2080 MXM

Really hoping I’ll be able to upgrade to 2080 SLI in my P870DM3 @Donald@HIDevolution !!! Also wondering if there will be a 2080ti MXM as specs for 2080ti Max Q did leak.

https://wccftech.com/nvidia-geforce-rtx-2080-2070-2060-mxm-mobility-specs-leaked/

 

rofl 2080ti Max Q that sounds cancerous. as powerful as it is, max Q just ruins it. until AMD doing some catch up we'll be stuck with this new standard.

 

Took them long enough...I don’t think the 2080Ti Max-Q is happening though, we’ll see.

 

I read the article I was referring to again and I was wrong. They discuss Max Q variants but the 2080ti is not Max Q. It’s the “laptop” variant but not necessarily Max Q or MXM. But gives you an idea of what we can expect once/if MXM 2080ti becomes available.

https://wccftech.com/nvidia-geforce-rtx-mobility-max-q-rtx-2080ti-rtx-2080-rtx-2070-rtx-2060/

 

Getting over excited about specifics is a bit premature at the moment.

 

I am extreeeeeeemmmeeellllyyyyy sceptical.

I hope we get real RTX cards, but looking how even the desktops cards are a disaster i HIGHLY doubt we have anything even closely resembling the real RTX cards.

 

I mean I would hope for a mobile 2080ti to be mxm only, but it's still unclear.

 

perfect timing to get the wallet ready for a 10 series MXM card

 

Mobile 2080ti was misinterpreted from a Zephyrus geek bench score with a 2080ti in an egpu box, but I wouldn't put it past Nvidia to come up with some new naming chicanery attached to sub-TU102 chips (like how we have i9 on Intel mainstream socket)

Power limits attached to the chips is what will matter. Turing desktop chips suck more than Pascal even without the tensor/RT cores in use so somethings gotta give

 

Nvidia moved away from that in the 10 series and I don't see a need to go back to it.

 

2080ti prob wont happen, just needs too much power than 1080, unless nvidia go crazy again with all the OEMS and we see a 250 or 300w card then we'll need some redesign and adaquate cooling, not my thing atm cause desktop zen2 looking gud

 

Looking good for multi thread on zen, still not clear on single thread.

 

What is the current wattage on the standard 1080MXM put into P870's? like 200W right?

If the 2080 can remain at that wattage, with a bigger die (More surface area to connect with the vapor plate) I don't see how colling will be any less. Am I wrong in this assessment?

I am optimistic about the 2080 cards, but I wonder how they will handle NV-link as the bandwidth is much higher now.

Even if it is a 15% to 20% increase, I will still fork over the money.

 

So, fake news?

 

2080Ti yes most likely...

2080 MXM
2070 MXM
2060 MXM

expect next year

 

http://forum.notebookreview.com/threads/rtx-notebook-possibly-debunked.826638/
no

 

Come back Jan 15th and say the same thing

 

Let's wait and see about CES, not long now.

 

2080 alone is a miracle on an MXM board, I would not hold my breath for a 2080 Ti .

What I am wondering, will Clevo cheap out regarding quality and will they use better quality Samsung rams or cheaper Micron rams?
If Micron was able to fix the manufacturing problem and the new laptop cards are getting v2 version of Micron rams, then no problem, but if Clevo purchased a left over batch of unsold Micron rams which may still have a few bad chips then prepare to get a free game with your brand new Clevo laptop

 

Rather expect all the remaining batch of bad vram from Micron will go directly into Dells gaming books @Vasudev

 

Returns cost a lot more on notebooks than GPUs.

 

What's that in the picture? A Matrix like screen or Pacman?
I don't think Micron was bad at all! I really prefer SK Hynix VRAM chips since they OC very high unlike Samsung VRAM. Samsung might not do very high OC but performance is same as SK Hynix High OC even with mild overclock and voltage.
AW/XPS team never released a VBIOS update that caused Micron VRAM bug and some of them resolved it thanks to a member who uploaded the new VBIOS with the fixes from Dell factory.

 

It's called Artifacting

 

Usually its red in color. I did have those when my Mem OC was too high.

 

That's why it can be nice to have the desktop as beta testers

 

NVidia never openly admitted but that issue was because 2 things:
1) bad batch of Micron rams. Some chips were defective and the reason why the issue was more often with RTX 2080 Ti cards is because those use 11 chips while the RTX 2080/2070 use 8. Even if only 1 chip is defective it will make the cards unusable. New tech and Micron
did not have a 100% perfect manufacturing process because they rushed it in order to get ahead of Samsung.

2) soldering the GDDR6 chips is a bit harder than previous gen so in some cases the above issue was due to bad manufacturing process and at least 1 chip was not soldered correctly. Samsung GDDR6 chips are better quality than Micron, however if the manufacturer cheaps out on the soldering process, even Samsung chips can produce the above mentioned issue.

We shall have to wait and see. Is Clevo stupid enough that in order to save some Canadian rubles, will they:
a) use cheaper Micron rams than higher quality Samsung rams? (on desktop side not even cheap Chinese brands like Inno3D use Micron anymore)
b) cheap out on the manufacturing and testing process to see if all components have been soldered correctly?

Unfortunately Clevo is notorious on cheaping out and saving money anywhere they can (that is why they are cheaper than other brands).

But I hope they won't repeat the problem which happened on desktop side.

 

Maybe Nvidia QA skipped checking of Micron,Samsung and SK Hynix VRAMs which they got in bulk.
Micron aren't trash. They don't OC well unlike Sam and SK Hynix, but for nvidia who sells GPUs in bulk wants GPU to just work since majority of users don't overclock.
I don't what Clevo will use in their MXM version of RTX GPUs.

 

As of now RTX cards are using Micron or Samsung rams. However due to the problems with Micron memories, most manufacturers now only use Samsung rams.
SK Hynix should be having GDDR6 memories too, but as of now no RTX cards using SK Hynix memory has been released to the public.
I did not say that Micron is trash as a brand, however their GDDR6 manufacturing process had issues (which may have been fixed by now). This was probably because Micron rushed things to grab the huge deal of "exclusive supplier" with NVidia.
However after the issues rised even Nvidia dropped the deal of only using Micron rams exclusively for their RTX cards and moved to another partner and that was Samsung. Because of this most manufacturers have also moved to Samsung chips.
Micron is good, but they screwed up with GDDR6+RTX cards big time and it will be hard for them to get their partners trust back. Maybe they corrected the issue, but as of now I do not know anyone who moved back from Samsung chips to Micron chips.

 

Everyone screws up sometimes. Just like Samsung with their batteries a few years ago, Nvidia with the soldering job on Tesla, AMD with Bulldozer etc. Micron will feel the repercussions for now, but i'm sure they will fix the issue and regain trust over time. Samsung alone can't produce enough GDDR6 for all our GPUs (especially if AMD's next lineup uses it), so we will definitely see Micron/Hynix GDDR6 in the near future.

 

I can run +750mhz OC (+1.5ghz) on my samsung vRAM

 

Do you have same 980M from Dell or MXM?

 

GTX 1060 ran easy +750mhz clock
GTX 1070 ran easy +750mhz clock as well.

All MXM cards.

I could try and see what my MSI GTX 980M says tho if you want to know.

 

The GDDR5 on the 9xx cards also loved a little extra voltage, got them even higher.

 

300MHz is my limit.

 

Where was the proof of this?
I never saw this proven anywhere.

 

I think Steve Burke from Gamer's Nexus who asked for samples of dead RTX cards inspected the cards and found most faulty cards were using Crucial or Micron VRAM.

 

That's still not proof.
You need hardware probing tools. You have to find out what failed and why. You can't just use voodoo science and say "oh if X then Y therefore X caused Y".

Sorry if that's rude or "toxic", but I don't tolerate or respect people talking out of their butthole about important things (i'm not referring to you, relax). I want evidence, proof, and scientific testing. Otherwise I don't believe what anyone says.

 

Yeah its okay. I think a simple resolder might solve those issues or even reflow. We might need an expert like @Khenglish to give us hints what might have caused those issues on RTX if he actually bought one for testing.

 

has nothing to do with vram, capaacitors blew up. Ironicly Ngreedia seemed to have done the same mistake as Eurocom MXM cards. Aat least from all the RTX i've seen that died.

 

+550
https://www.3dmark.com/fs/17679936

starts to glitch out at around 600, so limit is probably around 575mhz on my 980M.

 

I think my 780m went up to 1750-1800mhz.

 

Low score for the clocks.

https://www.3dmark.com/3dm/23330279

 

You do get reducing scores generally with GDDR5 before it gets unstable due to re-transmits rising.

 

This is my highest score on older drivers https://www.3dmark.com/fs/14850806
On newer drivers I don't reach that score.

 

Yeah, and if he increase core clocks there is no guarantee that he can run mem at that speed. And it's amusing seeing his locked down Skylake BGA killed by its predecessor


And the physics score ain't near the peak... https://www.notebookcheck.net/Intel-Core-i7-6700HQ-Notebook-Processor.149420.0.html

 

I doubt the VRAM can run at 6110MHz stably even without a core OC. That's why the performance dropped due to GDDR5 error detection kicking in. The memory on the 980M does not overclock well at all unless you volt mod it.

 

its not about the score. It's about the vram OC. The notebook I used is insanely bloated.
Also the core clock is slightly OCed stock due to custom vBIOS

Mind telling me what drivers you ran on?

 

Mine did close to that before the vrm additions added interference. Volt mods becore that would have likely seen it closer to 7ghz like my 780m cards.

 

I was using a bloated WIn 10 with all nvidia services and other junkwares in background.
The same driver as mentioned in 3dmark result 388.71 which caused core clocks to be stuck at 450MHz. If the clocks were alright I should have got 10-11k on Firestrike overall score.

 

I really doubt that. You would’ve scored much lower had you actually been throttling to 450MHz. And you weren’t getting 10-11K overall unless you were running an impossible OC, like 1500/8000.