible. it's possible for the mobo to have both connectors, like P377SM does, but there's no way to output both eDP and LVDS signals through one connector, their I/Os are different, their physical sizes are different.

 

you can use intel xtu to oc your cpu by 200 mhz (thats the max it supports) and nvidia inspector to oc ur gpu (+135 mhz on the core max unless we get a custom vbios, whereas video memory oc is not restricted )

enjoy!

 

40 pins is 40 pins, it doesn't really matter what signal they carry, so long as both sides agree on which wire carries which part of the signal. And since 40-pin eDP on the motherboard doesn't agree with 40-pin LVDS, it's up to Clevo to bridge the gap with a custom cable, which it sounds like they've done.

As far as circuitry on the motherboard to change what signals those pins carry, it seems plausible to me... and until or unless someone more knowledgeable proves it impossible, I have insufficient reason to doubt what Mythlogic is saying about it.

 

physical sizes?

 

Sure the connectors are different sizes, but the wiring in between the connectors is basically free-form (not restricted to a flat ribbon), and you can have different connectors on each end of the cable.

 

Duplicate post due to browser-site connection error.

 

Flat Panel Display (FPD) uses Low Voltage Differential Signaling (LVDS), while [embedded] DisplayPort ([e]DP) uses packetized data. They are drastically different technologies. You're going to be hard pressed to find any IC design today that doesn't prefer packetized data, hence why GPUs prefer to output DP and panels prefer to take DP. All other support than DP must be adapted or converted.

And this is just inside of notebooks and panels. For external connections, right now most people are "spoiled" by Display Port Dual-Mode (DP++) typically found on systems today. Dual-Mode includes HDMI and DVI-I (Signal Link) in the port, which works via any "passive" adapter. Without one, one would need an "active" (on-board IC in adapter -- powered by the +3.3V line in DisplayPort) to get HDMI or DVI-I (Single Link). An "active" adapter is always required for DVI-I (Dual-Link) and any analog (like VGA, DVI-A, etc...).

Spoiler

Flat Panel Display (FPD) uses Low Voltage Differential Signaling (LVDS) from the mid '90s, who's roots are actually in the '70s. You have a fixed clock and you send data over that clock. As CMOS semi and digital signal processing improved from the '80s to '90s, lower voltage was used and took over everything from displays to storage.

The main problem with LVDS is the synchronous clock, which has serious EMF/EMI issues at higher frequencies. Everything switching on a clock is a major EMF generator, and very suseptible to EMI, let alone pulls a lot of power because gates and other things are switching at the same time.

DisplayPort, like modern interconnect standards, from AMD HT to Intel QPI and even PCIe and newer Ethernet specifications, have moved to packetized data, where the signal control isn't a fixed clock, and data is embedded with the clock. That's why these technologies aren't listed with a clock/bits-throughput, but a number of transactions per sec with a maximum size per transaction. These could even be considered psuedo-asynchronous transmission protocols.

Packetized data drastically allows higher rates of data transmission per second and reduced EMF/EMI issues, and even reduce power as not everything shunts on the same, cyclic clock. Fixed clocks have always been the triple whammy (EMF, EMI, Power) that engineers have been trying to get away from, ever since mathematicians and computer scientists introduced them as the "control" in autonomous circuits back in the '60s (i.e., a mathematician is no longer "flipping a switch" to open/close a circuit).

This diagram is actually very limited, but it does show the "base" (legacy/original mid-'90s LVDS for) FPD designed for 6-bit color up to 1680x1050. There are 1080p and even 1440p FPDs out there, but they are fairly non-standard. Even eDP is still undergoing a lot of standardization in cabling/interfaces as it's fairly new.

 

however you can find a lvds to eDP converter, but it won't really fit in a LVDS laptop ^^.

 

which furthers proves that there's no single connector that works for everything

 

According to Mythlogic, the P650SE uses an on-board converter in the form of a Realtek RTD2136 that can be enabled by "some resistors on the motherboard".

But who knows, maybe they are blowing smoke to discourage people from doing their own screen swaps

 

EDIT: I see there is a new, universal connector standard from VESA. As long as the cable is designed for the specific EMF/EMI, grounding and other requirements of each standard, it can work. The years of being able to use a single 100, 50 or even 25mils generic cable are gone, and most cables are specifically designed for specific standards and tolerances in the '00s onward.

LVDS (FPD) and packetized data (DP) are utterly different. Almost everything in the semi world has moved to packetized data. In fact, video was one of the last bastions to finally move.

 

Yes and not really.

"Yes," for legacy, fixed clock interfaces. Prior to differential signaling, throughput was very limited. Then differential became commonplace in the '90s, but then we ran into its limits.

But "not really" in the world of packetized data. The idea is that you have a single set of modular connector(s) that really does work for everything. In the "worst case" it's just not capable of higher transaction rates, or has fewer channels (due to not connecting all of them).

AMD/API Networks (fka Alpha Processor, Inc) was really the first here with HyperTransport. Up to 16 bit-direction channels, and higher rates of transactions with each revision. It radically simplified system interconnect. PCI Express also does similarly for peripheral interconnect ("I/O slots"). Intel QPI/DMI for its interconnect solution to AMD (more of a "direct" v. "broadcast" different). All of them can "fall back" to lower rates of transactions, as well as reduced channels.

DisplayPort does more of the same. Packetized data is very easy to pin-out from ICs, so it drastically brings down prices too. You can use DisplayPort connectors of different standards and the end-user doesn't need to be familiar -- other than the fact that they may not be able to power a high resolution panel with an older system and/or an older rated cable connected.

However, interfaces/cables for high-res embedded DisplayPort panels are still converging in commonality. This also plagued FPD when panels starting going higher than the original 1680x1050 and 6-bit color limitations of the base design.

 

eDP's main reasons of existence are :
supporting variable refresh rate (an image that is still = 0hz whereas on lvds it will be refreshed 60 times/sec.)
Supporting higher resolutions 2 lanes = 1080p and 4 lanes = 3/4k.

 

Not sure if anyone has noticed, but XOTIC PC has uploaded unboxing and review videos on NP8651 (P650SE) on youtube.

Unboxing

<iframe width='560' height="315" src="//www.youtube.com/embed/j-DtKJon5DQ" frameborder='0' allowfullscreen></iframe>


Full Review

<iframe width='560' height="315" src="//www.youtube.com/embed/Qc4ZWlAEqP8" frameborder='0' allowfullscreen></iframe>

Hope these videos help in getting answers for some of the questions on P650SE model.

 

IIRC, Realtek RTD21 series are microcontrollers, although many variants have specific applications.

Pull-up/pull-down resistors are commonly used to enable/disable logic.
How they are implemented is up to the manufacturer.

They could require one to soldier the specific resistor(s) on-board.
More commonly, jumpers or DIP switches are used to enable/disable the bank of resistors.
Of course, they could be firmware controlled, however ... (next point)

Or preventing people from going into the firmware-CMOS and seeing what "LVDS" does. All of the sudden, their panel (using eDP) doesn't display any more, and they don't know why.

Having designed many products from the "engineering side" of things, you have to design everything for the case that a consumer is utterly ignorant, and will blame the product for doing exactly what they told it to do. My first, real lesson in this was writing documentation for Air Force techs, most of whom had barely a high school diploma with limited technical skills, assembling a multi-million dollar ballistic missile related control systems and launch computers. You'd be surprised how many pages it takes to just cover everything, step-by-step, with lots of "donts" unrelated to the procedure.

 

You guys are making my inner electrical engineer happy with this debate

Anyway, yes there is ONE connector on the motherboard that CAN serve either LVDS or eDP, remember its just a connector, 40 pins of put what-ever-signal you want on it. You are absolutely correct that LVDS and eDP are way different in signalling. However that doesn't mean much to a motherboard connector. Which is just 40 pins of electrical connection. Also yes eDP -> LVDS can be done on a small chip which is on the motherboard already. But LVDS -> eDP is not just a little chip Also that chip is only used when its needed (to make LVDS) not all the time. The P1x and the P3x all use that same realtek chip on all the SM(a) models as we already know that Haswell only spits out eDP but all those are LVDS displays and thats what does the heavy lifting to power our LVDS screens.

This is the first model that has this low profile 40 pin connector that can be either LVDS or eDP depending on what its fed. We got all this information by looking at the wiring diagram of the motherboard and that connector. The 8 datapins on it are either eDP (0-3) or LVDS (0-3) with the requisite clock/control pins also changing to account for the signaling. There are some other changes in backlight control as well depending on LVDS/eDP.

 

Ain't no documentation like Military technical documentation... :laugh:

 

I would strongly argue that cost is the much bigger one.
Again, video was really the last bastion to finally moved to packetized data.
But to your point, yes, resolution was always going to be the major PITA with staying legacy FPD (LVDS).

I wish everything was DP today.
Doing anything else just adds cost.

 

Okay, cool, they did come up with an "unified" connector for both FPD (LVDS) and eDP. I hadn't kept up with those developments. I know VESA had been pushing a number of designs around for eDP the last time I looked, but I had not seen an "unified" one with both.

 

I swear I wanted to pound my head against a wall sometimes. But the difference between a successful flight and one that undergoes "flight termination" was sometimes a matter of only what a $18K/year Air Force tech did.

 

we wish the SM generation was ALL eDP would have made things a little easier but they added cost to make the signal BACK into LVDS, so we feel the pain heh

 

Been there, done that, but for the Navy.

 

Its not "unified" as there is only one set of signal there at once. So its more of a switched connector

 

So the question still stands.. can the function of the Realtek eDP-to-LVDS chip be controlled through the firmware, or is it an actual hardware difference (presence or absence of on-board resistors)? Obviously it still won't work without the correct cable...

 

Maybe it's automatic through some sort of detection?

 

Doesn't look like it, as there are some clear pull-down resisters that control some of it, and none of the pins go back into the EC which would normally control this. That being said the EC on these is so complicated that they ran out of pins of logic on the EC so they had to get another EC on there to handle more pins of logic... Complicated buggers.