Clevo P650HS-G vs Clevo P950HR

Hello,
I hesitate which Clevo Laptop to buy. I will be using it for Gaming (obviously) and for school - at a lecture table, working at the library and etc. Therefore I need something portable and not very hard to carry around.
The laptops which I am hesitating between are the small, thin and lightweight Clevo P950HR and the somewhat bulkier and heavier Clevo P650HS-G.
Here are the two configurations:

Clevo P650HS-G around 1700-1800$ USD
Display
15.6” FHD 120Hz 5ms 16:9 Matte Display w/ G-SYNC Technology
Video Card
NVIDIA® GeForce® GTX 1070 (8GB) GDDR5 (Pascal) DX12
Processor
7th Generation Intel® Kaby Lake™ i7-7820HK (2.9GHz - 3.9GHz, 8MB Intel® Smart Cache)
Memory
16GB (1x16GB) DDR4 2400MHz Dual Channel Memory
M.2 PCIe NVMe or SATA SSD Slot 1
Samsung 960 EVO 250GB M.2 PCIe 3.0 x4 NVMe 3D-VNAND SSD
1st 2.5" HDD/SSD Bay
1TB 5400 RPM 7mm SATA 6Gb/s Hard Drive
Wireless Cards
Killer™ Wireless-AC 1535 ac/a/g/n 2x2 NGFF w/ Bluetooth 4.1
Dimension & Weight
15.16" (W) x 10.67" (D) x 1.13" (H) - 385mm (W) x 271mm (D) x 29mm (H)
5.84 lbs - 2.6kg with 4-cell Battery


Clevo P950HR around 1950$ - 2000$ USD
Display
15.6" 120Hz - 5ms Full HD (1920x1080) TN Antiglare Matte Type Display
Video Card
nVIDIA® GeForce™ GTX 1070 Max-Q w/ 8.0GB GDDR5
Processor
7th Generation Intel® Core™ Core i7-7700HQ Quad Core Processor, 2.8 GHz (Max Turbo Frequency 3.8GHz), 6MB Smart Cache
Memory
16GB (1 x 16GB) DDR4/2400MHz
M.2 PCIe NVMe or SATA SSD Slot 1
Samsung 960 EVO 250GB M.2 PCIe 3.0 x4 NVMe 3D-VNAND SSD -
1st 2.5" HDD/SSD Bay
1TB 5400 RPM 7mm SATA 6Gb/s Hard Drive
Wireless Cards
Killer™ Wireless-AC 1535 ac/a/g/n 2x2 NGFF w/ Bluetooth 4.1
Dimension & Weight
14.96"(W) x 9.80"(D) x 0.73"(H) - 380mm (W) x 249mm (D) x 18.5mm (H)
4.19 lbs - 1.9kg (with 4 cell battery)

Both of the laptops have a 120Hz, something which I was actually looking for. The P650 has a 7820HK processor, though I can go with a 7700HQ too. The main difference is the Max-Q GPU. Now, I've read a lot about Max-Q and there are sooo many different opinions that I got more confused than I was in the beginning.

My question is, is the P950HR is worth the extra price for an underclocked GPU (apparently with better thermals?) but insanely thin build? Looking at benchmarks, the Max-Q seems to be between the 1070 and 1060, however in real world (FPS tests for example) the Max-Q is only 5 to 10 FPS slower than the normal 1070. This is a huge dilemma for me right now, as I cant justify neither laptops as I need both a super powerful and portable machine and both of these devices seem great.

 

The Clevo P650HS-G has better cooling, performance and value. The only reason to consider the P950HR is if you really really need/want a thinner and lighter notebook and are willing to sacrifice some performance and thermals and pay more.

 

Thank you for your fast reply. I am leaning towards the P650HS-G right now.
The CPU thermals seem to be better on the P950HR, or at least according to these reviews:
http://forum.notebookreview.com/threads/sager-np8157-clevo-p650hs-g-review-by-htwingnut.801138/
http://forum.notebookreview.com/threads/sager-np8950-clevo-p950hp6-review-by-htwingnut.805339/
But the difference seems very minimal so that is not the point.

My other question is, do you think that the P650HS-G would be comfortable enough to carry around all day in a bag?
And would there actually be any noticeable difference between the performance of those two machines?

 

I would take the P650 if you could handle the slight weight/size bump. It is better cooled. I also highly suggest two sticks of RAM instead of one, and to ignore the 960 EVO in favour of an 850 EVO; you won't have any notice-able speed loss. Also get a 7200RPM drive for the P6, you don't want 5400RPM. Guarantee that. Also the 8265ac wifi cards over the Killer ones.

Sent from my OnePlus 1 using a coconut

 

At about 5.5 lbs, I'd say it's easy enough to carry around. Apples to apples, thermals should be just about identical.

 

Thank you everyone! I think (and by think I mean that I am almost 100% certain) that I will go with the P650HS-G. Pretty much everyone is saying good things about it ! The switch from a gaming PC would be something new for me, but, funny enough, I am looking forward to it as I want to become as portable as possible (also the fact that I can play my favorite games pretty muh everywhere makes me quite happy, hehe)

I will go with two sticks of RAM then . I was thinking about the SSD, but some benchmarks are showing that the 960 EVO is more than twice faster than the 850 EVO? Also what is your opinion on the FireCuda SSHD (or hybrid hard drives in general) ?

PS. After looking at topics I found that many people are having problems with the Killer cards, thank god I asked here .

 

Meaning dual-channel vs single-channel? A lot of benchmarks prove that the overall performance difference between either configuration is rather small (i.e. 1 x 16 GB vs 2 x 8 GB), at GamersNexus, HardwareSecrets and an independent Reddit review. This includes even non-gaming, productivity-related tasks like data compression, media transcoding and rendering, etc. The difference is much less so with gaming.

TL;DR: take the single largest capacity stick you can afford if you're not going to fill out all four slots.

It is more than 2x as fast as the 850 EVO, but take a closer look at the random, rather than sequential IO performance. They turn out to be not particularly different. This is what matters most in day-to-day use. Let's say your game loads in 2 seconds. Will it matter much if it ends up loading in 1.5 seconds? It may add up, but even then you'll get maybe an hour extra time per year after waiting less.

I mean, @D2 Ultima is entirely correct in saying that you're not exactly getting your money's worth. It's like buying a GTX Titan XP just so you can jack up your settings from 4096x4096 shadow quality to 16384x16384. You won't notice the difference unless you really, really look for it, and often, it is not worth the significant increase in price.

 

There is a rather large performance difference in CPU-intensive games between single channel and dual channel RAM.

 

Have you read both reviews? Excluding synthetic memory tests (MaxxMem), nearly every benchmark from CPU-heavy games (Witcher 3, GTA V) to CPU-only productivity tasks like transcoding with Adobe Premiere Pro or HandBrake shows little, or no difference between the two arrangements.

 

I saw the review. Those aren't CPU-intensive games. CPU-intensive games are for example PS2 96+ vs. 96+, BF4/BF1 64-player Conquest, 64-player RO2/RS/RS2, 32-player Insurgency/DoI, ARMA III, Crysis 3, PUBG, RTS games, etc.

 

Try BF4 unlocked FPS with single vs dual

 

Have you got benchmarks or any numbers? I'm open to changing my stand if you've got evidence, because I unfortunately don't own any of the games, and haven't personally got 2x4 1600 MHz DDR3L to test against one stick of my 2x8...

And here I was under the impression that GTA V and Witcher 3 were CPU-intensive. If the reviews I quoted don't show any difference in productivity-related matters like encoding, which is a 100% CPU-loaded task, then how come CPU-intensive games take a hit? Sounds counter-intuitive.

Edit: never mind. I found zWORMz Gaming's YouTube Channel, and he's got plenty of benchmarks. I'm sold: dual-channel is faster than single channel for CPU-intensive games.

But it doesn't affect productivity, which is really, really strange. An ffmpeg or HandBrake transcode hits all CPU cores at 100%, which would make me expect a similarly large 10-15% difference, but no, the difference is within margin of error.

 

Yeah unfortunately the non-deterministic nature of these online games makes benchmarking difficult, which is probably why most review sites ignore them in the first place.

 

Dual vs Single channel has been tested before and makes almost zero difference in games. Even "CPU-Bound" ones. This is mostly because DDR4 at even the base 2133mhz far exceeds any bandwidth requirements of any games. Even benchmarks specifically designed to test dual-channel memory bandwidth only sees a 25-30% increase (not 100% increase as most people incorrectly assume).

Here's a link with a BF4 test at 1080p ~135FPS:
http://www.hardwaresecrets.com/does-dual-channel-memory-make-difference-in-gaming-performance/3/

Those tests are done on DDR3 at 1600mhz. DDR4 blows it out of the water.
You'll notice some of the games actually GAIN a tiny bit of fps running single channel, most likely because the memory controller does so little work, that dividing up for dual channel actually slows things down.

The more important metric is latency rather than bandwidth (which is what dual channel gives you). With DDR4@2400mhz we're talking about something around 18GB/s bandwidth in single channel and 36GB/s in dual channel. Both of these numbers far exceed any typical gaming workloads. Most of what a game engine is doing is probably <1GB and the rest will be asset loading which will always be limited by storage I/O.

 

Sound point, but check out that channel I mentioned. I was surprised, because RAM bandwidth exceeds the bandwidth of everything else by so much that it ought not to matter, but it still does... And there's still the point of productivity tasks not benefitting.

 

That's completely false. Higher speed and better timings definitely improve a CPU bottleneck in many games that actually use a CPU heavily. This is not usually shown in maximum FPS, but in minimum fps, or in a CPU-bound test.

When the CPU's cache and RAM are being accessed a lot with dynamic scenarios, better speed RAM simply reduces CPU load. End of story. If your graphics are turned up and your GPUs are hitting near their limits then you'll clearly not witness any real benefit. Which leads me to your next example...

This is not a good test. Not only does broadwell's massive L4 cache help it in extremely odd scenarios (like firestrike physics tests giving it an extra 1000 points over most other competing chips, allowing it to blow past even newer Skylake chips), but they used "high" preset. You don't raise graphics to test the effect on the CPU. You lower them so the CPU *IS* the bottleneck, and then you see what the real difference is. I've had a friend with an i7-3820 and 16GB of dual channel channel memory at 1600MHz with a 970 who basically lived for high FPS in BF4. He took out two of the sticks of RAM to put in another PC and accidentally left his remaining sticks in single channel slots on the board. His FPS tanked from mostly 144 to about 90 in BF4, and he plays on a super-low configuration to get the most FPS he can get sacrificing graphics. We swapped the sticks back to dual channel slots and bam, fps back to 144 most of the time. I told him to get some 1866MHz quad channel sticks, and he put them in and informed me that he no longer dropped from 145fps anymore, no matter what he did.

That was my real-world test, and since then I've seen other people with single channel configurations who've been having weirdly "bad" performance in certain titles that are pretty CPU heavy and I've told them to switch to dual channel and that has helped.

And another thing, your virtual memory pagefile is important too. Raise it to match your RAM size on a SSD, you'll find less performance degradation as your RAM fills up. Something people don't seem to realize is even a thing. Buddy of mine was experiencing freezing in games, somewhat uncommonly but notice-able enough. Once I learned that it happened completely offline, I asked him for his specs. They checked out ok, so I asked his pagefile. He said he had it set to 500MB minimum and 1GB maximum on the SSD as described by Samsung's pamphlet for his 850 EVO SSD. I made him change it to 8GB and the problem went away entirely, and he said he'd been wrestling with it for months going as far as to reinstall his OS more than once. He simply never found a fix because he kept reducing his pagefile size after the reformats.

So yeah. RAM makes a big difference. If you're checking programs and synthetics? You're not going to see that much benefit. Most everything purely dependent on RAM speed is probably already fast enough too. But when you want to fine-tune an entire system? Nope. RAM is important. Hell, ask @iunlock about it. I got him so interested he logged CPU loads in Overwatch for three nights in a row, each night using a different kit of RAM. He noticed that bad enough timings while raising RAM speed can produce a negative effect compared to slightly lower speed and much better timings, but slight timings increases and higher speed can provide a benefit. He's on 2666MHz 15-17-17 in his laptop right now because of that testing, and we ended up appalled that dell was selling 2666MHz 19-19-19 kits to people standard if you take the upgrade on their website. 2400MHz 14-14-14 kingston sticks would blow those out of the water.

 

To add on, the fact that Ryzen's (and Threadripper's) Infinity Fabric clock rate is tied to RAM clock rates, means that RAM at 3 GHz makes nearly all usage scenarios about 20–30% faster than 2.133 GHz RAM. Not an insignificant benefit, and one that can be achieved for entirely free of cost by overclocking slower RAM.

 

I've been watching the videos and can definitely see the general jump in FPS. The issue is the videos are not terribly consistent (ie not using benchmark modes which are repeatable) and using DDR3@1600mhz.

The problem with his tests, is most of those issues are also solved with faster ram. Unfortunately I can't find the GTA5 thread where it was tested but the jumps from 1333mhz->2000mhz were quite noticeable and after 2133mhz the gains just fell off a cliff.

Your notes are all about ram LATENCY and not bandwidth. Bandwidth *was* a problem with DDR3 in some games but once you exceeded that it becomes all about latency. In which case dual channel or single channel becomes less of a factor.

As for testing minimum fps, again, that's impacted by latency. Not bandwidth.

I know it does, but it's more about the latency and clockspeed. Not whether it's dual-channel or not.

Hang on, so what you mean is he was actually running quad-channel at 1600mhz right? Then effectively dropped to a single-channel 1600mhz.

As I said, single 1600mhz IS a problem that will affect FPS significantly because of bandwidth limitations. Swapping the sticks around simply increased bandwidth which does almost nothing for latency.

 

You clearly didn't click the timestamped video in my link.

My other statements were that raising speed and latency both do not show very good results, but speed increasing with minimal latency increases will show the benefits.

Minimum FPS is determined mostly by when your CPU spikes a bottleneck at certain points in a game. The benefit of RAM is to reduce CPU load, and that is why speed bumps, OR heavy latency reduction, OR both, help improve minimum FPS, because you do not spike CPU bottlenecks any longer.

I'm disagreeing, even stating that I'd take quad channel at lower speeds/higher timings over dual channel at higher speeds/lower timings, within reason.

No, he did not have quad channel sticks and quad channel did not activate on his system. Also, he had some FPS drops into the 110-120s range as well, but "mostly" was 144 as I said. After using quad channel, however, this went away entirely. And he was even able to remove his CPU overclock and still maintain the perfect FPS count.

1600MHz to 1866MHz is nothing to write home about, and does not cut a minimum fps from being lower in the 110s/120s to never dropping. I don't know where you're getting this latency issue from. You seem to have not read what I said correctly, or not understood it at all.

 

I did. He tests Witcher3 where the main demonstration is clearly the difference between clockspeeds which is not necessarily implying bandwidth saturation.

Unless you're suggesting to me that Witcher3 is actually reading and writing out of memory at over 30GB/s and thus actually requires more bandwidth?

My god..we're going around in circles. The question was about memory channels.
I'll say it again, all that increased channel count does is increase how many modules you can access simultaneously. This means you can theoretically read or write twice as much data, per clock, to the RAM. But your absolute process time is still limited by latency and clock rate. Each channel is 64-bit wide, but if the program doesn't need to send more than 64-bits per clock then having more channels doesn't make any difference at all.

This is why games can benefit from lower latency ram even if it works at the same clock rate (and thus same bandwidth). The same relatively small amount of data can be read and written in less absolute time which means it can be processed faster. The exact same thing happens with video ram.

 

You have here in the last post a review with the P950HR vs P650HS, there are some performance diferences, you have to decide between 10 to 25% more performance but a bit heavier, or slim design and lighter but with less performance, ports and no g-sync.

http://forum.notebookreview.com/threads/new-clevos-with-max-q.805381/page-96

 

Exactly, if you write down the things you care about most see which system matches best against them given the above.

 

Yeahh, I can't really justify the 25% performance loss. I mean weight and size wise the devices aren't that different in the end. I had the opportunity to use an old Toshiba laptop today, weighting around 3kg and 33mm height and I was surprised how good it felt to carry it around, so I am 100% going with the P650HS-G.

If it is not a problem, I have a few more questions (sorry about my ignorancee)
Today I found out that there is a P65 1HS-G, what is the difference between the 650 and 651 as I can't seem to find any information?
Also do you think that i7 7820hk is an overkill since it is unlocked processor but overclocking it would bring the temperatures up, which is bad of course?

 

That 1 (or 0) signifies a cosmetic difference, there is no performance or hardware difference at all.

See this post. Open up the first spoiler in the first post, it gives you all the details for the models and the numbering. It may say P650SG, ignore the last two letters.

For the record, these letters denote the P650 'generation': SA/E/G was first, RA/E/G later, then RP/RS, and now, HP/HS: the first set was Maxwell GPU + Haswell CPU, second was Maxwell GPU + Skylake CPU, third was Pascal GPU + Skylake CPU, and the fourth, and current one, is Pascal GPU + Kaby Lake CPU.

Given that Coffee Lake will be launched a couple of weeks from now, I expect the next series to be something like either P650CP/CS or P650GP/GS.

As for the unlocked CPU—it's a good deal, you can get speeds up to 4.2 GHz while still undervolting it. But if you're not going to do anything very CPU-intensive, then best to pass it over.

 

Ahh that actually makes sense! Thank you so much!

Do you think that it might be a good idea to way a few weeks for the Coffee Lake processors to get released or it is not really worth waiting?

Given that I will work with some CPU intensive programs I will go with the i7 7820hk!

Thank you again!

 

Even if coffee lake releases today, laptops won't have it shipping for some weeks. Closer to 2 months, most likely. So in that case I'd say entirely don't bother waiting unless you can do without a unit for about 5 months (assuming that coffee lake launches sometime in october, that is; I'm not sure it will be so soon).

 

The delay is usually not that long between launch and ship.

 

We haven't even seen the MSI GT75VR new release in the US yet, and that was announced in May, and we are in August now - it'll be 4 months from announcement to availability.

Coffee-lake won't even be on the laptop maker's build list until the current crop of new laptops are ready to EOL, in about 6 months, starting in September, so March next year - 2018.

Desktop only is my guess for initial shipments.

Maybe Clevo will make a Coffee Lake LGA laptop, 1.5x TDP = 145w - I wonder if the CPU will require the addition of a 2nd PSU this time.

 

I agree thar coffee lake HQ chips will take a while to be released, soonest end of the year, most likely, next year. The U chips will be the first ones out for notebooks.
Hopefully on August 21st, intel will make the release dates clearer.

Sent from my SM-G930F using Tapatalk

 

Coffee Lake will be launched in two stages. Z370 chipset that supports CFL1 CPUs such as 8700K will be launched later this year and Z390 chipset that supports CFL2 CPUS will be launched more likely in Q2/18. These are desktop socket LGA1151 CPUs.

 

i'd be fine with a i7 8xxxHK till I get a decent desktop setup going

 

I got the P650hs-g and I haven't regretted it for a moment. I travel full time for a living, and the size has not been an issue. Thermals are excellent... if my room is a proper temperature, I usually never crack 80c, generally far below. Keyboard is excellent and it overall has a great quality feel to it. good luck!

 

Lack og G-sync and slower gpu due to max-q are reasons enough to avoid P950 in my opinion.

 

But the P950HR is 700g lighter than the P650HS. It's a trade-off.

 

Yeah, you could say that, but for gaming the g-sync and faster gpu to about the same price matters, Depends on how much gaming mean to you I, for instance, is not going back to a world without g-sync

 

I'll wait till they release this model with Coffee Lake....that might be a while

 

You mean your personal preference for size and weight will influence the kind of machine that's ideal for you?

 

Hahaha, leave it to @Meaker@Sager to put it like how it really is. I like how, despite you having a high end machine (P870?), you still understand the needs of different people, instead of bluntly categorising them as 'ignorant BGA lovers'.

 

You know what, lets test this with The Witcher 3 right now....stay tuned ! (must download it)

 

So here we are for a start : EVERYTHING STOCK, no repaste or anything... this is OOBE basically.



Stock install, firestrike : https://www.3dmark.com/3dm/22476985?



Here is HWMonitor while running the bench and downloading The Witcher 3 at same time on the default HDD.


For the record, Ambience was spot on 25C :

 

Some thoughts half way into this :

The following is a P957HR or EUROCOM Monster Q5, which retails for :

Google says that is :

So to be clear, see the following :


That above is clearly a different configuration, which is why I crossed the price out.
What I am testing is STOCK. That is the 60hz panel, which is what I'm using.
It is extra, see panels here :


Maybe adding an extra NVMe as well could be here : (NEWEGG, in CDN$)


What comes stock was a 500GB HHD, and it's pretty good, HDD have come a long way since SSD's have been taking over, it's not like they stopped getting better.

It's running cooler than I had expected out of the box, the picture when running Firestrike was amazing, quite impressed actually.

It does produce some noise, let me get my sound meter going :
(I'm in my studio office, so what I thought was LOUD, really was 12-14bBs... then again I can hear my cat breathing from 10ft away while it sleeps, my whole room is sound conditioned so I can game at night and hear the slightest audio in games for best immersion on my soundblaster ZXR on my main rig).

So main computer which is a desktop with 6950X and such and this laptop in tandem with it's fans running produces the following in sound:


Wait I upload pic from camera in sec... the shutter from the camera (which is not very loud at all) was more then double the ambience, lol.

I put the phone on my wallet to suspend it 1.5" from the table to pickup SOME vibration but not all, and positioned the mic close to where the most sound was coming from too, without being BEHIND the laptop. in some places infront of the laptop, it was less then 10dB..

You can see on the screen, and this is with fans going as in full use, no tricks.

So it's not silent, but not very loud either, these fans are relatively small. Theres 2 for the VGA and 1 for the CPU with a shared heatsink.

I will show that stuff later and my thoughts, when we try and optimise this laptop.

 

Damn it plays smooth, holyhell !

This laptop is amazing !

Temps went down...the firestrike benchmark was more demanding then this game was on max, or relatively the same.

 

Does anyone want to see anything in particular about this laptop ?

Something on insane settings, something more demanding then the witcher on max everything please, because that was smooth as hell, there has got to be something more demanding out there for this...right ?

latest TombRaider ? StarCitizen ? Otherwise I will simply move on to a breakdown/teardown and repaste and retest etc...

 

Could you fire up HWInfo and have a look at the USB 3.1 controller?

In particular, it'll tell us if it's:

A) Alpine Ridge
B) Potentially TB3 compatible via a firmware flash

I'm still surprised Clevo left TB3 support out of the 15" model when they included it in the 17" version.

 

Sure, let me try that now.

 

Alright so where do I find this info ?

 

You'll have to look under "Bus" instead of ports. If it's Alpine Ridge it's connected via x4 PCIE.

It also occurred to me, the controller may not even appear unless you have something plugged in to the Type-C port.

 

-- I had the same p950/p650 dilemma, but the better cooling/performance increase mentioned seems enough to offset the extra weight for me. Also, the p650 has room for plenty of room and options for future hard drive configurations.

 

Heya! Thanks for the benchmarking you did. Is there any chance I could ask about the temperatures after a long stress test (at least more than an hour).

Also, how do you find the speakers? Are they any good? In broad terms how would you compare it to the p650? Do you prefer it?

 

I left it on the "STRESS TEST" on firestrike for 2 days, the temps were the same, my ambience in the room was going up so I turned on air conditioning, and brought it to 23C (2 C less) and the temps never passed thier initial max on the laptop.
This means the cooling IS effective, and there is not much build up, BUT 2C ambience is a huge thing, I would have to retest at 25C to be 100% sure.

I thought for sure I was going to overheat the laptop and thats why I turned AC on. Was pleasantly surprised.

This is a way better laptop then anyone would give it credit.

I will tear it down probably tomorrow, and repaste and try and minmax things, and see if that changes thing, and I can rerun any benches for you.