Intel's announcement today that their 7 nm node is facing difficulties is being taken one of two ways: as an unmitigated disaster by some, and with a tentative carefulness (lest we see another 10 nm repeat) from others. However one looks at this setback, which means AMD will still enjoy a process lead over Intel for some extra time, this is good news for AMD in more ways than just that one.

Case in point: stock price. While AMD has a much lower market cap than Intel (calculated by multiplying the value of a single stock by the number of total issued stocks), today, for the first time since 2006, AMD's shares were more valuable than Intel's on a per-share basis. AMD's $70 billion market cap still pales in comparison to Intel's $215 billion. At time of writing, AMD's stock pricing is $18 higher than Intel, at $68.67 compared to Intel's $50.79. A first in many years for the green company.

Intel's silicon fabrication woes refuse to torment the company's product roadmaps, with the company disclosing in its Q2-2020 financial results release that the company's first CPUs built on the 7 nanometer silicon fabrication node are delayed by a year due to a further 6-month delay from prior expectations. The company will focus on getting its 10 nm node up to scale in the meantime.

The company mentioned that the 10 nm "Tiger Lake" mobile processor and "Ice Lake-SP" enterprise processor remains on-track for 2020. The company's 12th Generation Core "Alder Lake-S" desktop processors won't arrive before the second half of 2021. In the meantime, Intel will launch its 11th Gen Core "Rocket Lake" processor on the 14 nm node, but with increased IPC from the new "Cypress Cove" CPU cores. Also in 2H-2021, the company will launch its "Sapphire Rapids" enterprise processors that come with next-gen connectivity and updated CPU cores.

Intel Corporation today reported second-quarter 2020 financial results. "It was an excellent quarter, well above our expectations on the continued strong demand for computing performance to support cloud-delivered services, a work- and learn-at-home environment, and the build-out of 5G networks," said Bob Swan, Intel CEO. "In our increasingly digital world, Intel technology is essential to nearly every industry on this planet. We have an incredible opportunity to enrich lives and grow this company with a continued focus on innovation and execution."

Intel achieved record second-quarter revenue with 34 percent data-centric revenue growth and 7 percent PC-centric revenue growth YoY. These results were driven by strong sales of cloud, notebook, memory and 5G products in an environment where digital services and computing performance are essential to how we live, work and stay connected.

Taiwan-based tech publication CoolPC.com.tw published one of the first comprehensive performance reviews of the recently announced AMD Ryzen PRO 4750G, PRO 4650G, and PRO 4350G Socket AM4 desktop processors based on the 7 nm "Renoir" silicon that combines up to 8 "Zen 2" GPU cores with a Radeon Vega iGPU that has up to 8 compute units (512 stream processors). In their testing, the processors were paired with an AMD Wraith Prism (125 W TDP capable) cooler, an ASUS ROG Strix B550-I Gaming motherboard, 2x 8 GB ADATA Spectrix D50 DDR4-3600 memory, and a Seagate FireCuda NVMe SSD.

The benchmark results are a fascinating mix. The top-dog Ryzen 7 4750G was found to be trading blows with the Core i7-10700K, the i7-10700, and AMD's own Ryzen 7 3700X, depending on the benchmark. In CPUMark 99 and Cinebench R20 nT, the PRO 4750G beats the i7-10700 and 3700X while practically matching the i7-10700K. It beats the i7-10700K at 7-Zip (de-compression) and HWBOT x265 video encoding benchmark. The story repeats with the 6-core/12-thread PRO 4650G beating the Core i5-10600K in some tests, and AMD's own Ryzen 5 3600X in quite a few tests. Ditto with the quad-core PRO 4350G pasting the previous generation Ryzen 3 3300G.

Since the AMD Ryzen 4000 Series Desktop Processors with PRO technologies have launched today, MSI 500-series motherboards are well-prepared to fully support for the new processors' coming. Compared to the Ryzen 3000 series CPUs, the AMD Ryzen PRO 4000 Series Processors are built in a monolithic design based on the 7 nm architecture for both Zen 2 CPU and Vega GPU, which improves greatly in latency and bandwidth numbers with better efficiency in performance. Of course, MSI 500-series motherboards including X570 and B550 platform are perfectly compatible for the Ryzen PRO 4000 Series Processors.

AMD Ryzen PRO 4000 Series Processors offer greater CPU and memory performance for overclockers and enthusiasts to push benchmark to another level. MSI has showcased not only the best performance for memory frequency but also the memory stability with Memtest pass.

AMD today announced its 4th Generation Ryzen 4000G and Ryzen PRO 4000G desktop processors for pre-built OEM desktops. The company also expanded its entry-level Athlon 3000G series and debuted the Athlon PRO 3000G series. The Ryzen 4000G and PRO 4000G mark the Socket AM4 desktop debut of the 7 nm "Renoir" silicon, which combines up to 8 CPU cores based on the "Zen 2" microarchitecture, with a Radeon Vega 8 iGPU. These processors benefit from the 65 W TDP and increased power limits of the desktop platform to dial up CPU- and iGPU engine clock speeds significantly over the Ryzen 4000U and 4000H mobile processors based on the same silicon. The new Athlon 3000G-series and Athlon PRO 3000G-series parts are based on a 12 nm die that has "Zen+" CPU cores.

All of the processor models announced today are OEM-only, meaning that you'll only find them on pre-built consumer- and commercial desktops by the likes of HP, Lenovo, Dell, etc. Not even the system-integrator (SI) channel (eg: Maingear, Origin PC, etc.,) gets these chips. OEMs will pair these processors with motherboards based on the AMD B550 chipset, although the chips are compatible with the X570 chipset, too. The Ryzen PRO 4000G processors are targeted at commercial desktops that are part of large business environments, and launches along with the new AMD PRO565 chipset. Since they are OEM-only, the company did not reveal pricing for any of these chips. They did however mention that for the DIY retail channel, they do plan to update their product stack with processors that have integrated graphics at a later time (without going into specifics of the said time).

Semiconductor manufacturing is a difficult process. Often when a new node is being developed, there are new materials introduced that may cause some yield issues. Or perhaps with 7 nm and below nodes, they are quite difficult to manufacture due to their size, as the transistor can get damaged by the smallest impurity in silicon. So manufacturers have to be extra careful and must spend more time on the development of new nodes. According to industry sources over at DigiTimes, we have information that Samsung is struggling with its 5 nm EUV node.

This unfortunate news comes after the industry sources of DigiTimes reported that Qualcomm's next-generation 5G chipsets could be affected if Samsung doesn't improve its yields. While there are no specific pieces of information on what is the main cause of bad yields, there could be a plethora of reasons. From anything related to manufacturing equipment to silicon impurities. We don't know yet. We hope that Samsung can sort out these issues in time, so Qualcomm wouldn't need to reserve its orders at rival foundries and port the design to a new process.

AMD's top upcoming Socket AM4 desktop APU, the Ryzen 7 PRO 4750G, was put through Geekbench 5, as discovered by TUM_APISAK. The processor produced performance figures in the league of the popular Ryzen 7 3700X desktop processor. Both are 8-core/16-thread processors based on the "Zen 2" microarchitecture, but while the 3700X has additional L3 cache and added power budget for the CPU cores (as the processor completely lacks an iGPU); the PRO 4750G offers a Radeon Vega 8 iGPU with its engine clock above 2.00 GHz. Both chips were compared on Geekbench 5.2.2.

The single-core performance of both the PRO 4750G and 3700X are similar, with the PRO 4750G scoring 1239 points, and the 3700X scoring 1266 points. The 3700X has a slight upper hand with multi-core performance, with 9151 points compared to 8228 points of the PRO 4750G. This is attributable to the 3700X enjoying four times the L3 cache size. The Ryzen 7 PRO 4750G is expected to be the top desktop SKU based on the 7 nm "Renoir" silicon that features eight "Zen 2" CPU cores, and an iGPU based on the "Vega" graphics architecture, featuring 8 NGCUs amounting to 512 stream processors. The processor features AMD PRO feature-set that make it fit for use in commercial desktops in large business environments.

Apple has recently announced its transition from Intel-based Mac computers to custom Arm-based Apple silicon equipped Macs. The speculations for such transition have lasted a few years and we finally got that confirmation. So the question remains: who will manufacture Apple's custom processors for Arm-based Macs? The answer is pretty simple. It is TSMC who will again become Apple's main supplier of silicon. With its broad offerings of the latest silicon nodes, it was no brainer choice for Apple. Combined with the history of collaboration with Apple, TSMC was the only choice for new Apple silicon. Whatever the company will use the new 5 nm node or use the "old" 7 nm one, the question remains.

TSMC expects to see huge orders from Apple in the second half of 2021, for Apple silicon, so Apple will become perhaps the biggest customer of TSMC. It is also worth pointing out that Apple will be using ASMedia's USB controller for Arm-based Macs, as the original report suggests.

AMD's 3rd generation EPYC line of enterprise processors that leverage the "Zen 3" microarchitecture, could innovate in two directions - towards increasing performance by doing away with the CCX (compute complex) multi-core topology; and taking advantage of a newer/refined 7 nm-class node to increase clock-speeds. Igor's Lab decoded as many as three OPNs of the upcoming 3rd gen EPYC series, including a 64-core/128-thread part that ships with frequency of 3.00 GHz. The top 2nd gen EPYC 64-core part, the 7662, ships with 2.00 GHz base frequency and 3.30 GHz boost; and 225 W TDP. AMD is expected to unveil its "Zen 3" microarchitecture within 2020.

In a video message posted on her Twitter timeline, AMD CEO Dr Lisa Su confirmed that the company's next-generation "Zen 3" microarchitecture is coming out "later this year." Speaking in context of 7/7 (a year since AMD debuted high-performance CPU- and GPU- architectures on the same day, leveraging 7 nm), and the Ryzen 3000XT series processor announcement, Dr Su stated "As you know with Ryzen, we're always on a journey, a journey to push the highest performance that we can for our users and our fans. So Zen 3 is exactly that. Zen 3 is looking great in the labs, we're on track to launch later this year, and I can't wait to tell you more about it." Watch the video in the source link below.

Huawei has been readying the entire new breed of desktop PCs with a custom motherboard, custom processor, and even a custom operating system. Being that Huawei plans to supply Chinese government institutions with these PCs, it is logical to break away from US-made technology due to security reasons. And now, thanks to the YouTube channel called "二斤自制" we have the first look at the new PC system. Powered by Huawei D920S10 desktop motherboard equipped with Kunpeng 920 7 nm Arm v8 processor with 8 cores, the PC was running the 64-bit UOS operating system, which is a Chinese modification of Linux. In the test, the PC was assembled by a third-party provider and it featured 16 GB of 2666 MHz DDR4 memory and 256 GB SSD.

The YouTube channel put it to test and in the Blender BMW render test, it has finished in 11 minutes and 47 seconds, which is quite slow. The system reportedly managed to stream 4K content well but has struggled with local playback thanks to poor encoding. Being that it runs a custom OS with a custom processor, app selection is quite narrow. The app store for the PC is accessible only if you pay an extra 800 Yuan (~$115), while the mentioned system will set you back 7,500 Yuan (~$1,060). At the heart of this system is eight-core, eight threaded Kunpeng 920 2249K processor. It features a clock speed of 2.6 GHz, has 128K of L1 cache (64K instruction cache and 64K data cache), 512K of L2, and 32 MB of L3 cache.

AMD's 7 nm "Renoir" APU silicon, which features eight "Zen 2" CPU cores, has only a quarter of the L3 cache of the 8-core "Zen 2" CCD used in "Matisse," "Rome," and "Castle Peak" processors, with each of its two quad-core compute complexes (CCXs) featuring just 4 MB of it (compared to 16 MB per CCX on the 8-core "Zen 2" CCD). Chinese-language tech publication TecLab pubished a quick review of an alleged Ryzen 7 4700GE socket AM4 processor based on the "Renoir" silicon, and discovered that the chip offers significantly lower memory latencies than "Matisse," posting just 47.6 ns latency when paired with DDR4-4233 dual-channel memory.

In comparison, a Ryzen 9 3900X with these kinds of memory clocks typically posts 60-70 ns latencies, owing to the MCM design of "Matisse," where the CPU cores and memory controllers sit on separate dies, which is one of the key reasons AMD is believed to have doubled the L3 cache amount per CCX compared to previous-generation "Zeppelin" dies. TecLab tested the alleged 4700GE engineering sample on a ROG Crosshair VIII Impact X570 motherboard that has 1 DIMM per channel (the best possible memory topology).

VLSI engineer Fritzchens Fritz, famous for high-detail EM photography of silicon dies and annotations of them, recently published his work on AMD's 7 nm "Renoir" APU silicon. His die-shots were annotated by Nemez aka GPUsAreMagic. The floor-plan of the silicon shows that the CPU component finally dwarfs the iGPU component, thanks to double the CPU cores over the previous-gen "Picasso" silicon, spread over two CCXs (compute complexes). The CCX on "Renoir" is visibly smaller than the one on the "Zen 2" CCDs found in "Matisse" and "Rome" MCMs, as the L3 cache is smaller, at 4 MB compared to 16 MB. Being MCMs with disintegrated memory controllers, it makes more sense for CCDs to have more last-level cache per CCX.

We also see that the iGPU features no more than 8 "Vega" NGCUs, so there's no scope for "Renoir" based desktop APUs to feature >512 stream processors. AMD attempted to compensate for the NGCU deficit by dialing up engine clocks of the iGPU by over 40% compared to those on "Picasso." What caught our eye in the annotation is the PCI-Express physical layer. Apparently the die indeed has 20 PCI-Express lanes besides an additional 4 lanes that can be configured as two SATA 6 Gbps ports thanks to SerDes flexibility.

Alleged specifications of NVIDIA's upcoming GeForce RTX 3090, RTX 3080, and next-generation TITAN graphics cards, based on the "Ampere" graphics architecture, surfaced in tweets by KatCorgi, mirroring an early-June kopite7kimi tweet, sources with a high hit-rate on NVIDIA rumors. All three SKUs will be based on the 7 nm "GA102" silicon, but with varying memory and core configurations, targeting three vastly different price-points. The RTX 3080 succeeds the current RTX 2080/Super, and allegedly features 4,352 CUDA cores. It features a 320-bit GDDR6X memory interface, with its memory ticking at 19 Gbps.

The RTX 3090 is heir-apparent to the RTX 2080 Ti, and is endowed with 5,248 CUDA cores, 12 GB of GDDR6X memory across a 384-bit wide memory bus clocked at 21 Gbps. The king of the hill is the TITAN Ampere, succeeding the TITAN RTX. It probably maxes out the GA102 ASIC with 5,326 CUDA cores, offers double the memory amount of the RTX 3090, at 24 GB, but at lower memory clock speeds of 17 Gbps. NVIDIA is expected to announce these cards in September, 2020.

Benchmarks of the new Apple-exclusive AMD Radeon Pro 5600M graphics solution by Max Tech reveals that the new GPU is about 50% faster than the Radeon Pro 5500M, and within striking distance of the Radeon Pro Vega 48 found in Apple's 5K iMacs. The Pro 5600M is an Apple-exclusive solution by AMD, based on the "Navi 12" silicon that features a 7 nm GPU die based on the RDNA graphics architecture, flanked by two 4 GB HBM2 memory stacks over a 2048-bit interface. The GPU die features 2,560 stream processors, but clocked differently from Radeon Pro discrete graphics cards based on the "Navi 10" ASIC that uses conventional GDDR6.

The Radeon Pro 5600M solution was found to be 50.1 percent faster than the Radeon Pro 5500M in Geekbench 5 Metal (another Apple-exclusive SKU found in 16-inch MacBook Pros), and just 12.9 percent behind the Radeon Vega 48. The Vega 56 found in iMac Pro is still ahead. Unigine Heaven sees the Pro 5600M being 48.1% faster than the Pro 5500M, and interestingly, faster than Vega 48 by 11.3%. With 2,560 RDNA stream processors, you'd expect more performance, but this card was designed to meet stringent power limits of 50 W, and has significantly lower clock-speeds than "Navi 10" based Radeon Pro graphics cards (1035 MHz max boost engine clock vs. 1930 MHz and 205 W TDP of the Pro W5700). Find more interesting commentary in the Max Tech video presentation.

It looks like AMD's Ryzen 4000G line of socket AM4 desktop APUs based on the 8-core 7 nm "Renoir" silicon will be a lot wider than just a couple of SKUs. We've seen plenty of material on the top Ryzen 7 4700G part that maxes out everything on the silicon, along with increased power limits and clock speeds. It looks like the Ryzen 5 4000G series will consist of 6-core/12-thread parts. One such chip, the Ryzen 5 4400G surfaced on the 3DMark database, as dug up by TUM_APISAK. They earlier brought you a 3DMark score comparison between the 4400G, the top 4700G, and the entry-level 4200G.

The Ryzen 5 4400G (possible OPN: 100-000000143) appears to be a 6-core/12-thread part based on "Renoir," with the CPU clocked at 3.70 GHz base and possibly 4.30 GHz boost. The "Vega" NGCU count of the iGPU is unknown, but its engine clock is set at 1.90 GHz (max). With the "P" (performance) preset, the 4400G allegedly scores 4395 points in the 3DMark 11 graphics test suite (graphics score); with 10241 points physics score.

TSMC is reportedly planning a stopgap between its 5 nm-class silicon fabrication nodes, and the 3 nm-class, called N4. According to the foundry's CEO, Liu Deyin, speaking at a shareholders meeting, N4 will be a 4 nm node, and an enhancement of N5P, the company's most advanced 5 nm-class node. N4 is slated for mass-production of contracted products in 2023, and could help TSMC's customers execute their product roadmaps of the time. From the looks of it, N4 is a repeat of the N6 story: a nodelet that's an enhancement of N7+, the company's most advanced 7 nm-class node that leverages EUV lithography.

Earlier this week, we brought you a report about codenames of AMD processors that won't launch before 2022. It referenced "Raphael" being distant 5 nm "Zen 4" based successor to today's "Matisse." At the time, the codename for the 2021 release of AMD's mainstream desktop processor wasn't known. We're now getting a pointer as to what it is - "Warhol."

Named after American artist and filmmaker Andy Warhol, this processor combines CPU chiplets based on the "Zen 3" with a cIOD that retains PCI-Express gen 4.0, just like "Vermeer," but still qualifies as a new generation (and not a refresh). What's more, "Warhol" apparently sticks to a 7 nm-class silicon fabrication process. This means that "Warhol" could see AMD innovate on other fronts, such as leveraging an even more advanced version of TSMC's 7 nm node (such as N7+), to increase core counts over the chiplet that makes it to "Vermeer, "Genesis Peak," and "Milan."

AMD's RDNA2 graphics architecture, which sees real-time ray-tracing among other DirectX 12 Ultimate features, could see the company double the amount of stream processors generation-over-generation, according to a specs leak by _rogame. The increase in stream processors would fall in line with AMD's effort to increase performance/Watt by 50%. It may appear like the resulting SKUs finally measure up to the likes of the RTX 2080 Ti, but AMD has GeForce "Ampere" in its competitive calculus, and should the recent specs reveal hold up, the new "Navi 21" could end up being a performance-segment competitor to GeForce graphics cards based on the "GA104" ("TU104" successor), rather than a flagship-killer.

The RDNA2-based "Navi 21" GPU allegedly features 80 RDNA2 compute units amounting to 5,120 stream processors. AMD might tap into a refined 7 nm-class silicon fabrication node by TSMC to build these chips, either N7P or N7+. The die-size could measure up to 505 mm², and AMD could aim for a 50% performance/Watt gain over the "Navi 10." AMD could carve out as many as 10 SKUs out of the "Navi 21," but only three are relevant to the gamers. The SKU with the PCI device ID "0x731F: D1" succeeds the RX 5700 XT. The one bearing "0x731F: D3" succeeds the RX 5700, with a variant name "Navi 21 XL." The "Navi 21 XE" variant has a PCI ID of "0x731F: DF," and succeeds the RX 5600 XT.

AMD is planning to immediately update its product stack to counter the Intel 10th gen Core "Comet Lake-S" desktop processor family. Codenamed "Matisse Refresh," the processor will use existing IP, based on the 7 nm "Zen 2" microarchitecture, but could improve in areas such as clock-speeds. As it now stands, the Ryzen 9 3900X appears unfazed by the i9-10900K and i7-10700K at its new $410 price, however, competitiveness of the 3800X and 3700X could buckle under pressure from the i7-10700 series (K, KF, non-K, and F), as well as the Core i5-10600 series. To this effect, we're hearing rumors of a "Ryzen 7 3750X" and "Ryzen 7 3850X" seeing the light of the day soon, with an early-June announcement, and early-July market availability. References to the 3750X date back to October 2019.

Rumors of "Matisse Refresh" gained traction when WCCFTech editor Hassan Mujtaba tweeted a slide from a GIGABYTE AMD B550 motherboard series pre-launch presentation, which references GIGABYTE's own interpretation of AMD's roadmap. It lists out every CPU microarchitecture for the AM4 platform, and right next to "Matisse" is "& Refresh," confirming that "Matisse Refresh" is real. A microarchitecture "refresh" needn't even involve any physical changes to the processor design, core-counts, or architecture, and can sometimes even indicate something as simple as a second major wave of SKUs that replace existing SKUs in the market, leading to their phase-out (eg: Intel "Haswell Refresh" retaining the 4th gen Core model numbering). The slide also adds weight to the theory that desktop "Renoir," like its mobile counterpart, lacks PCIe gen 4.0. The slide also talks about AMD introducing the entry-level A520 desktop chipset in August, which will support PCIe gen 4 when paired with a capable processor.

Samsung Electronics Co., Ltd., a world leader in advanced semiconductor technology, today announced plans to boost its foundry capacity at the company's new production line in Pyeongtaek, Korea, to meet growing global demand for cutting-edge extreme ultraviolet (EUV) solutions.

The new foundry line, which will focus on EUV-based 5 nanometer (nm) and below process technology, has just commenced construction this month and is expected to be in full operation in the second half of 2021. It will play a pivotal role as Samsung aims to expand the use of state-of-the-art process technologies across a myriad of current and next generation applications, including 5G, high-performance computing (HPC) and artificial intelligence (AI).

Semiconductor Manufacturing International Corporation (SMIC), the state-backed Mainland Chinese semiconductor foundry, announced that it commenced mass-production of 14 nm FinFET SoCs for Huawei's HiSilicon subsidiary, a mere one month since Huawei shifting chip orders from TSMC to it. The company is manufacturing Kirin 710A is a revision of the original Kirin 710 SoC from 2018, built on SMIC's 14 nm node. The 4G-era SoC is capable of powering mid-range smartphones for Huawei's Honor brand, and uses an Arm big.LITTLE setup of Cortex A53 and Cortex A57 cores. This represents a major milestone not just for SMIC, but also Huawei, which has seen the company's isolation from cutting-edge overseas fabs such as TSMC. Much of Huawei's fate is riding on the success of SMIC's next-generation N+1 node, which purportedly offers a 57 percent energy-efficiency gain over 14 nm FinFET, rivaling sub-10 nm nodes such as 7 nm; enabling Huawei to build 5G-era SoCs.

Here is the first picture of the AMD Ryzen 7 4700G, the company's upcoming socket AM4 APU based on the 7 nm "Renoir" silicon, courtesy of VideoCardz. The picture reveals a standard-looking socket AM4 chip with commercial name and OPN markings (100-000000146), matching the Igor's Lab OPN code leak from earlier this week. The Ryzen 7 4700G offers an 8-core/16-thread CPU based on the "Zen 2" microarchitecture, and an integrated graphics solution that combines the SIMD machinery of the "Vega" graphics architecture, with the updated display- and media engines of "Navi." The iGPU is configured with 8 CUs (512 stream processors), which on the 4700G has an impressive maximum engine boost clock of 2.10 GHz, according to the Igor's Lab story.

The 8-core/16-thread CPU of the Ryzen 7 4700G has a nominal clock speed of 3.60 GHz, and a maximum boost frequency of 4.45 GHz, with several Precision Boost power-states in both directions of the nominal clock. The CPU features 512 KB of L2 cache per core, and 8 MB of shared L3 cache (4 MB per CCX). The iGPU engine clock goes all the way up to 2.10 GHz, which could help it overcome some of the CU deficit vs. "Picasso," which has 11 CUs (704 stream processors), but clocked only up to 1.40 GHz. Since the Ryzen 5 3400G has an unlocked multiplier, it stands to reason that even the 4700G could. If the platform I/O of "Renoir" in its mobile avatar is anything to go by, then the 4700G could feature a limited PCI-Express x8 lane setup for its PEG port. AMD is rating the TDP of the 4700G at 65 W.

Not long ago, Intel's Raja Koduri claimed that the Xe HP "Ponte Vecchio" silicon was the "big daddy" of Xe GPUs, and the "largest chip co-developed in India," larger than the 35 billion-transistor Xilinix VU19P FPGA co-developed in the country. It turns out that NVIDIA is in the mood for setting records. The "Ampere" A100 silicon has 54 billion transistors crammed into a single 7 nm die (not counting transistor counts of the HBM2E memory stacks).

NVIDIA claims a 20 Times boost in both AI inference and single-precision (FP32) performance over its "Volta" based predecessor, the Tesla V100. The chip also offers a 2.5X gain in FP64 performance over "Volta." NVIDIA has also invented a new number format for AI compute, called TF32 (tensor float 32). TF32 uses 10-bit mantissa of FP16, and the 8-bit exponent of FP32, resulting in a new, efficient format. NVIDIA attributes its 20x performance gains over "Volta" to this. The 3rd generation tensor core introduced with Ampere supports FP64 natively. Another key design focus for NVIDIA is to leverage the "sparsity" phenomenon in neural nets, to reduce their size, and improve performance.