Nvidia Pascal Announcement

Is it the same for the A9X on the Ipad ? .... And will it be the same on the A10 ? because for what i know Samsung still have a good part of the deal for them ( shared with TSMC )
Tsmc will be only manufacturer for iPhone 7
 
Is it the same for the A9X on the Ipad ? .... And will it be the same on the A10 ? because for what i know Samsung still have a good part of the deal for them ( shared with TSMC ).

This said, we have seen all and nothing, here TH see a difference in performance who seems averaged around 0.5% difference, with a lower difference at 0.1% and one figure at 3.7% ( strange case ).

But battery life is better on Samsung ... http://www.tomshardware.com/news/iphone-6s-a9-samsung-vs-tsmc,30306.html

Honestly this one review, i have not been able to find the same numbers on different reviews, result seems all over the place.. because smartphone are nearly impossible to benchmark and to been tested correctly.

benchmarks suites for smartphone are a nightmare, let alone the batter life tests.

Even if you had multiple reviews..there is still variation in silicon which is not accounted for. You'll probably need a sample size of a few dozen of each to draw any sort of conclusion
Tsmc will be only manufacturer for iPhone 7

Yep..that's the rumour so far..and extending all the way to 7nm apparently.
 
We will see if Samsung 14nm is really worst than TSMC 16FF+ or if AMD is just too much behind Nvidia in Power efficiency optimization (uarch and/or die layout).
Yum! I love me some all other things equal comparison!

Edit: BTW I think Nvidia moving to Samsung is to be free of TSMC limited wafers capacity (thanks to Apple...)
And increased bargaining power with TSMC. I'm not convinced at all that Apple has first dibs on TSMC wafer capacity. It would be a very short-sighted way to run a business.
 
NVIDIA GeForce GTX 1080 Mobility GPU Pictured – First Notebook Chip With GDDR5X Memory, 20% Faster Than The Original Titan X

NVIDIA is going to unleash the worlds fastest notebook chip ever, the GeForce GTX 1080. Based on the Pascal architecture, the GeForce GTX 1080 (mobility) solution will be embedded inside high-performance notebooks and laptops from various manufacturers. The GeForce GTX 1080 (mobility) will also provided a big performance jump compared to the flagship Maxwell graphics card, the Titan X.

As for specifications, the chip is based on the GP104 die which houses 2560 CUDA cores, 160 TMUs and 64 ROPs. The chip is the first mobility solution to utilize GDDR5X memory or G5X in short. This memory is clocked at 10 GHz to deliver a bandwidth of 320 GB/s. The memory runs along a 256-bit bus interface.

http://wccftech.com/nvidia-gtx-1080-mobility-gddr5x-gpu/
 
Only for very custom Notebooks though. Non-MMX edit: MXM-board required.

And, according to one of the pictures there:
NVIDIA-GeForce-GTX-1080-Mobility-Features.jpg

(credit: wtftech or whoever they took it from)

„Geforce GTX 1080 is the king of GPUs,…“

Maybe someone should tell the mobility division that:
a) a graphics card aka SKU is not a GPU and
b) GP102 is available.
 
Last edited:
No GloFo is worse Samsung does it right. GloFo tried to copy Samsung's process and screwed up doing it.

In reality, they dont "copy" it, as it is implemented in collaboration with Samsung ( Glofo and Samsung have a colllaboration deal for FF ), if something went wrong in Glofo, knowing that engineer of Samsung are deeply involved in it, it should be fixable.
 
In reality, they dont "copy" it, as it is implemented in collaboration with Samsung ( Glofo and Samsung have a colllaboration deal for FF ), if something went wrong in Glofo, knowing that engineer of Samsung are deeply involved in it, it should be fixable.
Is that the exact wording of the contract? This should be gospel and taken for granted at all costs. Always. So say we all. ;)
 
Looks like they copied the playbook of the 480 4GB which has slower memory than the 8GB. In a way, both AMD and Nvidia will strengthen the already common perception that more memory equals better performance. ;)

Actually AMD's specs for both the 4 GB and 8 GB cards just call for a minimum of 7 Gbps. However from what I can gather, the reference 8 GB cards sent to reviewers by AMD did indeed have 8 Gbps memory. Slightly lower memory bandwidth is not as bad as a cut down GPU like NV is doing, but I don't like it either. That perception is why I was half expecting a 4GB 128 bit part instead.
Only for very custom Notebooks though. Non-MMX board required.

You mean I cant use it with my 166 Mhz Pentium??? :runaway:
 
Actually AMD's specs for both the 4 GB and 8 GB cards just call for a minimum of 7 Gbps.
Yes. But when AMD sends out 480 4GB configurations with 7Gbps, that's obviously how they want to position that particular GPU: one where less memory is also inherently slower.

However from what I can gather, the reference 8 GB cards sent to reviewers by AMD did indeed have 8 Gbps memory.
That's just a short term work-around to fix a logistics issue. In 4GB mode, they were still clocked at 7Gbps.

Slightly lower memory bandwidth is not as bad as a cut down GPU like NV is doing, but I don't like it either.
Yes.
 
Actually AMD's specs for both the 4 GB and 8 GB cards just call for a minimum of 7 Gbps. However from what I can gather, the reference 8 GB cards sent to reviewers by AMD did indeed have 8 Gbps memory. Slightly lower memory bandwidth is not as bad as a cut down GPU like NV is doing, but I don't like it either. That perception is why I was half expecting a 4GB 128 bit part instead.
Actually AMD specifies 4GB's at 7 Gbps and 8GB's at 8 Gbps, but notes that AIBs could use different clocks. It's in the RX 480 launch slides
 
Actually AMD specifies 4GB's at 7 Gbps and 8GB's at 8 Gbps, but notes that AIBs could use different clocks. It's in the RX 480 launch slides
Yes, understood. And they set the dial to 7Gbps for the reviewers. Meaning: this is how they prefer it to be reviewed.
 
Yes. But when AMD sends out 480 4GB configurations with 7Gbps, that's obviously how they want to position that particular GPU: one where less memory is also inherently slower.

That would be the logical conclusion..but 3 out of 4 RX480 4GB cards listed on Newegg have 8Gbps memory so who knows what AMD is really thinking ¯\_(ツ)_/¯ -
http://www.newegg.com/Product/ProductList.aspx?Submit=ENE&N=100007709 601203793 600007787&IsNodeId=1&Description=rx480 4 gb&name=Desktop Graphics Cards&Order=BESTMATCH&isdeptsrh=1
Actually AMD specifies 4GB's at 7 Gbps and 8GB's at 8 Gbps, but notes that AIBs could use different clocks. It's in the RX 480 launch slides

Like I said..AMD specifies the minimum speed at 7 Gbps. Partners are free to use 8 Gbps if they want to.

Anandtech Article up as well - http://www.anandtech.com/show/10564...series-for-notebooks-unveiled-launching-today

Interesting that Nvidia chose not to lower the TDPs with the benefits of 16nm and possibly even increased the TDPs. Unsurprisingly, there is a significant increase in performance compared to the equivalent Maxwell parts (Though with the GTX1060M, we should compare it to the GM206 based 965M and not the GM107 based 960M). That additional performance dosen't come for free though..NV says they expect prices to start at $1300. A bit expensive compared to current 965M laptops which start at about ~$950. Either ways..these GPUs will still be niche products..the real volume will be with GP107 and GP108.

I am a bit surprised that all the announced notebooks still seem to be using Skylake though. Maybe Kaby Lake still isnt ready to ship.
 
Last edited:
Interesting that the mobile 1070 has more units unlocked than the desktop one. Maybe Nvidia thought a 1070 with 2048 SPs would compete against the 1080 too favorably and would affect sales in the desktop market.
 
Interesting that the mobile 1070 has more units unlocked than the desktop one. Maybe Nvidia thought a 1070 with 2048 SPs would compete against the 1080 too favorably and would affect sales in the desktop market.

If we follow the old adage, wider and slower is more power efficient than narrower and faster. The 1070 was cut down a bit more than expected with only 15 out of 20 SMs enabled (75%) compared to the GTX 970 with 13 out of 16 SMs (82% but with only 88% the ROPs). Of course GM204 was released on a very mature 28nm process so it could simply be due to lower 16nm yields at this point. I expect we might see a 16 SM GTX 1070/1075 refresh at some point in the future.
 
Back
Top