Nvidia Used Maxwell GPUs Masquerading As Pascal In Drive PX2 Demo

Several reports have surfaced alleging that Nvidia's Pascal GPUs could be in deep trouble after Pascal was notably absent from the Bulldoze PX2. Nvidia appear that this 250W automotive compute box is powered by two Tegra fries and two Pascal GPUs. Nonetheless the Drive PX2 lath Nvidia'south CEO Jen-Hsun Huang held on phase carried two Maxwell based GTX 980 MXM modules instead. As was evident from the date inscripted on the fries, the size of the fries and the configuration of the modules.[i][2][3]

Nvidia Pascal Bulldoze PX2 photo by Anandtech.com

Nvidia's Pascal Is MIA, Could Be In Trouble Reports Criminate

Many journalists constitute this quite the eye brow raising affair. Some of whom constitute this to exist reminiscent of the company's Fermi troubles in 2022. Fermi was Nvidia's showtime GPU compages to utilize 40nm technology, and due to various challenges the company faced during this important node transition a like incident took identify.

Extremetech.com
These issues with Pascal and the Bulldoze PX 2 repeat the Fermi "forest screw" even of 2009. Back so, Jen-Hsun held up a Fermi board that was aught but a mock-upwards, proclaimed the bit was in total production, and would launch before the end of the twelvemonth. In reality, NV was having major problems with GF100 and the GPU only launched in late March, 2022.

Use of mockups and prototypes in stage-craft is not something that's new to the industry or Nvidia, only it's utilise in place of the real thing has always been limited to situations where the real affair isn't set, in this case Pascal. This isn't the first time that nosotros've seen a mockup being used in a Pascal declaration either. Last yr Jen-Husn Huang held Pascal lath, which did not actually take a Pascal GPU. Just rather a placeholder chip made to correspond Pascal. This epitome was made to showcase what a mezzanine form-factor board would look like, and so a functioning Pascal GPU wasn't pertinent to this goal.

NVIDIA Pascal GPU Prototype
The same argument can be made for the Bulldoze PX2 sit-in. And perhaps this is merely a marketing decision on Nvidia'south role to withhold showing any actual Pascal silicon until GTC. But the question remains why weren't actual Pascal GPUs used in whatsoever demo to date. More than importantly why did Nvidia try to pass sometime Maxwell silicon every bit Pascal?
This may wait particularly troubling for Nvidia when on the other side of the money we take AMD demoing functioning Polaris chips to the public and to the press, Charlie Demerjian notes. Does Nvidia take working Pascal silicon? Is the roadmap still on track? we can simply know in due time. One thing's for certain, if Pascal silicon doesn't bear witness up at GTC this April people may have genuine reason to worry.

Pascal Still Set To For A 2022 Launch - Featuring HBM2, 16nm, Mixed Precision And NV-Link

There are four hallmark technologies for the Pascal generation of GPUs. Namely HBM, mixed precision compute, NV-Link and the smaller, more than power efficient TSMC 16nm FinFET manufacturing procedure. Each is very of import in its own correct and as such we're going to break down everyone of these 4 separately.

#i HBM : Stacked memory will debut on the green side with Pascal. HBM Gen2 more precisely, the second generation of the SK Hynix AMD co-developed high bandwidth JEDEC retentiveness standard. The new memory will enable memory bandwidth to exceed 1 Terabyte/s which is 3X the bandwidth of the Titan X. The new memory standard will also allow for a huge increase in memory capacities, 2.7X the memory capacity of Maxwell to be precise. Which indicates that the new Pascal flagship will feature 32GB of video memory, a mind-bogglingly huge number.

We've already seen AMD take advantage ofHBM memory technology with its Fiji XT GPU. Which volition characteristic 512GB/S of memory bandwidth, which is twice that of the GTX 980. AMD has likewise stated that it plans to utilize the second generation of this new retention engineering in its Arctic Islands family of GPUs in 2022. So we're likely to meet both red and green rocking second generation stacked HBM side by side twelvemonth.

HBM achieves this astonishing improvement in retentiveness bandwidth and capacity by employing a very broad through-silicon-via memory interface. Each HBM cube is connected to the GPU with a 1024bit wide retention bus. HBM modules really operate at low frequencies compared to GDDR5 but thank you to the significantly wider memory interface they manage to be up to 9 times faster than standard GDDR5 memory modules.

Nosotros've already covered this revolutionary new memory applied science exclusively and in-depth last twelvemonth. HBM will rapidly replace GDDR5 as the standard memory technology for high performance graphics solutions. It's fair to say that HBM is the hereafter.

#ii Mixed Precision / Half Precision / 16FP Compute

One of the more pregnant features that was revealed for Pascal was the addition of 16FP compute support, otherwise known equally mixed precision compute or half precision compute. At this mode the accuracy of the consequence to any computational trouble is significantly lower than the standard 32FP method, which is required for all major graphics programming interfaces in games and has been for more a decade. This includes DirectX 12, 11, x and DX9 Shader model 3.0 which debuted near a decade ago. This makes mixed precision manner unusuable for whatsoever modern gaming application.

However due to its very bonny ability efficiency advantages over FP32 and FP64 it tin can exist used in scenarios where a high degree of computational precision isn't necessary. Which makes mixed precision computing peculiarly useful on ability limited mobile devices. Nvidia's Maxwell GPU architecture characteristic in the GTX 900 series of GPUs is limited to FD32 operations, this in turn ways that FP16 and FP32 operations are candy at the same rate by the GPU. Still, adding the mixed precision capability in Pascal means that the architecture will now be able to process FP16 operations twice as speedily as FP32 operations. And every bit mentioned above this tin can be of great do good in power express, light compute scenarios.

#3 NV-Link : Pascal will as well be the showtime Nvidia GPU to feature the visitor's new NV-Link technology which Nvidia states is five to 12 times faster than PCIE iii.0.

NVLINK_3

NVLink is an energy-efficient, high-bandwidth communications channel that uses up to three times less free energy to motion data on the node at speeds 5-12 times conventional PCIe Gen3 x16. Start available in the NVIDIA Pascal GPU architecture, NVLink enables fast advice between the CPU and the GPU, or betwixt multiple GPUs. Effigy iii: NVLink is a key building cake in the compute node of Summit and Sierra supercomputers.

VOLTA GPU Featuring NVLINK and Stacked Retentiveness NVLINK GPU high speed interconnect 80-200 GB/s 3D Stacked Retentiveness 4x Higher Bandwidth (~1 TB/s) 3x Larger Chapters 4x More than Energy Efficient per bit.

NVLink is a key technology in Summit's and Sierra's server node compages, enabling IBM POWER CPUs and NVIDIA GPUs to admission each other's memory fast and seamlessly. From a programmer's perspective, NVLink erases the visible distinctions of data separately fastened to the CPU and the GPU by "merging" the retentivity systems of the CPU and the GPU with a high-speed interconnect. Because both CPU and GPU accept their own retentivity controllers, the underlying retentivity systems can be optimized differently (the GPU's for bandwidth, the CPU's for latency) while still presenting as a unified memory organisation to both processors. NVLink offers ii distinct benefits for HPC customers. First, it delivers improved application performance, simply by virtue of profoundly increased bandwidth between elements of the node. Second, NVLink with Unified Memory technology allows developers to write code much more than seamlessly and still achieve high operation. via NVIDIA News

NVLINK_4
#iv 16nm manufacturing procedure : Pascal volition the outset Nvidia GPU to be built on TSMC's 16nm FinFET manufacturing procedure. The new process promises to be significantly more power efficient and significantly more dumbo than 28nm. Which would enable Nvidia to build significantly more complex and powerful GPUs all the while significantly improving ability efficiency.

TSMC's 16FF+ (FinFET Plus) technology tin can provide above 65 percent college speed, around 2 times the density, or seventy percentage less power than its 28HPM technology. Comparing with 20SoC technology, 16FF+ provides extra twoscore% higher speed and 60% power saving. Past leveraging the experience of 20SoC technology, TSMC 16FF+ shares the same metal backend process in social club to rapidly improve yield and demonstrate process maturity for time-to-market place value.