Overall I agree, just couple points I want to make.
GP100 Tesla with HBM2 wont be coming until at least august due to HBM2 mass production until 2H 2016 and I imagine quite a bit of QA is needed for Tesla compare to Geforce.
Whats likely is going to happen is a GP104 GDDR5X based Tesla in April/May. Followed by a crippled GP100 Tesla with HBM2 in August/September. A launch of titan pascal is likely next year March to May, it follows Nvidia big die launching time frame and it makes sense.
GP106 is probably going to focus on power consumption. I imagine with the heat density on 16nm finfet with TSMC's yield, the first gen is more likely to be 10% below 980m than above. Just a more conservative estimate on that.
Agreed on GP104 card at Q3. It makes sense to launch a Geforce Product sometimes shortly after polaris' line launch to gauge whats need to defeat AMD. I would imagine by Q3, there is enough binned Tesla GP104 nvidia can afford to launch a Geforce. I am not sure about the CLC requirement on stock. Thats not the style of nvidia, I can see a crippled GP104 flagship though if the yield is really that bad.
I somehow doubt GP104 will be running HBM2 either, nvidia needs them for their cash cow.
-
Yeah the only part I disagree is the CLC, since nVidia has an image to preserve, and strapping on a CLC basically sends the message "our card RUNS ASS HOT and can't be cooled using normal means". Although that said, those who bashed the CLC on Fury X seemed to be pretty happy when it was rumored that Pascal would use a CLC, so nVidia could probably pull it off because there are just too many braindead rabid fanbois.
-
-
PrimeTimeAction Notebook Evangelist
If this is the case then for me the main question is how likely is the P870DM (its custom MXM or even standard one for 980m) be compatible with GPUs 3 / 4 years down the line (pascal or volta). If it is likely to be compatible then i think the best option to get one now and wait for pascal to mature and then upgrade. If not then i dont know what to do.
To be honest iam leaning more and more towards a desktop now. Even if pascal/nnvlink needs new motherboard, it will still be better then seeing your entire laptop becoming obsolete in 1 year. -
nvlink is not going to take off ever in consumer....PrimeTimeAction likes this. -
PrimeTimeAction Notebook Evangelist
But if its MXM is likely to be supported in future then I think its the best strategy.hmscott likes this. -
hmscott likes this.
-
PrimeTimeAction Notebook Evangelist
And if I do go the desktop route, it will not be small. It will be mid/full tower (probably Lian Li PC 08) with 980TI (probably kingpin) and maybe custom water cooled CPU and GPU. I mean if I have to get tied up to one place, why not go all out.Last edited: Feb 23, 2016hmscott likes this. -
That's my major complaint about eGPU. You have an expensive box, high end GPU, and probably an expensive external monitor hooked up to the box, an external keyboard and mouse, and none of it is useful without the laptop plugged in.
You could have a fully functional 2nd computer hooked up with all those goodies, and have a 2nd screen laptop that runs independent of the desktop computer.
After spending $300+ for the eGPU box, $600+ for the 980ti - or $1000+ for a new Pascal top end GPU, and $500+ for the monitor, you have invested more than enough for a computer with those parts.
Why force your laptop into service as a desktop, when you can have a desktop and a laptop as independent computers?
File sharing? Seems like overkill to me. You won't be installing games on the laptop if it's gonna be a 6500u low power cpu, you can save the space there and install games in the desktop.
If you are want a skinny laptop to carry around, then take the savings between what it costs and a full gaming laptop, and invest that in a CPU/memory/disk for your desktop parts.
Nothing has changed really, it's just that the eGPU math helps define what we already knowLast edited: Feb 23, 2016jaybee83, Robbo99999 and PrimeTimeAction like this. -
You're all making wild assumptions...
I'd be shocked if GP106 wasn't faster than 980M.......when overclocked.....tgipier likes this. -
-
PrimeTimeAction Notebook Evangelist
-
As for GP106, maybe overclocked. Probably not stock?
I will laugh nvidia is gonna brand GP106 as x080m. If thats the case, yeah probably. Which would be very sad..... -
FinFets hit high clocks... Very high clocks...
PrimeTimeAction likes this. -
PrimeTimeAction Notebook Evangelist
-
Don't new architectures for nvidia usually give about a 20 to 30 percent performance increase? Historically speaking, I have no idea about Pascal though.
-
A performance of 10% ahead of 980m is enough to rebrand it as x080m.
And launch the GP104 as a "desktop" card later with an huge premium.
I really think the performance depend on what nvidia's branding decisions for GP106 will be. If it is 1060m, it might decide to leave some room for GP104. If it is 1080m, then well...
Either way, I guess it depends on how well GP106 performs and if nvidia wants to make room for GP104.
Do you know anything about GP102 or just some rumor made by rumormills?Ethrem likes this. -
oh he knows
he just cant tell explicitly
only via rumor-style talk
Mr Najsman, Ethrem, tgipier and 1 other person like this. -
-
Either that or GP102 is a version of GP100 used for gaming and GP100 itself is a pure computing die for pro market only. Or they are flipped.
Edit: I think the second one is much more possible.Last edited: Feb 23, 2016 -
More interesting news comrades! Seems like AMDs release of Polaris in April-May may have forced Nvidia to release their Pascal on mobile in June before desktop!
http://videocardz.com/58338/nvidia-pascal-to-be-unveiled-at-computex-2016Mr Najsman likes this. -
HA!
NAILED IT on the TSMC might fail us once again bit.
-
"However if any difficulties are met on new TSMC 16nm FinFET process, this launch could transform to ‘paper launch’." -
yeah that part
-
Ok, but that parts a given right ? If the manufacturer fails to deliever, it can be delayed. Thats what that part basically said right ?
-
Well if I'm not mistaken, AMD's high end Polaris parts would also use TSMC, and only the low-mid range GPUs actually use Samsung's 14nm process. So a TSMC fail would impact both red and green unfortunately.
-
This isn't good if AMD switched... As it stands right now, AMD would be in a better position in the short term........ At least Samsung and GloFo have tested processes unless this means that Samsung killed the US Exynos because of yield which bodes terribly for us all... Including those of us upgrading phones this year, even the iPhone 7. If there is an issue with 14nm LPP, its not good for anyone.
Edit: wow... So for those of you comparing the scores of the Snapdragon 820 vs the new Exynos 8890... The SD820 is a quad core...
Looks like TSMC is on top this time... I only heard of one delay and it doesn't affect consumers but the Exynos 8890 and to a lesser extent, the 820, have issues... This doesn't look good for AMD 14nm LPP when Samsung had to basically concede defeat to the quad core Snapdragon using their same process.
Things might be shaking up... And not in AMD's favor. Not to say that nVidia isn't having issues either.Last edited: Feb 25, 2016 -
jaybee83 likes this. -
Cant wait for more. You think Nvidia would feel concerned if there is an delay ? Or perhaps they are still sitting on their high horse and just brushing it off.
-
nVidia could release another 480 and it would still sell out within minutes. So no nVidia isn't concerned at all.
I'd say at this point AMD would have to deliver more performance at literally half the price before people would even consider them. Sad state of affairs but that's just reality.Ethrem, i_pk_pjers_i and Omnomberry like this. -
-
i_pk_pjers_i Even the ppl who never frown eventually break down
-
It would be worth comparable price/performance, but it would be nice if they would come in a little cheaper, as the high end is a little over priced as exists now.
With 2 makes at the same performance range, dropping prices should start happening.tgipier and i_pk_pjers_i like this. -
Would it be worth it to upgrade from GTX 980Ti to GP100(Titan if reasonably priced or else 1080Ti)?
For gaming on 1440p all settings max(including AA) -
I need more money for all this technology coming out this year! I've just bought a Galaxy S7 and now I've got a save up for a new laptop haha oh my poor bank balance
Last edited: Feb 26, 2016 -
-
-
meh, mate 8 king of the hill biatches
but nice to see we all agree
Sent from my Huawei Mate 8 NXT-AL10 using Tapatalk -
i_pk_pjers_i Even the ppl who never frown eventually break down
I have a 980Ti already so I probably won't 1080Ti - I doubt it will be that big of an upgrade.Splintah likes this. -
The fact that I already have three really good SSDs in my P651SG means I will be able to spend less this time around. Hopefully I can get away with only £1200ish maximum. I would need to get the top Pascal GPU though, that's just who I am -
"NVIDIA Pascal to be unveiled at GTC in April" - Source
Hope so. Makes sense to me. We'll see soon enough!
I can't believe it's almost March already... What the hell happened with time?Last edited: Feb 26, 2016 -
keynotes together with a guy from toyota, yay....so its gonna be all about automated cars and their computing power, boooooooring
Mr Najsman and Phase like this. -
Phase likes this.
-
-
too many "ifs" involved to make any such statement with any degree of certainty...lets wait and see
Sent from my Huawei Mate 8 NXT-AL10 using Tapatalk -
If they don't have a functional demo by GTC, then they probably won't release / launch Pascal desktop until late June or early Q3. However, if they do present the next "Pascal" Titan, then they have likely entered mass production and will ship/sell them within days after the official announcement.
Let's just hope they have a functional demo and no more "mock-ups" using Maxwell.Last edited: Feb 29, 2016Robbo99999 likes this. -
TBoneSan, jaybee83, triturbo and 1 other person like this.
Pascal: What do we know? Discussion, Latest News & Updates: 1000M Series GPU's
Discussion in 'Gaming (Software and Graphics Cards)' started by J.Dre, Oct 11, 2014.