Jump to content

Mobile Geforce GTX980 incoming...


chap

Recommended Posts

As usual we as Clevo community will probably have to take care of the P7xxZM (especially for the g-sync part), P570WM, P37...

As much as I'm disgusted with the way laptops are going, I feel sadness when I consider tearing myself away from them for a future system because of this community. Even though as Mr. Fox quoted there's quite a few people who seem to be willing to accept the bad direction and move on, people like you make me want to stay.

I feel terrible about the tech industry on the whole these days. It's not hard to give us something good... but they try SO HARD to give us things that are bad.

  • Thumbs Up 1
Link to comment
Share on other sites

I would have also preferred 990M or even 985M over 980. Just doesn't sound right without an M being in a laptop.

It's more that it sounds wrong. The Mobile GTX 980 not being the 980M confuses everybody, and I think the reason is that nVidia does not want to make a new line. They probably want Pascal to be 1000 series, or they might forego the whole 1000 deal and make new GPUs. That's probably the reason their desktop cards are in such a state right now too. Kepler had a disgusting lineup for the consumer on launch, and it was fixed with the 700 series. The 670 was too close to the 770 for business, so it was delegated to an obscure 760Ti card for OEMs (yes, that EXISTS, just like the 192-bit mem bus GTX 760 with 1.5GB/3GB vRAM which is wholly inferior to the 3GB 660Ti models) and the 760 was designed to combat the mismatched memory issue of the 2GB 660/660Ti while (adding more ROPS and memory bandwidth) as well as making a "weaker" card for midrange so people wanted to buy the 770.

But look at what we've got now... a midrange *60 card which has a pitifully slow 128-bit memory bus... unable to even match the GTX 285 from what, seven years ago? Even if OC'd to 8000MHz effected memory clock; just barely able to match a 660Ti's bandwidth? Unable to match my STOCK 780M's bandwidth? And this is taking into account maxwell's memory bandwidth improvements too, even, as a 1.15 multiplier on end-result bandwidth for an extra 15% benefit? It doesn't matter if the core is good (which it isn't for that tier of card; considering it's considered GM206 (even though the 965M which is the same card is GM204) and there's no 960Ti, and the fact that it's 1/2 a 980's cores) if the memory bandwidth can become such a limiting factor. I always say vRAM isn't a limiting factor on a GPU at all, but memory bandwidth is, and that card should be 192-bit at worst.

And then the lovely 970, marketed as a 256-bit card to this day (even though it is effectively a 224-bit card, as the "memory bus width" is found by the number of memory controllers working in tandem added together, which is 7 x 32-bit in this case) and that vRAM problem in itself is absolute aids for anybody who is affected (even if 95% of people are not)... it should have been a 3.5GB card and it'd have been left alone and we'd have had a perfect little card. But no. nVidia screwed up, and tried to hide it, and refuse to admit any mistakes. And they probably won't until Pascal is out; because admitting that you made a mistake on a current-gen still-selling non-recalled card is corporate and PR suicide.

Then the 980, the only actual good card in the lineup... still tiered too high for its own good, and costing too much for its own good ($520+ for a 4GB midrange card when a 6GB top-end card can be found for as little as $120 more, and is 40% or more stronger?) . If this card would just drop to $420-450 we'd have no problems. But no. nVidia.

And then the 980Ti, the only actual well-done, well-priced, well-made card in the WHOLE lineup of maxwell desktop cards. ONE card that has the whole package. One. Lovely. Nice choices, nVidia.

And the Titan X. Overpriced garbage that tops out in OCing around where the 980Ti does; $350+ more for an extra 6GB of vRAM and some extra cores that don't help due to its inability to OC as high as its younger brother (except maybe for people like john, but he's an exception). It doesn't even have a double precision block to make itself worthy of the "Titan" name.

And now the crown jewel. The "Mobile GTX 980" not to be confused with the "GTX 980M" not to be confused with the "GTX 980". Still with low voltage vRAM.

  • Thumbs Up 2
Link to comment
Share on other sites

Still with low voltage vRAM.

Except for the 6x0m series, all GDDR5 mobile GPUs run the memory at the same 1.5V as desktop cards. The reason why they clock lower is because there are major crosstalk issues on the Nvidia 256-bit mxm board. AMD cards have a thicker pcb to combat this. While 6ghz is a struggle on any Nvidia card, I have a 7970m which does 6.21ghz with no mods 100% stable. I overvolted another and was running 6.4ghz. Keep in mind these are 3 year old cards with inferior memory chips, and in the desktop world Nvidia cards tend to be able to clock memory higher than AMD cards.

Link to comment
Share on other sites

Except for the 6x0m series, all GDDR5 mobile GPUs run the memory at the same 1.5V as desktop cards. The reason why they clock lower is because there are major crosstalk issues on the Nvidia 256-bit mxm board. AMD cards have a thicker pcb to combat this. While 6ghz is a struggle on any Nvidia card, I have a 7970m which does 6.21ghz with no mods 100% stable. I overvolted another and was running 6.4ghz. Keep in mind these are 3 year old cards with inferior memory chips, and in the desktop world Nvidia cards tend to be able to clock memory higher than AMD cards.

780Ms pretty much did 6GHz without issue, and the hynix vRAM MSI cards with Titan memory were easily able to cross that too. 880Ms as well did higher vRAM clocks as far as I know. But the 980Ms I've *NEVER* seen a stable 6GHz card.

If it's simply crosstalk issues, why is that so? What makes the crosstalk so much worse on the 980Ms than the 700M/800M series?

Link to comment
Share on other sites

One thing to NVIDIA's defense:

At least they went full board for their last Maxwell and we are going to get a full blown GM204 to 'play' with.

Earlier rumblings suggested the plan for only 1664 Shading Units (aka mobile GTX 970) for the refresh as they had ditched that early on for the original GTX980M.

Link to comment
Share on other sites

Well yes, they did, though haven't we had full board cards for quite a few gens?

My 280M was full blown G92b

485M/580M/675M were full blown GF104/GF114

780M/<s>broken garbage</s>880M was full blown GK104

The 680M was a 670's core with terrible vRAM, but Kepler was re-done twice, with much better cards each time except the unmentionable.

I simply more "expect" that they do something like that. But with the terrible VRM state and such of the 980Ms... at least they're doing SOMETHING halfway decent; they had no real reason to; no competition whatsoever.

*sigh*

Question though: Do you think the P370EM's slave GPU heatsink (lapped) would be capable of handling a Mobile 980's heat? I know the 980Ms would be no problem, but I'm unsure about that last one. I want the better CPU cooling, but if the GPU won't hold very well... I'll have to make a hell of a choice.

Link to comment
Share on other sites

Well yes, they did, though haven't we had full board cards for quite a few gens?

My 280M was full blown G92b

485M/580M/675M were full blown GF104/GF114

780M/<s>broken garbage</s>880M was full blown GK104

The 680M was a 670's core with terrible vRAM, but Kepler was re-done twice, with much better cards each time except the unmentionable.

I simply more "expect" that they do something like that. But with the terrible VRM state and such of the 980Ms... at least they're doing SOMETHING halfway decent; they had no real reason to; no competition whatsoever.

*sigh*

Question though: Do you think the P370EM's slave GPU heatsink (lapped) would be capable of handling a Mobile 980's heat? I know the 980Ms would be no problem, but I'm unsure about that last one. I want the better CPU cooling, but if the GPU won't hold very well... I'll have to make a hell of a choice.

The 485m was an interesting card. The desktop world never saw a fully enabled GF104.

680m (clevo version) and 780m had the same exact 5ghz rated Samsung chips. However this rating only applied for 1.5V, and the 680m ran at 1.35V. I think the memIO voltage was lower as well.

The slave GPU heatsink if lapped should be OK with the 980m if I believe, but you probably can't overclock it crazy hard. Have you already lapped the heatsink and tried it on your 780m?

Link to comment
Share on other sites

The 485m was an interesting card. The desktop world never saw a fully enabled GF104.

680m (clevo version) and 780m had the same exact 5ghz rated Samsung chips. However this rating only applied for 1.5V, and the 680m ran at 1.35V. I think the memIO voltage was lower as well.

The slave GPU heatsink if lapped should be OK with the 980m if I believe, but you probably can't overclock it crazy hard. Have you already lapped the heatsink and tried it on your 780m?

Yeah, GF114 was the one the world saw for desktops. I remember the day the 470M came out and I sat there doing math on the cards and determined it was better than the 480M and people were telling me I was crazy on NBR =D. When they found it was true the world flipped out.

This is most likely true.

I was considering mobile GTX 980, not the 980M. I know it should work fine for 980Ms even with some slight OCs. I don't have the heatsinks yet; I had a job I'm aiming for in a new company early next year when I was planning to start purchasing parts. But the Mobile GTX 980s that appear in MXM formats... THAT I don't know.

Link to comment
Share on other sites

Looking at the image Nvidia released I'm a bit concerned:

post-5693-14495000748884_thumb.jpg

If that's the only form factor for the 980 mobile, then that won't work for any system. For Clevo systems the middle post between the GPU and CPU would need to be removed. In addition the core is shifted slightly leftward, so heatsinks will need to be heavily modified. The left post for screwing the card and memory heatsink is moved leftward as well, and oh yeah, making contact with those left side memory chips will be fun. Hopefully this is an ASUS card as they always make funky shaped cards. I'm hoping for a standard size card with a thicker pcb like AMD cards to get memory clocks up, but I'll likely be disappointed.

Also the card in that image has 4 cores phases.

Link to comment
Share on other sites

Looking at the image Nvidia released I'm a bit concerned:

[ATTACH=CONFIG]15987[/ATTACH]

If that's the only form factor for the 980 mobile, then that won't work for any system. For Clevo systems the middle post between the GPU and CPU would need to be removed. In addition the core is shifted slightly leftward, so heatsinks will need to be heavily modified. The left post for screwing the card and memory heatsink is moved leftward as well, and oh yeah, making contact with those left side memory chips will be fun. Hopefully this is an ASUS card as they always make funky shaped cards. I'm hoping for a standard size card with a thicker pcb like AMD cards to get memory clocks up, but I'll likely be disappointed.

Also the card in that image has 4 cores phases.

ASUS card has always been such way.

I hope othre vendors aren't that way.

Link to comment
Share on other sites

GTX 980 in notebook?!

I can't imaginable that, LOL.

Nvidia introduced it in their site, but I think it will be too expensive to buy :(

However, it will be "the best" gaming laptop :)

The P570WM will have a word with any laptop that has a single 980 in it. A word like this: NVIDIA GeForce GTX 980M video card benchmark result - Intel Core i7-4960X,Clevo P570WM powered by PremaMod.com

  • Thumbs Up 1
Link to comment
Share on other sites

It looks like SLI system.

I'm also using SLI, but it disappointed me... a little bit.

A benchmark program can use SLI or CF 99%, but in real, I live in Korea, many online games not support SLI or CF.

So, if I have a chance to get 980M SLI or 980, then I will choice 980 notebook :)

Hmm... Am I miss the point what you want to say? If that is true... Sorry for my bad English ;)

Link to comment
Share on other sites

Hi there!

What are the expectations of this card turn out to be compatible with clevo P150EM? Will only be a matter of a bios update or hardware modifications will be necessary? Or both ? Or is it impossible to accomplish? Is there any price estimated for it?

Link to comment
Share on other sites

It looks like SLI system.

I live in Korea, many online games not support SLI or CF. So, if I have a chance to get 980M SLI or 980, then I will choice 980 notebook :)

It is a SLI system.

I understand your pain. A lot of the games I've been playing recently have not been very nice with SLI. In terms of single GPU notebooks, a P870DM with the mobile 980 is the way to go, definitely.

What are the expectations of this card turn out to be compatible with clevo P150EM? Will only be a matter of a bios update or hardware modifications will be necessary? Or both ? Or is it impossible to accomplish? Is there any price estimated for it?

Likely, the mobile 980 as you're seeing advertised right now won't fit. There is almost certainly going to be a MXM version, but you may want to consider getting a 240W PSU for your P150EM if it's compatible. You'll need I believe an adapter mod LIKE THIS, as well as a Dell PA-9E 240W PSU.

  • Thumbs Up 1
Link to comment
Share on other sites

Except for the 6x0m series, all GDDR5 mobile GPUs run the memory at the same 1.5V as desktop cards. The reason why they clock lower is because there are major crosstalk issues on the Nvidia 256-bit mxm board. AMD cards have a thicker pcb to combat this. While 6ghz is a struggle on any Nvidia card, I have a 7970m which does 6.21ghz with no mods 100% stable. I overvolted another and was running 6.4ghz. Keep in mind these are 3 year old cards with inferior memory chips, and in the desktop world Nvidia cards tend to be able to clock memory higher than AMD cards.

Just my two cents - every single 7970m I've tested (about 5) has had absolutely no problem getting to at least 6ghz on the mem.

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...

Important Information

By using this site, you agree to our Terms of Use. We have placed cookies on your device to help make this website better. You can adjust your cookie settings, otherwise we'll assume you're okay to continue.