Clevo 2019

Ah, an MXM upgrade path, it will be interesting to see how those 20 series MXM GPU's are priced - gonna be hard to justify desktop level prices for subpar performance due to power throttling. That probably won't stop them being overpriced though.

 

If prices are not reasonable according to the performance upgrade I won't buy anything. 1070 is still doing nice things.

 

It seems to me that tdp is embargo now. They must give higher tdp limits otherwise these cards will not overcome the old 1080. the update will be pointless, nobody will buy it.

 

Even with a difference 70 and 80 class cards are usually designed to let you skip a generation.

 

Yep, maybe. But the problem is the ability to upgrade two generations after.

 

Ever undervolt a desktop 1080? I went from 250w to 150w without a real dip in performance, on the order of losing 1-3 of my frames.
Thats more of a power loss than bumping down 215 to 150w. Also keep in mind that this isn't anything but tdp (Thermal Design Power) meaning that it could be a card that draws 280w but only needs 150w of heat dissipation.

 

That really is the question. Can the TDP throttled RTX mobile cards boost and hold it beyond their factory spec?

 

I hope it will be as you say + the new turing chipsets are much bigger than Pascal
so it will be easier to collect heat from it and spit out through the cooling system

 

At least there's the advantage of better heat dissipation over a larger surface area.

 

Only if it's all active if course, the shaders themselves will still be similarly thermally dense. A bit like the igp not helping cooling on CPUs.

 

So where did that other 130W of heat energy go?

 

Isn't heat energy, simply enough. The 280w with a 150w tdp part with a larger area simply doesn't need as intensive thermal solutions as the part that is 200w and 190w tdp because it has more surface area. Think of an elephants foot vs a high-heel. If you have 200mm² putting out 70c vs 500mm² doing the same, you don't need as much concentrated heat solution, thus the TDP is lower.

 

RTX cards are now listed on taobao too, not sure when they'd ship though.

 

What to you guys think, when can we expect the first RTX laptop reviews to pop up?

 

I’ll have mine up post embargo on the 29th.

i9 + RTX 2080

 

Lol. This is nonsense.

 

That's not even remotely how thermals work...I think you're getting confused with the concept that coolers can have a "rated TDP", which is not really good.

TDP is literally energy. The amount of energy created in heat is not related to surface area, it's related to how much energy goes in and how efficient the chip is.

What surface area does do, is allow you to transfer that heat away more efficiently. That's because different materials have different thermal transfer properties in relation to surface area (as well as direction in some cases).

 

Buildzoid (actually hardware overclocking) did a video recently on this topic of "thermal density" comparing CPUs from Sandy Bridge thru CFL-R (9900K)



Assuming the whole die area is 'active', (even tho that's really not the case yet with turing's dark silicon aka tensor/RT cores......) the calculations of thermal density of turings vs pascals are why we're assuming turing will be cooler *IF* they are putting out the same amount of heat on the same heatsink

 

They start shipping on the 29 january.
A sellor confirm accepting shipping directly with SF express overseas.
I will see what are the us-eu price to compare where to buy it.

 

Can anyone tell the ram brand from those picture? Is it Micron or Samsung ?

 

+rep.

A Watt (W) is a unit of power. 1 J/s (Joule per second) = 1 W (Watt).

So for twice the wattage, you get roughly twice the heat output.

Of course, the 2nd Law of Thermodynamics dictates that nothing is 100% efficient, so you may not get ‘exactly’ twice the heat output, but you will certainly get more.

 

Coolers will have a thermal density they can cope with and a thermal max. A bit like power and current limits on a chip, then one or the other can be the restriction on performance.

 

Going back talking about the new Clevos, is it just me or on online shops you can preorder any model but the PB50/70? Incidentally, that's the notebook I'm most interested to, and there's not even a decent picture about it...

 

Concentrated heat is much less effective to move than heat spread out across a board. I've actually replaced not only the copper heatsink on both a p870dm with indium and silver, but also have experimented in that type of vein with larger and smaller loads to find the crosspoint between wattage of tdp, rating of heatpipes, and so on and so forth. A heatpipe of 95w theoretically would move more heat than three 30w heatpipes, but in practice, those 30w heatpipes remove more heat because of their area and the contact they make with the heatsink/baseplate.

Think of thermals less as a wave and more as liquid. 100psi in 3 3" tubes moves water faster than 1 9" tube at 33.33-psi. heat removal is definitely a complex situation because not only is the heat of a part a concern but also is the materials that act as heat transfer. Indium, for instance, is about 3-4 times as effective as silver solder, silver is a little more effective than copper.

What I did find out is that the copper on the heatsink of a 1080 is less a material to pull heat away, but more a material to spread that heat out so the heatpipes were more effective to remove it. I assumed that a CCHP would maybe work better but was surprised to see that in order to pull away the amount of heat that was required, a larger surface area was needed across multiple heatpipes, which is why the heatspreader was so important. Essentially, then, a higher density heat needed more surface area, thus required a heatspreader, in order to act as an efficient transfer of said heat. With a larger die, whether it is active or not, that becomes less an issue, and can effectively use a lower amount of working materials to accumulate and transfer that heat, lowering the TDP.

The TDP does not mean power used, rather power dissipated.

 

Looking like an NDA until the 29th.

 

Yes, exactly. A RTX can consume 240w but only put out 150w of heat.

 

That's not want I meant and also physically impossible.

Watt/mm^2 would be density of heat.

Energy cannot be created or destroyed so all energy consumed by a chip will he emitted as heat (since they dont produce noise, light etc)

 

Yeah, but I thought that ALL models were under NDA, not just the PB5/7... too bad then, I'll just wait

 

delete

 

Indeed, we're basically saying the same thing with different words. Heat concentrated in a small area is harder to pull away.

Yes, I'm also very careful not to say that TDP = Power. However, the relationship between TDP and power consumption is simple and small enough that they're often just interchangeable.

In the case of modern Nvidia GPUs they are using a new "TGP" value (Total Graphics Powerdraw). In this case TGP is the hard wattage limit enforced by the vBIOS. So TGP does indeed equal max power draw in these cases.

The TDP itself doesn't go up or down, as it's generated by the chip. What does change is how much of that is absorbed by the cooling system. You could just glue a big block of aluminium to the chip and the TDP would still be the same.

The chips do emit some non-heat output, otherwise the GPU would literally do nothing but generate heat, but we want it to generate graphics
The output energy is just miniscule in comparison to the amount of power required to switch all the transistors.

 

It is true that heat concentrated in a smaller area is harder to dissipate (fast enough), but a bigger surface doesn't necessarily translate in significantly less heat density if there is a significant increase in nr. of cores (of the same size) that are mostly responsible for the generated heat or in the case of CPUs, the iGPU occupies a significant portion of the die, but most of that surface doesn't help with heat dissipation (only part of it, that is near the CPU cores) when the iGPU isn't used.

But all this talk about heat to surface area ratio, isn't relevant when discussing the TDP, because the TDP was already calculated for that specific SKU. Nvidia (and Intel in the case of CPUs) also give design guidelines which are to be used in conjunction with the specified TDP, so if the OEMs screw up something, it is their fault, because they have all the necessary data to design a proper cooling system.

The HS/cooling system designer is going to input the provided power value when calculating the HS's potential, not a lower power value. Even though the higher surface contact area might improve the efficiency, the rating of that HS is still going to be at least the same value that the chip has, because it is actually dissipating that total amount of heat in watts (or joules per second).

While the specs of the chip remain the same, the TDP rating of the cooling system can differ (but still needs to be at least the same value as the provided TDP).

OEMs have the liberty to design a more efficient cooling system if they want to, and they usually give some thermal headroom, especially for the OC-able SKUs.

Just like a lawyer interprets the law, an OEM can also interpret the given guidelines, so they can design a cooling system that is barely at the limit and still be within specs due to the complex throttling mechanisms or the high tolerances in some cases (for ex. the +90C avr. temps that are considered within specs).

I agree with Meaker, we can assume (for simplicity) that all of the energy that is used to switch the transistors, ends up as heat (the insignificant parts that end up in other forms, for eg. radiation, are not relevant). Basically, that part of the GPU is just like an electrical heater that uses the energy in a more useful way.

Now of course that the GPU uses a small part of the total energy for communication, but even so, the significant power values that are not converted to heat, are left out when the TDP is calculated:

https://www.nvidia.com/content/dam/...ure/NVIDIA-Turing-Architecture-Whitepaper.pdf

See ^^^ page 15 and the bottom note:

For Intel CPUs:
https://www.intel.com/content/dam/w...heets/8th-gen-core-family-datasheet-vol-1.pdf

Start ^^^ from page 87 "Thermal Management".

https://www.intel.com/content/dam/www/public/us/en/documents/guides/xeon-scalable-thermal-guide.pdf

Start ^^^ from page 39.

 

Yep, power consumed = heat out. You can ignore everything else.

Data is not energy so does not count

 

Alleged marketing material from Terrans Force (Clevo's child company) alleges RTX 2060 is about 90% of GTX 1070 in laptop context
(EDIT: it mentions T7 model which I believe is the new P970).

 

RTX2060 not being at 100% GTX1070 it's pretty bad IMO.

 

Yes, that would leave almost no reason to buy a 2060 over a cheaper 1070 laptop, all things being equal.

 

That graph does not even match the numbers lol.

 

glad im not the only one who noticed haha

Sent from my Xiaomi Mi Max 2 (Oxygen) using Tapatalk

 

*Graph not to scale

 

Another pic of the alleged new P775 cooling

 

Holy orientation batman!

There, that's better:

Spoiler

Is that silver part of the GPU heatsink (left) nickel plating or aluminum? I'd feel a lot less secure using liquid metal on the GPU if the latter.

Here is the previous heatsink for comparison. It's all bare copper.

Spoiler

 

It looks more anemic compared to the heat sink of Alienware 51m

 

Definitely aluminum. Have seen too much of clevos aluminum heatsinks to not know it lol

 

i don't know that site but they sell the 2080mxm for 2000$ lol... at this ****ty price i prefer to sell the latpop and built a mATX

 

New P750 cooling, deja vu on that 3+U+3 layout.

 

Bad graphs set me off

Certainly looks that way to me

 

Looks like a 2060. Curious to see how this handles temps to the previous 2 versions.

 

Huh, the P750 has more heatpipes than the P775 this time around.

 

How are you counting that?

 

By the number of heatpipes?