New Clevos with Max-Q?

Odd, I found review of MSI GS32 updated with Kaby Lake CPU : https://www.notebook-center.ru/overview_1779.html
And there's also Youtube korean review :



But there is no announcement from MSI regarding this. Their global website also doesn't show this updated product.

 

There's just too many things working against the practicality of eGPU now.

High cost, lack of bandwidth and more importantly recent Pascal mobile GPUs being VERY capable for the price. A mid-ranged 1060 laptop like the P650 can play most games at 1080p/high/ultra because game requirements haven't really gone up much. They also took too damn long to finalise the spec and sort out drivers. Not to mention lower end 1050/ti models will play most of the popular games with ease (Dota/LoL/CSGO/Overwatch/etc).

 

Agreed. But eGPU real benefit is actually for sub 14" ultrabook / business / non-gaming laptop.
You can go to work/school bringing your small thin slim ultrabook (such as HP Spectre X360), then comes home and plugin the eGPU and play the latest games.
Yes you can always have gaming desktop + ultrabook, but having 1 device for everything is better for some people.

 

I have a good feeling that the eGPU may get a comeback after people realise what a Max-Q GPU is . Anyways I dont understand why eGPUs are 'too late', I absolutely love the idea of an ultra thin long lasting laptop being able to transform into a gaming laptop whenever I like. Whats the problem that is holding eGPUs back currently? Isnt the thunderbolt port fast enough to provide a connection between the GPU and the laptop? If so I would really like to see manafacturers making a laptop with a eGPU compatible port to increase efficiency and speed.

 

To a degree, but at the same time, you can get 14" ultrabook-like machines with decent GPUs in them anyway. Machines like the Aero14 already prove you can shove a massive battery in there and get 8 hours of battery out of it which makes it suitable for work. Lets also not forget, Ultrabooks which actually have a TB3 port tend not to be cheap. A Spectre X260 where I am costs as much as a decent 1050ti equipped machine or even a minimum spec P650HP.

On top of that the bottlenecking is a very real problem. Either bottlenecking due to a weak ULV cpus in ultrabooks or due to Thunderbolt3 only having x4 PCIE lanes AND it has to share that with SATA/Audio/Networking/USB back to the CPU via DMI. If you have a ULV only, there's not much point putting anything more than a GTX1060 in the dock anyway as the CPU will be the bottleneck. If you've got a 7700hq then DMI becomes the problem and your upper limit is around the GTX1080.

I'm not saying the concept or idea itself is bad, they just completely missed the boat with the implementation.

 

Razer Core eGPU

"In real world performance (not synthetic benchmarks), there seems to be little benefit to using even a 1080 in a Core over an integrated 1060, and in some situations the integrated card actually performs better."

 

Clevo p70hs with gtx1070 and thunderbolt3
http://www.tw003.com/3/35332.html

http://forum.notebookreview.com/threads/clevo-pa70hs-g.805708/#post-10542219

 

@sicily428 FYI PA70 is not MAX-Q ;-) HS and HP are the same codes as other chassis.

 

really!? it is great. thanks

clevo is great!

 

Could you use dual tb3 cables for egpu and double the bw? How about driving the internal display through a separate dp cable?

 

Some of the connection still will rely on the CPU that's what I meant, a U series CPU is likely to bottleneck performance even with an eGPU.

 

Basically there was a huge difference between mobile and desktop GPUs in the Maxwell generation, and a way to attach a Maxwell desktop GPU to a laptop would have been a major improvement (and sorta worked with those little 13" MSI docking station models, docking clamp problems aside). But all those machines had a big proprietary connector and so you couldn't do it with any old system (lack of bandwidth). TB3 wasn't really much of a thing until Pascal machines, which made an eGPU less necessary. Then consider that TB3 was mostly available on machines that already had about the best GPU you could get. It's making its way to lower performance machines, but not quickly. So the situations where it's been useful have been fairly few comparatively.

 

wow, 17" slim laptop with 3.4kg weight??

I agreed.
But again, I think the appeal of eGPU are for business laptop owners.
Let say you bought a Thinkpad X1 Yoga priced at $2500, it's great for work, presentation, have amazing battery life, body built like a brick, can withstand firestorm etc.
But it only comes with Intel HD 620. eGPU solve this.

Scenario #2 :
it easier and cheaper to upgrade GPU inside eGPU case rather than buy a new laptop.
i bought my laptop with 980m SLI in late 2015, it's still powerful enough, but 1 year from now, I probably won't able to play VR games with it.
everything else about the laptop is still great and fast, only the GPUs are absolute. upgrading MXM module is a no go and even if it can, it cost an arm and a leg.
eGPU solve this.

I'm fully aware that there's performance hit if the graphics from eGPU are relayed back to the laptop screen instead of output it into external monitors.
but i think once the technology matured, it will be great... for consumers that is.
for manufacturers and resellers... might not look so good

 

A fair point, and I didn't mean to say that nobody would find utility with it, just that they're not taking off because they overlap too much with solutions that are better for most consumers.

 

scenario #3
You have an old acer vn7-592g with a gtx960m or something like that and you could use an egpu for gaming with a gtx1070.

scenario #4
You need an ultrabook 1kgs/1.5kgs for traveling (lightweigt+long battery life) like dell xps 9360 and you don't want a desktop cpu or a gaming cpu. you could use an egpu

scenario #5
You need a laptop with a quadro gpu for cad3d in solidworks/catia/creo and you would like a lightweigt laptop. You could choose a msi gs63 7re or a vn7-593g and you could use a egpu+m3000 or p3000

 

(Max-Q) ASUS Zephyrus - What's Inside the Thin Gaming Beast?

 

Half the power consumption and only lose 10% performance . Nvidia's logic! or should I say magic ?
Yeah , just like how you - Nvidia claimed GTX 1080 would be twice as fast as GTX 980 last year . Turn out it wasn't that fast ...
# $h!t Manufacturers Say
While GTX 1080 Max-Q is only 1~4% faster than GTX 1070 in benchmark
Real GTX 1070 mobile can be as fast as the desktop version . This is how you should do it , Nvidia :


If I wanted a thin gaming laptop , I would rather buy an ultrabook under $1000 and this Gigabyte eGPU + GTX 1070 bundle for $600 :
http://appleinsider.com/articles/17...xternal-thunderbolt-3-enclosure-in-600-bundle

 

It just boggles the mind how Nvidia weedles it's way around trying to present Max-Q in best light from all angles at the same time, even though those are all mutually exclusive - you can't have them all at the same time - features.

The quiet mode with whisper mode frame limiting to 30-40 FPS, while the monitor is 120hz with G-sync at 120 FPS.

I am assuming these benchmark numbers weren't run with Quiet or Whisper, or even Max-Q mode's enabled - it's mentioned you can disable all of it to run at full potential of the GPU / CPU - which has to be a turbine jet engine sounding thermal and power throttling mess in that tiny chassis:

Notice how not one game was able to reach 120 FPS even on a 1080. We need "real" 1080 / 1070 / 1060 laptop results tested at the same game settings for comparison - strangely not provided - the #1 comparison everyone wants.

Notice how it says "*with* Max-Q Design", but it doesn't say *using* or *enabled* Max-Q design, I am assuming these numbers being fairly high are with all the Max-Q controls disabled. Otherwise why not show results with both Max-Q enabled and disabled in the same laptop.

This whole Max-Q thing is going to be a great litmus test for reviewers to see who is reporting straight - finding the Emperor has no clothes, vs. the reviewers that report on how fine the Emperor's new cloths look

 

You'll want some popcorn for when those independent reviews start showing up. No matter what the verdict I predict some 'spirited' discussion.

 

Unigine Heaven 4.0 Extreme Preset (our own figures)

Standard (not MAX-Q) GTX 1060 = 70.8 fps

Standard (not MAX-Q) GTX 1070 = 109.1 fps

Comparing the fps in @hmscott 's post image;

View attachment 146250

Rise of the Tomb Raider seem very close to non MAQ-Q fps, The Division is 13-15% lower than standard GTX versions, Hitman seems to run 40% faster on MAX-Q than on the normal GTX versions. So there's something rotten in Denver with those figures, they are all over the place.

I'll leave the conclusions to you guys!

 

Part of the problem is what I alluded to in my post that we need them "tested at the same game settings for comparison", because any testing without settings sync'd to the same are going to come out wonky - ok I Ieft out the "wonky"

Even the Unigine tests can be skewed with a little offness in settings from standard. I've run into this with people running at different Unigine settings all of a sudden thinking they got a "magic" GPU

 

Ungine Heaven score, Oh'well the number talk by them self



From NBC ("Normal working" Gtx1080-Laptop)

 

Check Nvidia's stated graphics settings again, Ultra in Division, but Very High in RotTR. "All settings Maxed for Hitman, SMAA"... again - every little setting different can make a difference.

So Max-Q RotTR is even worse than the Ultra scores you highlighted, Very High in that same list is *waaaaay* higher.

Max-Q performance has to be pretty bad to require lowering RotTR quality from Ultra to Very High on a laptop with a 1080...

 

Ungine Heaven for 1070-1080 (Laptop) by NBC

 

I mean I think it's pretty clear that when considering performance of a MAX-Q laptop that we just move 1 tier down.

1080MQ = 1070
1070MQ = 1060
etc.

The 4% or whatever difference is not really a "difference" in any meaningful way. For now, we are all but certain of the way MQ laptops behave performance wise.

For me, personally, now the only swing-vote is cost. If a 1070MQ laptop costs the exact same as a 1060 laptop, and offers within 4% performance, then all things considered, I see them as standing on equal footing. The principle of it doesn't really get my goat.

But what are the chances of a 1070MQ laptop costing the same as a 1060 laptop? If I was a betting man, i'd say nil.

 

The more interesting one is the 1080Q vs the 1070N (because that's at least the "theoretical" use case for Max-Q).

Based on NBC, the 1070N averages 78 on The Division and 92 in RotTR. So that's a decent bump there.

Hitman as a benchmark is super wonky though. Mine varies by up to 10% every time.

 

The Asus Zephyrus price is already quoted as being MSRP of $2700

That's right there firmly in the 1080 pricing territory, way above what a "real" 1070 costs.

That's why I've been calling out the Max-Q 1080 laptops as being $1000 more than their true performance value.

 

Wow, those "normal" 1080 / 1070 scores are hugely higher than the Max-Q scores!!

 

Here's mine by the way, stock vbios Clevo 1080

The asus one in the chart @Papusan mentioned ran 99Mhz higher than the Clevo (1681 vs 1582). So here's a bench with +100Core

Adding +500Mhz more to the memory increased the fps by around 5 (+100Core/+500mem)

PS: The screenshots were done to have an apples-to-apples comparison. Normally, we used to do ultra preset but at a 1080p resolution. The fps drops by around 40 e.g 150->110 making the numbers much lower compared to using 900p.

 

So why post them if they're not comparable?

EDIT: for clarification, the Nvidia table and NBC benches are done at 1080p

 

Are you sure the NBC benches were done at 1080p? Here's what I'm having at 1080p extreme. It's impossible to reach 140fps for a single gpu 1080N at 1080p on Unigine Heaven.

 

They often don't add it to the title on recent benchmarks I've noticed. But if you look at some older pages you can see they use 1080p which they then list with the Comparison graph (confirming it's not a 1-off bench):
https://www.notebookcheck.net/Nvidi...sktop-Review-Pascal-has-arrived.165500.0.html

They do mention running Tessellation at Normal. No mention of AA though. Based on my benching (comparing my P650 to theirs) it looks like it'll be Off or 2x.

 

For the records. Some answers regarding dif settings. Ultra vs. High in 1080p
NR.1 Ultra.


NR.2 High.

 

So unless there's some other features specific to the models, it's basically paying for form factor (including noise in form factor)

 

^^^^ That's the whole point of MAX-Q though, thinner chassis with limited dB cooling systems but with the same GPU models which are found in the current "thicker chassis".

That's not my take, it's the MAX-Q message and when I say "same GPU models" I don't mean same performance.

 

Yes, a bit like cars where the same engine is tuned down and fitted with a smaller gearbox to etc to target a different size of vehicle.

 

I've seen impressions of Asus and Acer machines, any on Clevo MAX Q ones?

 

clevo p950hr
https://www.notebookcheck.net/Ultra...us-is-Razer-s-latest-competitor.224654.0.html

 

To be honest , I would rather buy a super thin Gigabyte Aero 15 (GTX 1060) than this crap , at least that GTX 1060 is real . GTX 1070 MQ (a.k.a fake GTX 1070 or GTX 1070 in name only) = GTX 1060 , no more no less .

 

The TDP on the previous machines could still be altered. This is formalising the process and widening the range.

 

MAX-Q on a Clevo? It's an abomination. As a matter of fact the whole idea is an abomination.

 

It seems a little counter productive considering the image Clevo has made for itself in the realm of performance. In other OEMs it makes more sense, even if it's not ideal.

 

The Dell guy in this video is speaking English... starts at 00:37

So apparently the 1080 Max-Q in the Dell product is set to 110w, which is the "fastest Max-Q 1080 laptop out there", "other Max-Q 1080 laptops are so slow in comparison that they may benchmark slower than a full power 1070", quotes from the Dell guy...

So there you have it, confirmation from "Dell" that "other" Max-Q 1080's are as slow or slower than a full power 1070 laptop

Dell and Alienware at E3. GTX 1080 and 1060 Max-Q [ger.]

 

They just need to implement MAX Q seriously. Like make it seriously thin but well cooled. No 1080s in .6" chassis's with a TDP of 100W with like a 180W psu.

I'm mostly just opposed to these chips being called 1060 1070 and 1080, and the pricing. A 1080 Max-Q should cost like 100 bucks over a 1070 performance wise, but it won't.

 

Nvidia 2016: no more low power x70 sold as x80M mobile GPUs! Buy our new full power desktop parts!
Nvidia 2017: buy our new low power x70 GPUs we call x80MQ!

Gotta laugh really.

 

Exactly, going backwards. facepalm

 

Spoiler

 

Getting rid of the machine series was always silly IMO.

 

Yeah, Exactly. Dell claim they use max powered graphics vs. other Companies who only push out graphics with crippled power limits Maybe he is the author of this statement as well?

 

Oh no, I smell something and apparently it ain't a good one too.