NVIDIA to Reuse Pascal for Mobility-geared MX300 Series

[Raevenlord]

NVIDIA will apparently still be using Pascal when they launch their next generation of low-power discrete graphics solutions for mobile systems. The MX300 series will replace the current crop of MX200 series (segregated in three products in the form of the MX230, MX250 10 W and MX250 25 W). The new MX300 keeps the dual-tiered system, but ups the ante on the top of the line MX350. Even though it's still Pascal, on a 14 nm process, the MX350 should see an increase in CUDA cores to 640 (by using NVIDIA's Pascal GP107 chip) from the MX250's 384. Performance, then, should be comparable to the NVIDIA GTX 1050.

The MX330, on the other hand, will keep specifications of the MX250, which signals a tier increase from the 256 execution units in the MX230 to 384. This should translate to appreciable performance increases for the new MX300 series, despite staying on NVIDIA's Pascal architecture. The new lineup is expected to be announced on February.



View at TechPowerUp Main Site

 

[TheoneandonlyMrK]

despite keeping on AMD's Pascal architecture.


Pardon, correction in order bro.

 

[Assimilator]

Didn't know Pascal was ever on 14nm.

 

[cucker tarlson]

Didn't know Pascal was ever on 14nm.

1050 was always 14nm glofo

 

[Assimilator]

1050 was always 14nm glofo

Right you are, I thought the article was talking about Maxwell for some reason. But wasn't NVIDIA's 14nm at Samsung?

 

[Anymal]

1050 was always 14nm glofo

Hell no, no GloFo. Sammy was, as for Ampere will be.

 

[QUANTUMPHYSICS]

The 1060 and up can run Crysis in maximum settings on a laptop.

I'd really like to see more laptops pushed to a minimum of 1080 power considering a RTX level 2060 laptop is around $1000.

 

[lexluthermiester]

This development could be a very good thing for the mobile PC market. Solidly performing GPU parts for reasonable prices.

 

[Tsukiyomi91]

Pascal to power the MX300 Series? This is interesting...

 

[IceShroom]

So now people has no problem with 4 year old architecture.

 

[Totally]

So now people has no problem with 4 year old architecture.

If they're going to charge "hey, this stuff is 4 year old" prices I couldn't be bothered to care.

 

[Ruru]

So now people has no problem with 4 year old architecture.

Fits perfectly with Intel CPUs.

 

[Ferrum Master]

Well, at least there will be a sure long term driver support for Pascal cards.

 

[IceShroom]

Fits perfectly with Intel CPUs.

Yep. For 1000$-1500$ Comet Lake based Laptop.

 

[Xajel]

While I hate rebranding, it's understandable from both AMD and NV, especially with these lower end parts as it's hard to keep the cost low while spending R&D for each generation.

But, being also for mobile, I wonder why they didn't invest in 12nm, is the cost doesn't worth the extra power/efficiency.

 

[Cheeseball]

It's similar to how AMD is still keeping Vega 3 to 11 in their APUs. It's still GCN (without HBM). If it works, don't fix it, especially if it's a low-power part.

940MX (Maxwell) = MX 150 (Pascal) = GT 1030 = MX 250 (Pascal) = MX 350 (Pascal @ 640 SPs)

It's great for 14" laptops aimed at light-gaming and media consumption (light CUDA & NVDEC).

 

[Tartaros]

It's great for 14" laptops aimed at light-gaming and media consumption (light CUDA & NVDEC).

Yeah, I'm quite happy with my hp x360 with mx250, it does its job. I suppose they will move on when the gt1030 is succeeded, but yeah, they could have lowered the process if is just for some performance gain / lower consumption.

 

[prtskg]

AMD would be happy with this. Navi will be able to compete well in laptops too.

But, being also for mobile, I wonder why they didn't invest in 12nm, is the cost doesn't worth the extra power/efficiency.

One reason could be cost. Another could be getting more experience with Samsung for their gpus. After all, their next gen gpus are using Samsung 7nm.

 

[lexluthermiester]

So now people has no problem with 4 year old architecture.

You're suggesting there was a problem to begin with..

 

[IceShroom]

You're suggesting there was a problem to begin with..

Pascal in 2020 in $1000+ laptop which you cannot swap out. No problem!!

 

[lexluthermiester]

Pascal in 2020 in $1000+ laptop which you cannot swap out. No problem!!

And? Can't do that with CPU's on laptops anymore either. Your point was?

 

[newtekie1]

If people want to talk about pricing, I bought my ASUS laptop back in 2017 brand new for $750 with a GTX1050 in it which is supposed to be equivalent to the MX350.

 

[Tartaros]

Pascal in 2020 in $1000+ laptop which you cannot swap out. No problem!!

Mine was 800 and is a convertible. So not really.

Most of them are convertibles and ultrabooks, which are more expensive than regular laptop counterparts. Following that reasoning why would anyone buy an ultrabook when a regular laptop with similar hardware can cost less?

 

[IceShroom]

Mine was 800 and is a convertible. So not really.

Most of them are convertibles and ultrabooks, which are more expensive than regular laptop counterparts. Following that reasoning why would anyone buy an ultrabook when a regular laptop with similar hardware can cost less?

Who is saying that ultrabook should not a dGPU? What I am saying that new ultrabook releasing in 2020 should have gpu which have new architecture, not the architecture released in 2016.

And? Can't do that with CPU's on laptops anymore either. Your point was?

Well both Ice Lake, Comet Lake new iGPU that is not released in 4 years ago.

 

[newtekie1]

Who is saying that ultrabook should not a dGPU? What I am saying that new ultrabook releasing in 2020 should have gpu which have new architecture, not the architecture released in 2016.

At the end of the day, there isn't anything equivalent to the GP107. It might be from 2016, but it's only one generation old. It also uses about 20w less than the equivalent(but yes faster) TU117. Yeah, i'd be nice if nVidia released a lower end TU sku, but I'm guessing 12nm production is far enough behind, so it makes sense to stick with the 14nm GP107 SKU in lower end machines. It isn't like GP107 isn't quite capable at 1080p.

Well both Ice Lake, Comet Lake new iGPU that is not released in 4 years ago.

And yet, still insanely slower than GP107... No one cares when the product was released, performance is what matters.

If someone said I could get a 2020 Dodge Caravan or a 2016 Corvette for the same price, and my intent was to race them on the track, I'm not buying the Caravan just because it's newer.