*** Official Clevo P870KM1/P870KM1-G/Sager NP9876 Owner's Lounge! - Phoenix 3.0 ***

Just to let you know that the app in my signature now includes P870KM
Please report bugs, broken stuff!

 

Yes and performance consistency under heavier loads.

 

video workload thats mostly sequential read/write so you would definitely benefit in raid 0, if you read/write to the from raided sector to itself which also note most SSDs can't handle mix workload very well and performance take a huge hit.

also, since its mostly sequential for video workload, it is less strain on flash compare to random writes, like torrenting for example, which hits harder in both SSD and HDD. u might be editing a video say 10GB or torrent something at 10GB, torrent would count as 2-4x worse in terms of page write in SSD. most endurance dont specify if they are random/sequential, so i take them as mix writes, only for server ssds they would probably tell u its for server work load, how many write drive per day.

 

If all you have used is HDD arrays in the last however it is a huge improvement whatever option you go with.

 

I just started reading about this new model today

Any major changes from P870, or the most major change is running z270, which is questionable if it brings any changes to the table.

 

See no reason not to go for P870KM1 / P870KM1-G if you are going to use Windoze X. There is not much difference. Only benefit with Z270 chipset is if you want 3x M.2 NVMe SSD's.

 

seriously doe, they need 4x M.2 and 2x 2.5", that would be perfect, though i would prefer more 2.5" but 2 would have to do. 3 M.2, cant really do two raid 0 array and that pisses me off, which ever remaining PCIE SSD wont be able to take advantage IRST write cache.

btw fan control software is good? for cpu of course..

 

I'm confused now, why cant you do 2x RAID 0 on the m.2 ?

On my MSI laptop, it has 3 m.2 slots, 2 are NVMe and one is standard m.2 so I put 2x 960 PRO 2TB in RAID 0 and the m.2 non NVMe slot is a single 1TB 850 EVO m.2 SSD

 

What do your WRITE numbers look like with 2 x 960 pro's in RAID0 ?

 

NICE !

 

I finally decided to take the plunge and order my EVOC P870KM1 from HIDEVOLUTION. I have been watching the "evolution" to this configuration for quite a while. Now that the 7700K processor, the Samsung 960 (Pro & Evo), plus the motherboard upgrade to 270 are available, I feel comfortable to proceed. Thanks to Donald at HIDEVOLUTION for his patience, help and excellent advice in guiding my choices.

 

excellent choice!

Please make a post in the Reseller Feedback Forum and share with us your order details / specs

 

see the bolded part two raid 0 array each raid 0 array = two x M.2 SSD, two raid 0 array means 4 M.2 pcie ssd.. which is not doable lol

 

That single regular M.2 slot can be raided with his normal ssd, thus giving him another raid array.....If i'm not mistaken.... But since he has a 4TB and a 1TB drive, raid 0 would not be beneficial.
Not sure there is any one laptop with 4 M.2 pcie 3.0 lanes. They are mixed. Sata and pcie.

 

By my estimation, 4 x NVMe drives should be capable for Kaby Lake... most of these laptops don't even have enough USB/TB ports to saturate all the lanes they have. I mean, 8 x PCI/e 2.0 lanes was the original chipset limits for Haswell/broadwell and prior, I believe. They were able to shove 5 USB ports and such on those chipsets from before; even if you toss 4 x NVMe drives that's still 8 x PCI/e 3.0 lanes left.

 

The chipset starts shutting down lanes at 3 (reduced to 2x SATA) then I think 4 would disable SATA totally.

 

Maybe within the theoretical I/O capabilities of the PCH, but the way Intel implemented it, only lanes 15-18 and 23-30 are usable with IRST. Plus GbE, WiFi, etc all eat a lane each as well since they can't share, and TB3 requires a massive 4 lanes.

 

Interesting

 

Here is what it looks like under the keyboard..

 

yea thats exactly the reason. theres none with 4 m.2 pcie ssd unless its desktop, laptop max at 3 with z270 cihpset and thats because of intel. a pcie SSD raid with sata is kinda pointless cause in terms of latency, sequential read/write and random read is all faster than two sata SSD raided together. only real benefit is the write cache which trade off isnt worth it.

that would also take up one of the 2.5" SATA slot, leaving just 1 for storage, not enough, 2 is already very little.

 

need a custom build pcb for mxm 3.0 to pcie ssd adapter. that way choose just 1 GPU and more SSDs, though probably not bootable or can't be raided.

 

pcie SSD raid with sata is kinda pointless cause in terms of latency,
Who in their right mind would be doing something like that????

1 pcie NVMe 2TB Samsung 960 Pro drive is faster and bigger than your 3 raids put together.....I would kind of start there first. This is what @Phoenix did . He basically just about maxed it out on both ends of the spectrum

And in raid 0 you gain 1000 megabytes on write and a little bit more snappiness out of the setup, but still capped at 3500/3100 no matter what. Single Drive is 3500/2100.

This is just to say that you can get a lot out of just two slots my friend. And with 3 you can raid 0 that as well, but then it would be over kill since you still can't go higher than 3500/3100, but it would surely be very snappy on response and added storage space to the raid array.

Side note:
Also, there is no desktop board with 4 m.2 slots. You would need Aic's, but for now....Here is one with 3.
https://www.pcper.com/reviews/Stora...-Pro-Z170-PCIe-NVMe-RAID-Tested-Why-So-Snappy

 

2x Samsung 960 PRO in Super RAID 4 on my MSI Titan Pro 7RF laptop

 

2x Samsung 960 EVO in Regular RAID 0 on my Sager NP9873/P870DM3 laptop.

 

Perhaps one advantage. Aka help those with heat issues on NVMe SSD's
"Reduced SSD loading leads to lower operating temperatures and less chance of thermal throttling"

 

Sorry bro, I don't have those issues so I never worry about them.

 

Ha Ha. But some are so concerned about the heat problem, fully looking all the time with lights and lanterns on the web after all sorts of M.2 heatsink. Even using copper foil under the bottom lid, which dissipate heat from all sorts of home made M.2 heatsink

 

Sorry double post. Still almost impossible to edit post on the forum.


@Phoenix Did yooo test ATTO Disk Benchmark v3.05 ? And if so, what was the result with the M.2 heatsink?

 

theres a lot of things i didnt go into details but yeah, in QD1 of 4k read/write, 1 PCIE SSD would top out even 2 or 3 in raid0. multiple ssds in raid 0 would increase latency overhead but just a bit, would benefit more from QD4 and up which would be nice for running multiple games and vmware, not saying its bad having 3 in raid 0 but 2 imo is good enough, was just hoping another slot for the remaining two to also be in raid 0.

i know desktop got no 4 slots cause ultimately they are still the same chipset z270 so max 3. what im saying is desktop can use it's pcie lane from CPU if someone for go their SLI setup, they would get another SSD with PCIE card adapter, that type. those can't go in raid though which is an issue cause they dont go through chipset.

this is why im eagerly waiting for x299 skylake-x to be in a laptop but all hopes fly away

 

@ole!!! No I totally get what your saying, but there comes a time when one should just be thankful for what is possible.

All i see now a days after a brand new product hits the market is...
I'm waiting for z670 this or nvidia 1580 that or amd 64 core mobile cpu or 10k ram sticks running at a gazillion ghz. and on and on...
Then I go and look what hardware they are rocking and it's stuff from 1999.

When does it stop?

 

Ok, probably a bone headed question and I have no interest in purchasing one, just wondering. The KM1 has 3 m.2 drives. Can you NOT Raid 0 #3 960 Pro drives in that machine? Is it 2 pcie slots and 1 sata slot?

 

Three M.2 2280 SSD PCIe Gen3 x4 interface (RAID 0/1)

 

honestly, you COULD raid them, but it really doesn't make sense to me to do so.

 

Thank you @Papusan @D2 Ultima
After reading through some of the threads I was getting confused.

 

The third KM1 M.2 drive is "optane ready". This means you can shove an optane drive in there and use it like a hyper-fast cache for your existing SSDs... which I still feel is pointless, because it only speeds up your ability to access data on the drive. It's like hitting data in RAM before pulling from the HDD. If you still need to load from the storage, it'll be no faster or slower than before. Also it should be limited to the DMI bus speed, which means it can't work faster than 4GB/s. Now a 4GB/s cache is great, but still.

This tech needs lots of consumer upgrades before it's really good. Maybe if we get some PCI/e 4.0 with lots of lanes and a DMI limit of about 10GB/s or so, we might get somewhere. But that'll probably have to wait till whatever the next socket is, a few years from now.

 

Just from myself being in the electronics design/ prototyping/ testing industry for about twelve years and knowing how that goes; you can bet that Clevo already has a KM2 revision in prototyping/ lab testing and KM3 is probably very early on in conceptual design. Its just how long it takes to move through a development cycle for any computing product. When you see questions like "let Clevo know this..." it will end up on being things that happen in KM2/ KM3 (six months to a year out).

They most certainly have manufacturer prototype Optane modules that they are already testing on with the KM1 and I would suspect that DP 1.4 /dual TB 3 would be a KM2 type feature.

As has been said, these are incremental improvements; Not major game-changers on a deployment roadmap. As the P870xx series is just about the top of the heap for gamer/enthusiast systems (barring a few others like MSI) I would not expect that they come out with something radically different until the AMD chip hits the marketplace. Even then I do not suspect that they will be "top end" systems as AMD is playing catch-up to Intel's development.

What would be groundbreaking is an improved power delivery system to get past the approximately 1-2 hour life that we can squeeze out of a P870 with SLI'ed graphics cards. Either loads will need to decrease (CPU/GPU optimization and we know that is tied to Microsoft) or a completely different power source (batteries based on nanotechnology or ultra-capacitors?).

There are things "I would like" that just are not happening on these types of systems (integrated GPS for position and precision time reference), a better camera than the 2 MP thing we have today that is very "1995", an integrated LTE modem (or at least the slot for one).

As far as the heat-sink fitment issue; Yes, Foxconn should be spanked over that one and Clevo takes the blame too for not holding a supplier to a manufacturing standard. You know (HINT HINT) that if someone came up with a "jig" that could realign the old heat sinks to proper specification (brackets, screw holes, mating surfaces) and lapped the unit you could probably get a really good yield of premium heat sinks out of what is essentially copper scrap.

 

I guess I was fortunate that when I placed my order in December (OriginPC, yes, complain away about my choices) I was able to cancel the DM3 order just a few days before shipping and change it out to a KM1 (Kaby). The price actually went down by about $100 so it is not that the KM1 is more expensive than the DM3, it is a wash as far as the dollars you spend.

You just need to endure the additional delay in getting an order and some of the still questionable compatibility issues with certain components (memory of different speeds) and that you will be locked down to Win10 (no big deal to me).

I am skeptical of Optane; while I hear how much it will help out as a fast cache for spinning drives I no longer have any of those in a computer I own. I started moving to SSD's +6 years ago and have not looked back. Initially the price-point will be high, sizes will be small, just like how SSD's were six years ago. Some of the Optane-tech alternatives are about devices more than just the M.2 slot, they are talking about modules that will be able to fit in to a DDR4 position and act as persistent memory across reboots (the entire instant on OS).

 

Now this one I do remember...Only the guy who actually started on this path was trying to get the memory slots to run ssd's

 

Actually, it will operate as a cache for SSDs as it supposedly has 1/10th the latency of NAND cells! Question is how does that 3rd card's bus compare against the two other PCIe x4 buses.

I believe that is one of the long term goals of 3DXPoint is to basically have persistent memory.

 

Depends on what you are doing as response times can actually improve with multiple drives as the controllers are under less load each.

 

And same time less heat

 

I fundamentally oppose the notion that Samsung should get two sales of high end equipment because they choose to release the product with no thermal solution and thus an inability to maintain its advertised speed

 

For the final time, please do not put words in my mouth. I have specifically stated multiple times I asked out of curiosity, and not to bash you for your vendor choice.

Yes, that makes sense. They have to plan models far in advance. That's why it slightly annoys me when super-basic refreshes like the DM3 --> KM1 happen; they could have done more.

dual TB3 is already on the DM3/KM1 though, I don't see why it needs to be on the KM2. But it doesn't have DP1.4 though, you're right about that.

honestly after having and fiddling with my DM3 for a few hours I can say if Clevo gave a crap, this could be a perfect high performance laptop. Lots of little engineering designs like the way the SATA drives are held is amazing. If their heatsinks, thermal pad designs, and such were proper (ironically like MSI's) and didn't really require any of the post-sale optimizations we call "standard", there would not even be competition, really.

We need new battery tech for sure. Even if we could make the 1080Ns draw less power and cut down the CPU a lot, it'll still pull huge amounts more power than an iGPU system would due to the hugely different power plane of the cards. "Optimal" power still drops things massively for the 1080N cards; they downclock all the way from near 2000MHz to 165MHz core and 135MHz memory or thereabouts (I checked last night and the massive clockspeed drops were actually a bit surreal) but it still used quite a bit of power I'd say.

Doesn't this have a slot or capability for 4G/LTE? Or do you mean something else?

That's interesting, I have only seen the KM1 be more expensive. That's a good find for you.

That's mostly what I'm thinking. It won't be much use with existing tech. It'd be pretty great if it could function as a full SSD at that speed, though. But inb4 it overheats into a crisp with the poor M.2 slot design.

Yeah, that's why we're trying to get some of the OEMs to send back heatsinks in bulk if they don't have very proper fit. Maybe Clevo will get the hint, since they have had this problem for literally as far back as I can remember except possibly the D900F unit I had.

 

brother johnkssss i thought you'd know me better. that should be reading AMD 128 Server cores and 10x 2.5" SSDs . yep we should be thankful i think its great clevo is doing well means at get to continue enjoy socket machine.

tbh im real hyped about mass storage, especially optane storage unfortunately i'd say it won't come to consumer anytime soon. Lenovo has it in one of their upcoming book and its used for caching the HDD which imo its kinda a joke.

optane sounds nice and all with super low latency but gotta note that due to it's low latency, its contradicting to connect it via chipset, because it increases latency. so best optane should be running off directly from CPU's pcie lanes or ram slots, for fastest performance, which in that aspect its clearly targeted for enterprise market. though it might still be faster than 960 pro in a pcie slot off chipset lanes, but it'd increase its latency, and as we know if we want them to be running off of cpu lanes then we'll need 20 cpu lanes or else good bye to SLI configuration.

 

wow desktop people can get one when this come out prob with a hefty price but as expected, mobo pcie so probably set that go to through cpu lanes rather than chipset for fastest performance, at only 375GB.. http://www.tomshardware.com/news/optane-3d-xpoint-intel-p4800x-cold-stream,33624.html

i'd think the 4k is capable of going so fast, it'll need to be on a HHHL format with heatsink to be able to cool off, for enterprise workload though, might be okay for regular use. but look at that endurance, 375GB for 30 DWPD LOLLLLL

"Latency and performance consistency are also two key attributes, and Cold Stream doesn't fail to impress--it provides sub-10 microsecond read/write latency and a sub-.150/.200 milliseconds 99.999% Quality of Service (QoS) read/write measurement at QD16"

latency that low for QD16 what happens when its QD1/2? superb low? in a normal high performance enterprise PCIE NVMe SSD QOS at QD16 is around 2-4ms, this optane tech is ground breaking, now we just need software to catch up using more cores and newer extensions.

also look at this article and its first graph measurements. https://www.pcper.com/news/Storage/...tane-SSD-DC-P4800X-Enterprise-SSD-Performance

500k iops in 4k means around 2000MB/s in random read/write performance.. this would destroy 960 pro completely, if ony we can fit one in laptop.

 

The intel servers can have 4-8 CPU attched m.2s easily for a total of 12.8-25.6GB/Sec of storage speeds

 

"For the final time, please do not put words in my mouth. I have specifically stated multiple times I asked out of curiosity, and not to bash you for your vendor choice."

DM3, I never meant that as a bash against you specifically, I have had other people question my choices as well. My apologies if it seemed targeted at you.

 

On the subject of batteries; since we all know that the P870's all have poor battery life it might of be a better choice of going with an ultra-capacitor in the future. They have a much higher capacity for short interval, very high current demands, a near infinite recharge/discharge rate and do not face the same thermal challenges as a battery when recharging. They could handle the current peaks that run for a few seconds or minutes during very high GPU utilization.

Even if they left the existing battery system in place ultra-capacitors still could have their place, It would reduce the power supply sizing requirements. Right now the switching power supply has to be sized for the absolute maximum power peaks as switchers do not have the ability to ride out a temporary high demand peak for more than a few seconds. (it makes the magic smoke packets escape from the solid state devices that make up the switching transistors on the primary side of the power transformer). When running on battery it would give the system a little more breathing room before the poly-fuses (self resetting fuses that are inside of most lithium battery packs) pop open and need to cool down to reset.

Ultra-capacitors are often able to do a million charge/discharge cycles before losing capacity (lithium batteries maybe a few hundred), they have about sixty times the energy density (per weight) of a lithium battery (8-10 Wh/kg), do not contain pyrophoric lithium metal that poses such a fire risk and is toxic (they contain graphine nanopowders) and are about 30% more efficient than batteries.

I believe that we are still a few years away from broader adoption of ultracapacitors (supercapacitors same thing) but for small devices with a high energy requirement (like a laptop, tablet or phone). One of the engineering challenges will be in incorporating a small switching supply as ultracapacitors have a very linear voltage drop as they discharge (where batteries are flatter until they approach total discharge where the voltage drops off of a cliff). The switching supply would handle when the capacitor is at high charge (much more than 19 volts) all the way down to where the capacitor is nearing depletion (5-6 volts) but is still very capable of supplying current until it hits 0 volts. The upside of capacitors is that you can charge them in seconds (if your supply and conductors can handle the hundreds of amps for that type of charging).

If you tried that type of charging with a lithium battery it would explode in your face.