Nvidia Pascal GPUs spotted en route to testing facilities (Fudzilla)

MentalIlness

Platinum Member
Nov 22, 2009
2,383
11
76
Nvidia Pascal GPUs spotted en route to testing facilities



http://fudzilla.com/news/40167-nvidia-pascal-gpus-spotted-en-route-to-testing-facilities


f6c3925709349198f820042714a9a4ce_L.jpg


Shipping values priced between $608 and $810 USD per unit Any time a new batch of GPUs are ready to be sent from a manufacturing facility overseas to a supply partner for testing and metrics, there are always debugging and testing parts sent ahead of time to ensure that silicon is quickly brought to market without suffering silicon respin and unforeseen manufacturing delays.

Over the past two months, several sources have begun seeing Pascal GPU testing parts sent from China and Taiwan to India for testing. More recently, however, some actual Pascal-based GPU units have been shipping over the course of February and are on track to make a splash at Nvidia’s 2016 GPU Technology Conference next month.

Zauba is an India-based import and export database that allows corporate clients to track and access daily import and export shipment records to and from different countries and regions, discover new markets and gain new insights into international trade and new market opportunities.

As we wrote in February, at least four Pascal-based GPUs have been shipped from TSMC manufacturing facilities in China to Nvidia’s testing and validation facilities in India. All four models contain the same serial number prefix beginning with “699.” The four models are the 699-2H403-0201-500, 699-1H400-0000-100, 699-1G411-0000-000, and the 699-12914-0071-100.

According to the Zauba database, the most recent shipment for the 699-2H403-0201-500 was February 5, 2016, originating from China with a value of $608 USD per unit.

The most recent shipment for the 699-1H400-0000-100 was December 16, 2015, originating from China with a value of $668 USD per unit.

The most recent shipment for the 699-1G411-0000-000 was February 26, 2016, originating from China with a fluctuating value between $606 and $751 USD per unit.

The most recent shipment for the 699-12914-0071-100 was February 4, 2016, originating from China with a value of $810 USD per unit.


vr-zone-pascal-10x-maxwell.jpg


TweakTown senior GPU editor Anthony Garreffa notes that the higher-priced 699-1G411-0000-000 and 699-12914-0071-100 units could very well be the successors to the Maxwell-based Geforce GTX 980 Ti and the Geforce GTX Titan X, both based on the flagship 28-nanometer GM200 GPU. These chips are expected to come with an impressive 16GB and 32GB of High Bandwidth Memory 2 (HBM2) VRAM achieved using 4-Hi and 8-Hi stacks.

nvidia-nvlink-vs-pci-e.jpg


The flagship GP100 (“Geforce Pascal 100”) GPU will be the successor to the GM200 (“Geforce Maxwell 200”) GPU and features up to twice the performance-per-watt of its predecessor. Nvidia’s GP100 will contain a total of 17 billion transistors, over twice the number found in the GM200.

It will also feature Nvidia NVLINK, a new proprietary ultra-high speed interconnect technology exclusively for Nvidia Geforce GPUs that will enable communication between CPU and GPU, and between GPUs at data rates between 80GB/s and 200GB/s.

The company is hosting its annual GPU Technology Conference from April 4th to 7th in San Jose, California, and we expect this will be a prime opportunity for CEO Jen-Hsun Huang and senior Nvidia graphics architects to show the international GPGPU developer community and the rest of the world the incredible quantum leap in performance that Pascal will offer compared to 28-nanometer Maxwell chips.

We will be at GTC 2016 this year and look forward to sharing more concrete details on Pascal part numbers, retail pricing, and last but not least, the SKU naming scheme that Pascal will adopt as coming after the Maxwell-based “Geforce GTX 900 Series.”
 

moonbogg

Lifer
Jan 8, 2011
10,731
3,440
136
This is getting exciting. Any GPU upgrades for me will likely be dictated by BF5's performance.
 

Aristotelian

Golden Member
Jan 30, 2010
1,246
11
76
Will upgrade straight away \o/ Aslong as it's the full GP100 that comes out.

Yeah I'm looking for two or more of the Titan X replacements - if Nvidia gets those to the market before its competitors do (the successor to high end parts) they'll be getting a few thousand from me ;)
 

Mahigan

Senior member
Aug 22, 2015
573
0
0
NVLink is not compatible with x86 CPUs due to the Intel point to point interconnect used on consumer motherboards (PCIe). You could still pair multiple GPUs, with NVLink, but not x86 CPUs as they would interface through a PCIe switch. NVLink is compatible with IBM and ARM processors with NVLink integrated onto the motherboard.

Since dual GPUs don't saturate the PCIe bus, then it remains to be seen how useful NVLink would be on the desktop right now.
 
Last edited:

Ancalagon44

Diamond Member
Feb 17, 2010
3,274
202
106
If true, it means that Nvidia's upper mid range and higher GPUs will be competing against AMD's low and mid range GPUs. From the die size estimates of Polaris 10 and 11, it seems they are aimed at low and mid range.

It is only when Vega comes out later this year or Q1 2017 that we have a decent comparison.
 

exar333

Diamond Member
Feb 7, 2004
8,518
8
91
We'll probably get fed the $600.00 mid range parts first.

Fine by me. The 680/670 was a fantastic card and lasted for a while and was a great step-up from the 4xx/5xx series.

My philosophy has been buy the 2nd best card on the x104 release and then get the best on the x110, depending on your buying timing. That has worked well for me and resale is solid too.
 

exar333

Diamond Member
Feb 7, 2004
8,518
8
91
Christ, when did $600 cards become "mid range"

It's not, just some folks get butthurt when they don't get the GPU they feel entitled too....$600 is high end. :)

By that logic, I guess SB-E 6 cores were 'mid range' because the die was a Xeon 8C. Just crazy...