*** Official Sager NP9877 / Clevo P870TM-G Owner's Lounge! - Phoenix 4 ***

There is a MASSIVE difference in speed between SSD and NVME. You might wanna check out why that is and then decide from there.

 

Can't say I'm astounded but, if HIDEvo does a similar job to my CPU as they did yours then, I'm in good shape. Most ppl have indicated that their temps are higher than yours so, either u got a primo piece of silicon ,or u on some shyt that no one else who has a similar CPU is on.

Shyt I play IS CPU and GPU intensive so, we'll see how mine compares. I'm curious about your screenshots n why would u max ya fans if ya shyt isn't running hot to begin with? U like the extra fan noise? If my CPU idles at 50 and isn't taxing any of the games I push, fans would be on low...

I'll be happy with a 4.8ghz @ 60ish c's and runs around 80c with max fans on playing a game like Destiny 2 or The Witcher. U got dem "damn near 2 good to b true" temps right now.. I'm especially curious about those screenshots now...

Sent from my SM-G950U using Tapatalk

 

This is after about a half hour of Overwatch.

 

U got a hell of a CPU...#salute

Sent from my SM-G950U using Tapatalk

 

Who the hell would run fans on high when the machine is idle? I run fans on high during gaming sessions. I use headphones for sound so I couldnt care less what the machine sounds like when I'm gaming. The rest of the time the fans are on auto.

 

@Ogg and this is without VC? i wonder how it will looks like when it finally arrives. Keep us updated

 

Got the following installed from the OEM:

Silicon Lottery Binned,
Delidded and Tested,
Factory Overclocked 8th Generation Intel Core i7-8700K 6 Core-12 Thread Core Processor,
4.7 GHz (HIDevolution Overclocked to 4.9GHz, rated to 5.1GHz) - GUARANTEED Performance (Thermal Grizzly Conductonaut)

 

I know there is a massive difference between M2 sata III and M2 nvme, and I know why as well. I'm just concerned as to whther or not I'd lose lanes provided to the gpu's if I add an nvme ssd. Both of the cards are running at x8 right now for a total of 16 lanes, which is all the cpu supports. So I'm curious as to the lanes the nvme ssd would use. I've read it would use the chipset's lanes, however, those lanes have to come from somewhere too.

 

Thanks @Prema for the suggestions!
3DM11P - https://www.3dmark.com/3dm11/12537394

3DM11X - https://www.3dmark.com/3dm11/12537384

*Spent the rest of the day figuring out how to make Catzilla work. It turns out that installing the Z170 chipset from my P870DM2 and transferring to the P870TM did the trick.
Catzilla 576p

Catzilla 720p

Catzilla 1080p

Catzilla 1440p

Catzilla 4K is grayed out and I'm unable to test.

@D2 Ultima @Papusan , 4.5Ghz is doable after all. Temps are between 77-79C (fans on OC profile, 1.14V). Sorry I panicked a little early.

 

This...
Bout to hit them up and up my chip...


Sent from my SM-G950U using Tapatalk

 

SO, after some new information, I'm at a crossroad.

My CPU is paid for but hasn't been built yet. This will mainly be for gaming both on the laptop monitor and out to a 34" 3k Widescreen Monitor @ 120hz GSync. Here is my issue...I want the performance but I'm trying to maintain decent thermals. I also want the upgradability. (Provided the next chipset is viable and compatible with the Coffelake MB)
Of the two options below, which would you choose and why? Which one would you choose and why? They both would be around the same price as I would add additional options like a longer warranty, RAM and RAID M.2 SSD's to option "A".

A - Silicon Lottery Binned, Delidded and Tested, Factory Overclocked 8th Generation Intel Core i7-8700K 6 Core-12 Thread Core Processor, 4.7 GHz ( Overclocked to 4.9GHz, rated to 5.2GHz) & NVIDIA GeForce GTX 1080 8GB GDDR5 X 200W + Vapor Chamber GPU & CPU Heat Sink Set - NVIDIA G-Sync Enabled

OR

B - Delidded - Unlocked, Under Volted and Overclocked 8th Generation Intel Core i7-8700K 6 Core-12 Thread Processor, 4.7 GHz ( Overclocked to 4.8GHz) & DUAL NVIDIA GeForce GTX 1080 8GB GDDR5 X 200W + Vapor Chamber GPU & CPU Heat Sink Set - NVIDIA G-Sync Enabled

 

I'd stay at the option you are at man, if I were you. And just tweak it yourself. I personally am going to do just that.

 

Definitely option B.

I personally enjoy the 2 1080's over just one. Sure SLI isn't supported by every game, but it's still more horsepower for the ones that do support it, and just in general. Plus, I think you'll be battling cpu temps with anything over 5ghz, just my opinion. Yeah, that particular chip may be silicon lottery, but the gains from 4.8ghz to 5.0ghz for gaming are going to be rather minimal. Not worth it to me considering once again, I feel like you'd be battling temps at that point. Plus OC'ing is something you can save your money on and do it yourself man.


Ask, but they may also let you buy the warranty after the purchase too...if that may be more feasible for you later on. It's worth asking anyway.

Like I was telling you yesterday, you got the same build as me. I'd stick with that.

 

Don't even bother

Forget the silicon lottery one, take two cards if they come out to the same price

 

Thx fellas

 

Agree with others go sli. Imo Not worth all that cash for .2 higher overclock that will keep ur processor cooking. Much rather have another gpu

 

@Phoenix
@Papusan
@Mr. Fox
@Meaker@Sager
@D2 Ultima
and anyone else with knowledge on this,

Correct me if I'm wrong, but I think I have finally done enough research to understand the pcie lanes stuff. Would just like some extra eyes to make sure I understand this right before I order my ssd to put in this laptop.

The CPU provides the GPU's with 16 lanes and the mobo chipset provides 24 lanes, but only 8 lanes useable(since something on intel's site said that the mobo would only have as many lanes as the cpu lanes, that part still confuses me.)
So, the 16 lanes coming off my 8700k will go to my SLI cards, both cards at x8 performance.

Then, adding my M2 pcie ssd is going to take up 4 lanes, leaving me with 4 lanes left over, for another M2 pcie ssd or whatever else, in the future. However I probably wouldn't add another M2 pcie ssd beyond 1, because of possibly being bottlenecked by the DMI link between the chipset and the cpu since it only uses 4 lanes for that link.

Did I get that right fellas?

Also, is there a motherboard manual for the motherboard in the P870TM1(beyond just the laptop manual I received from HID)? I'd like to know which, if any, sata ports would get turned off if I did go this route and add an M2 pcie ssd.

 

i'd go option A.

reason being volta mobile GPU might just be around the corner and no point paying top dollar for something almost 2 yr old architecture. volta 12nm with HBM2 titan GPU just out iirc so theres something that will soon follow it. on the other hand, no doubt option B will get you a CPU but you don't know if it'll be the lowest end of the binned side, 4.8ghz might really be all you can get and no more, which would mean high temps high voltage. going for A ensures you of a much better CPU while leaving you room to upgrade on GPU (assuming clevo doesnt screw around again).

 

The DMI 3.0 link between the CPU and chipset is effectively 4x PCI-E. Any link you see from the CPU is independent, everything that is connected to the chipset has to come back across the DMI 3.0 bus to the CPU.

 

I see. So I understand how that works, but what about the part about how many lanes would the chipset of this laptop provide me for things such as an M2 pcie ssd.... 8 usable lanes? Did I understand that part correctly?

 

24 lanes which at any point can't exceed a total of 4 lanes bandwidth total. As I said all devices on the chipset have to talk to the CPU so are limited by the DMI bus.

 

this is why x299 is attractive because you can do raid boot off vroc directly placed within the processor, while having additional lanes on the chipset to use. lower latency, but cons is basically less matured of a product as it is brand new, as well as higher cpu usage involving high QD workload.

 

Also you need a specific workload to worry about 4x bandwidth to be a limit for you.

 

Both of your 1080s in SLI should run at x8. The way the motherboard has been designed by Clevo, the P870 family never runs any MXM at x16 even with one 1080 installed. Not sure why or what they did to it, but it has always been that way. Adding two M.2 NVMe PCIe SSDs will not reduce them from x8 speeds. But, it's still enough bandwidth with each running at x8. I don't think it would change the benchmark scores at all with single or dual 1080 running x16. If you just accept that it doesn't run x16 and move on, everything will be fine. It used to bug me.

 

Adding in switches costs money and a full x16 connection offers no benefit in single card mode for that cost.

 

Why are clocks down to 4.6GHz ? I mean you said you run your chips 50 or 51x. You run 4.6GHz all 6 cores? Thanks

HeHe. 4.3GHz would be a big disappointment

 

x8 for SLI is a problem for games, but not for benchmarks, which are designed to stress the core and not cross-talk between cards. It's why four-way SLI scales well with them.

Single GPU though, doesn't even matter if it's a 4x lane width on PCI/e 3.0, it's enough.

 

We can source the B173QTN01.4 and provide it as an option for $175 installed.

However it won't have G-Sync. According to @Prema G-Sync will be stopped at the driver level.

Other than @SirSaltsAlot, is any else interested?

 

In what way? I ask because I've never experienced any game problems with SLI to speak of. Except for the coding abortions, I've have nothing but good experiences 99% of the time.

 

Answer: Clevo firmware. It needs @Prema magic to not throttle the CPU. This is not new news to any of us. Same applied to systems with SKL and KBL CPUs. At least it can be fixed. Enabling and maxing out TDC for SA and IA can minimize it a little bit, but that's not enough to make it go away.

 

Scaling is inferior in games that generally do not like SLI, or use techniques that rely on previous frame framebuffer data. There isn't enough bandwidth between the cards. Forms of AA like TAA, Unreal Engine 4 on a whole, etc. They are not built with SLI in mind and the problem is bandwidth (which the HB bridge helps, but does not completely fix).

If you check the final section in my SLI guide there's a few links to tests with Witcher 3/Rainbow 6 Siege/etc and even some other games. It has been an issue since Maxwell, and every SLI-capable Pascal generation card suffers from it.

Benchmarks have no issue because they are designed to NOT stress inter-GPU bandwidth with such tech. Same with games that are designed to be AFR-friendly, like Sniper Elite 4 (which is what all games should strive to do, honestly). You wouldn't even notice that you'd have a problem unless you had the ability to compare lane for lane. So someone with say, a 5960X would be able to test by using PCI/e 3.0 x16/x16 and PCI/e 2.0 x16/x16 (since 2.0 is roughly half the bandwidth of 3.0, it emulates a PCI/e 3.0 x8/x8 connection), and then test the different games.

The need for bandwidth is so bad that some games even exhibit negative scaling (especially at higher resolutions) with the flex bridge. The important thing to note is that GPU utilization is not an indicator of positive or negative scaling with complete certainty. Sometimes a bad SLI profile will not let your utilization % cross about 50% for anything but short bursts and that's an indicator of negative scaling, but other times you will be getting a solid 90%+ utilization on each card and simply not be getting proper scaling % (scaling % being the % increase over single card, so 60fps on one card and 78fps on two cards is low scaling %, whereas 110fps on two cards is much better and more desirable scaling %).

For my favourite example, non-HB bridge on x8/x8 grants negative scaling in Fallout 4 at 4K:

Spoiler

I can also attest that I got very low scaling when i tried fallout 4 on my system, consistent with the HB bridge on x8/x8 here. And then you add more PCI/e lane bandwidth and...

Spoiler

 

Does this mean g-sync would be enabled with a prima bios update?

Sent from my LG-H932 using Tapatalk

 

G-Sync will never be enabled on any after-factory notebook LCD upgrade unless or until the Green Goblin adds it to their list of approved panels for that system and collects the ransom fees. So, if it is not an "official"' panel that Clevo offers, the short answer is "no" LOL. NVIDIA doesn't do anything nice for anyone. No pay, no play. But, if it is a 120Hz panel or higher, there is little point in G-Sync anyhow. You won't really be missing anything compared to a 60Hz panel with G-Stink. It will still be better than a 60Hz panel with G-Stink.

 

Oh yeah, one more thing @Mr. Fox, a piece of info you'd love. Apparently when the HB bridge came out, it added even more bandwidth to multi-GPU than it does now. In a driver series somewhere around 372.54 (or one or two before it), they forced the excess bandwidth over the LED bridge into frame pacing (making mGPU smoother). This effectively hurt scaling and made x16/x16 more important than it was on release. Welcome to Nvidia.

And since the Titan Volta launched, with NVLink connectors, people reached out asking about NVLink on bridge-only format for SLI. They were told that SLI doesn't work on NVLink and they didn't even turn on the ability to use multi-GPU rendering via NVLink... so basically, it's even worse. Not that they're "not giving us" NVLink in bridge-form for consumers, but rather it's being funneled completely away from consumers, where it would fix the entire bandwidth issue available right now and even re-allow 3-way and 4-way SLI to function better than ever before.

 

Gotcha, thanks for the clarification.

I'll probably go with a 1TB nvme ssd then now that I know if won't affect my GPU lanes and bandwidth and that the DMI link can handle at least one nvme. Doubt I'll ever install more than one though knowing about that DMI bottleneck now. Thanks for the clarification man, I appreciate it.

 

They do, they're both running at x8 speed. Like you said, I shouldn't ever have to worry about maxing that bandwidth at x8 really. I'm fine with that, as under most games, I shouldn't even get close to that x8 limit anyway, from what i've read at least. I'm extremely happy to finally get an answer between @Mr. Fox and @Meaker@Sager and know that adding an nvme won't affect my x8 SLI cards' bandwidth, it will instead use4 of the 24 lanes provided by the chipset, with 4 max on the DMI link. I didn't think you added the 16 cpu lanes and 24 lanes from the chipset, that part is still a little hazy.

I'll probably only ever do one though, since I finally understand the DMI link bandwidth limit and I don't want to limit other items coming across that link. I'm ocd with that stuff lol, I try to be efficient as possible haha. I'm weird, I know. Anyway, Thanks guys. Extremely appreciated.

 

You don't ever max the storage anyway so even if you got 3x NVME and have 3.5GB/sec read and write your other devices would not suffer.

 

no man, gotta go 10000GB/s to the moon

 

I think der8auer got 26GB/sec with his testing on 8x drives in the beta firmware.

 

yea but we all know thats at high QD and at low QD consumer usage its practically junk. not to mention the 4k random performance is a huge disappointment lol. right now VROC itself is not mature and takes huge performance hit when raid 0 SSDs together, better to have a single optane as boot drive via PCIE lane that is connect to the CPU, can shave off additional latency through chipset that way.

 

But why would a consumer want to go to the levels of bandwidth you are talking about?

Raid will always add a latency penalty too.

 

Agreed. But in this case its a 120hz gstink vs a non gstink 120hz. Is the newer revision screen worth losing gsync? I intend on getting a single 1080 setup. Im not sure it can push games at 120fps all the time.

Sent from my LG-H932 using Tapatalk

 

Forgive me if this is a niave question. Will this laptop run Linux? This is mainly for numerical processing on the GPUs (via CUDA) and CPU cores. I have heard there is a hot for Linux vs Windoze with CUDA. Any insight for this particular model would be great.

Sent from my Pixel 2 XL using Tapatalk

 

Though I havent tried it myself there is absolutely no reason the P870TM wouldnt run Linux. As a HUGE bonus there is no garbage iGPU to get in the way!!

 

You can turn off secure boot so yes.

 

Picking up my TM1 from FedEx tonight. I'm glad it came 2 days early since I leave the country in a few days. How does the single GPU vapor chamber differ from the dual Gpu vapor chamber? Or are they the same part minus the second gpu?

 

Didn't the single card version have two fans, one for the GPU heatpipes and the second for the MXM VRM heatpipes, then the third fan for the CPU itself? While the SLI version has the same two fans, but has a block type of assembly that goes over both GPU's (the heatpipes are barely visible from top on that one)?

 

Same exact part minus the second GPU.

 

There's resellers out there that offer Linux and Windows 7 so they definitely work. I don't know if they have to set it up differently but it's offered.

 

Congrats! If you have a chance, would you mind posting some temps once you have personalized it and have it up and running? I hear good things about the Vapor Chamber.