News Posts matching "HBM2"

Return to Keyword Browsing

BIOSTAR Announces Racing B350 Motherboards and Radeon RX Vega 56 Graphics Card

BIOSTAR recently released the RADEON RX VEGA 56, which is a must-have for miners with the ability to crank out crazy hash rates for mining, while gamers can expect high performance graphics processor power with the same graphics card. Combined with the BIOSTAR RACING B350 series, which continues to be popular amongst gamers with its price-performance and gaming features, these make up for a great solution for mining during the day and gaming at night.

The Vega 56 has hash rate/power draw ratio that puts it in a class by its own, surpassing that of the RX 580, which was previously the graphics card of choice for mining. With the hype surrounding its mining capabilities, lets not forget that Vega 56 is a top-of-the-line graphics card that includes 56 compute units, 21/10.5 TFLOPS with 8GB of 2048-bit High Bandwidth Memory (HBM2) for performance gaming that gives you the best graphics possible.

AMD To Change Suppliers for Vega 20 GPUs on 7nm, HBM2 Packaging for Vega 11

AMD's RX Vega supply has seen exceedingly limited quantities available since launch. This has been due to a number of reasons, though the two foremost that have been reported are: increased demand from cryptocurrency miners, who are looking towards maximizing their single node hashrate density through Vega's promising mining capabilities; and yield issues with AMD's Vega 10 HBM2 packaging partner, Advanced Semiconductor Engineering (ASE). It's expected that chip yield for Vega 10 is also lower per se, due to it having a 484 mm² die, which is more prone to defects than a smaller one, thus reducing the amount of fully-enabled GPUs.

AMD's production partner, GlobalFoundries, has historically been at the center of considerations on AMD's yield problems. That GlobalFoundries is seemingly doing a good job with Ryzen may not be much to say: those chips have incredibly small die sizes (192 mm²) for their number of cores. It seems that Global Foundries only hits problems with increased die sizes and complexity (which is, unfortunately for AMD, where it matters most).

RX Vega Achieves 43 MH/s @ 130 W in Ethereum Mining

AMD's RX Vega is more along the lines of an original computing card that was moved over to the consumer segment for gaming workloads than the other way around. Raja Koduri himself has said something along those lines (extrapolating a little more than what he can actually say), and that much can be gleaned with at least a modicum of confidence through AMD's market positioning and overall computing push. In the argument between gamers and miners, Raja Koduri didn't have all that much to say, but for AMD, a sale is a sale, and it would seem that after some tweaking, RX Vega graphics cards can achieve much increased levels of mining efficiency than their Polaris counterparts, further showing how Vega handles compute workloads much better - and more efficiently - than traditional gaming ones.

AMD RX Vega 56 to Vega 64 BIOS Flash - No Unlocked Shaders, Improved Performance

A ChipHell forum user has done what probably others have already done in relative obscurity: trying (and succeeding) to flash a Vega 64 BIOS onto a Vega 56 graphics card. The result? Well, apparently the shaders won't unlock (at least not according to our very own GPU-Z), but interestingly, performance improves all the same. The lesser amount of shaders on the Vega 56 silicon (3585 Shaders / 224 TMUs / 64 ROPs compared to Vega 64's 4096 / 256 / 64 apparently doesn't hinder performance that much. It appears that the improved clockspeeds of Vega 56 after the BIOS flash do more than enough to offset performance loss from the lesser amount of compute resources available, bumping RX Vega's clock speeds of 1471 MHz core boost clock and 800 MHz HBM2 memory up to Vega 64's 1545 MHz core boost clock and 945 MHz HBM2 clock.

This means that Vega 56 can effectively become a Vega 64 in performance (at least where 3D Mark Fire Strike is concerned), which isn't unheard of in the relationship between AMD's top tier and second-best graphics cards. Now naturally, some Vega 56 samples may even be further overclocked than Vega 64's stock clocks, which means that there is the potential for Vega 56 to have even better performance than Vega 64. The BIOS swap should allow Vega 56 to access higher power states than its stock BIOS allows, which is one of the reasons it can unlock higher core and memory clocks than an overclocked, original BIOS Vega 56 would. However, the fact that a Vega 56 at Vega 64 clocks and a Vega 64 deliver around the same score in benchmarks definitely does raise questions on how well the extra computing resources of Vega 64 are being put to use.

Sources: ChipHell, via Videocardz

AMD RX Vega Supply Issues to Persist At Least Until October - Digitimes

DigiTimes is reporting, through "sources from the upstream supply chain", that AMD's current shortage of RX Vega cards to distribute to the retail market will continue at least until October. The tech reporting site says that sources are pointing towards the package integration of HBM2 memory (from SK Hynix or Samsung Electronics) and the Vega GPU (manufactured on Global Foundries' 14 nm FinFet) as being at fault here, due to low yield rates for this packaging effort. However, some other sources point towards the issue being with the packaging process itself, done by Advanced Semiconductor Engineering (ASE) through use of SiP technology. Whichever one of these cases may be, it seems the problem lies with AMD's choice to use HBM2 on their Vega graphics architecture.

As a footnote to its story, DigiTimes is also reporting that according to industry sources, NVIDIA has, in light of RX Vega's performance, decided to postpone the launch of Volta-based GPUs towards the first quarter of 2018.

Sources: Digitimes, via HardOCP

ASUS Announces ROG STRIX Radeon RX Vega Series

ASUS today introduced the Republic of Gamers (ROG) STRIX Radeon RX Vega 64 O8G graphics card, among its first (and probably the first) custom-design RX Vega 64 to hit the markets (model: ROG-STRIX-RXVEGA64-O8G-GAMING). The card combines a custom-design PCB by ASUS, with the company's latest generation DirectCU III cooling solution the company deploys on its STRIX GTX 1080 Ti graphics card. The cooler features a heat-pipe direct-contact base, from which the heat-pipes pass through two aluminium fin-stacks on their two ends, which are ventilated by a trio of 100 mm spinners. The fans stay off when the GPU is idling. The cooler features RGB multi-color LED lighting along inserts on the cooler shroud, and an ROG logo on the back-plate.

Moving over to the sparsely populated PCB (thanks in part to AMD's HBM2 move), the card draws power from a pair of 8-pin PCIe power connectors, conditioning it for the GPU with a 13-phase VRM. The O8G variant features factory-overclocked speeds that are close to those of the RX Vega 64 Liquid Edition, although ASUS didn't specify them. There's a "non-O8G" variant that sticks to reference clock speeds, boosting to around 1495-1510 MHz. What ASUS is really selling here is better clock sustainability under load, lower noise, and zero idle-noise; besides all the ROG STRIX bells and whistles. The card also drives two 4-pin PWM case fans in-sync with the cards, like most ROG STRIX graphics cards from this generation. ASUS also rolled out the ROG STRIX RX Vega 56, which features the same exact PCB, and sticks to AMD reference speeds. The company didn't reveal pricing.

AMD RX Vega 56 Benchmarks Leaked - An (Unverified) GTX 1070 Killer

TweakTown has put forth an article wherein they claim to have received info from industry insiders regarding the upcoming Vega 56's performance. Remember that Vega 56 is the slightly cut-down version of the flagship Vega 64, counting with 56 next-generation compute units (NGCUs) instead of Vega 64's, well, 64. This means that while the Vega 64 has the full complement of 4,096 Stream processors, 256 TMUs, 64 ROPs, and a 2048-bit wide 8 GB HBM2 memory pool offering 484 GB/s of bandwidth, Vega 56 makes do with 3,548 Stream processors,192 TMUs, 64 ROPs, the same 8 GB of HBM2 memory and a slightly lower memory bandwidth at 410 GB/s.

The Vega 56 has been announced to retail for about $399, or $499 with one of AMD's new (famous or infamous, depends on your mileage) Radeon Packs. The RX Vega 56 card was running on a system configured with an Intel Core i7-7700K @ 4.2GHz, 16 GB of DDR4-3000 MHz RAM, and Windows 10 at 2560 x 1440 resolution.

Everything AMD Launched Today: A Summary

It has been a huge weekend of product announcements and launches from AMD, which expanded not just its client computing CPU lineup on both ends, but also expanded its Radeon graphics cards family with both client- and professional-segment graphics cards. This article provides a brief summary of everything AMD launched or announced today, with their possible market-availability dates.

AMD Announces the Radeon RX Vega Nano

AMD had two press days over the weekend to cover a whole bunch of announcements- Ryzen Threadripper, RX Vega, Vega Pro and more. They provided details galore on the Vega microarchitecture, and the two RX Vega versions. Well, make that three now. AMD at Capcaicin SIGGRAPH 2017 also announced and showed off the RX Vega Nano.

Raja Koduri, Senior Vice-president of AMD's Radeon Technology Group, came up on stage minutes ago and handed over one of the very few working samples of this new card to Tim Sweeney, founder of Epic Games and Unreal Engine. The RX Vega Nano is an update to the R9 Nano from 2015 which was based off AMD's "Fiji" microarchitecture and was the then king of efficiency and small form factor support. We were only just talking about how RX Vega missed an opportunity with HBM2 to provide another mITX GPU and looks like AMD agreed.

AMD Says Vega Frontier Edition "Gaming" and "Pro" Modes are Not Placeholders

AMD's Vega Frontier Edition was a release that seemingly left most users either scratching their heads in bewilderment or - more specifically - disappointed. Some of this disappointment seemed to stem from a desire to see the long-awaited RX Vega consumer graphics card performance in the wild - or at least snagging a preview of it. Alas, the Frontier Edition's gaming performance was a disappointment when one considers the expected performance of AMD's underlying hardware - 4096 Stream processors and 16 GB of HBM2 memory - as well as the fact that this is AMD's first high-performance architecture since the Fury line of graphics cards. But to be fair to AMD, they did warn us - the Frontier Edition isn't the right graphics card for gamers.

One of the points of contention for this new release was that AMD delivered a graphics card that straddled the prosumer equation - offering both Pro drivers for professional workloads, and a Gaming Mode which should allow developers to seamlessly jump from development mode to testing mode through a driver toggle. However, when used at launch of the Frontier Edition - and even now - this toggle is little more than a dud. Mostly, what it does is remove the Wattman control panel.

Samsung Increases Production of 8 GB HBM2 Memory

Samsung Electronics Co., Ltd., the world leader in advanced memory technology, today announced that it is increasing the production volume of its 8-gigabyte (GB) High Bandwidth Memory-2 (HBM2) to meet growing market needs across a wide range of applications including artificial intelligence, HPC (high-performance computing), advanced graphics, network systems and enterprise servers.

"By increasing production of the industry's only 8GB HBM2 solution now available, we are aiming to ensure that global IT system manufacturers have sufficient supply for timely development of new and upgraded systems," said Jaesoo Han, executive vice president, Memory Sales & Marketing team at Samsung Electronics. "We will continue to deliver more advanced HBM2 line-ups, while closely cooperating with our global IT customers."

Liquid Cooled AMD Radeon Vega Frontier Edition Now on Sale for $1,489.99

The liquid cooled version of AMD's latest graphics card meant for the "pioneering crowd" of prosumers has been made available over at SabrePC. It sports the exact same GPU you'd find on the air-cooled version, featuring all the same 4096 Stream Processors and 16 GB of HBM2 memory. The only differences are, and you guessed it, the higher cooling capacity afforded by the AIO solution, and the therefore increased TDP from the 300 W of the air-cooled version to a eyebrow-raising 375 W. That increase in TDP must come partially from the employed cooling solution, but also from an (for now, anecdotal) ability for the card to more easily sustain higher clocks, closer to its AMD-rated 1,630 MHz peak core clock.

You can nab one right now in that rather striking gold and blue color scheme, and have it shipped to you in 24H. Hit the source link for the SabrePC page.

Sources: SabrePC, Computerbase.de

NVIDIA Announces the Tesla V100 PCI-Express HPC Accelerator

NVIDIA formally announced the PCI-Express add-on card version of its flagship Tesla V100 HPC accelerator, based on its next-generation "Volta" GPU architecture. Based on the advanced 12 nm "GV100" silicon, the GPU is a multi-chip module with a silicon substrate and four HBM2 memory stacks. It features a total of 5,120 CUDA cores, 640 Tensor cores (specialized CUDA cores which accelerate neural-net building), GPU clock speeds of around 1370 MHz, and a 4096-bit wide HBM2 memory interface, with 900 GB/s memory bandwidth. The 815 mm² GPU has a gargantuan transistor-count of 21 billion. NVIDIA is taking institutional orders for the V100 PCIe, and the card will be available a little later this year. HPE will develop three HPC rigs with the cards pre-installed.

AMD Announces the Radeon Instinct MI25 Deep Learning Accelerator

AMD's EPYC Launch presentation focused mainly on its line of datacenter processors, but fans of AMD's new Vega GPU lineup may be interested in another high-end product that was announced during the presentation. The Radeon Instinct MI25 is a Deep Learning accelerator, and as such is hardly intended for consumers, but it is Vega based and potentially very potent in the company's portfolio all the same. Claiming a massive 24.6 TFLOPS of Half Precision Floating Point performance (12.3 Single Precision) from its 64 "next-gen" compute units, this machine is very suited to Deep Learning and Machine AI oriented applications. It comes with no less than 16 GBs of HBM2 memory, and has 484 GB/s of memory bandwidth to play with.

AMD Radeon Pro Vega Frontier Edition TDP and Pricing Revealed

AMD Radeon Pro Vega Frontier Edition goes on sale later this month (26 June). It is designed to provide a "gateway" to the "Vega" GPU architecture for graphics professionals and game developers alike, with the consumer graphics product, the Radeon RX Vega, is bound for late-July/early-August. Radeon Pro Vega Frontier Edition, being a somewhat "enterprise-segment" product, was expected to have slightly lower TDP than its consumer-graphics sibling, since enterprise-segment implementations of popular GPUs tend to have slightly restrained clock speeds. Apparently, AMD either didn't clock the Radeon Pro Vega Frontier Edition low, or the chip has extremely high TDP.

According to specifications put out by EXXACT, a retailer which deals with enterprise hardware, the air-cooled variant of the Radeon Pro Vega Frontier Edition has a TDP rated at 300W, while its liquid-cooled variant has its TDP rated as high as 375W. To put this in perspective, the consumer-segment TITAN Xp by NVIDIA has its TDP rated at 275W. EXXACT is claiming big performance advantages in certain enterprise benchmarks such as SPECVIEWPERF and Cinebench. In other news, the air-cooled Radeon Pro Vega Frontier Edition is reportedly priced at USD $1,199; while the liquid-cooled variant is priced at $1,799. Based on the 14 nm "Vega 10" silicon, the Pro Vega Frontier Edition features 4,096 stream processors and 16 GB of HBM2 memory across a 2048-bit memory interface.

Apple iMac Pro, the Most Powerful Mac Ever, Arrives This December

Apple today gave a sneak peek of iMac Pro, an entirely new workstation-class product line designed for pro users with the most demanding workflows. The all-new iMac Pro, with its gorgeous 27-inch Retina 5K display, up to 18-core Xeon processors and up to 22 Teraflops of graphics computation, is the most powerful Mac ever made. Featuring a stunning new space gray enclosure, iMac Pro packs incredible performance for advanced graphics editing, virtual reality content creation and real-time 3D rendering. iMac Pro is scheduled to ship in December starting at $4,999 (US).

In addition to the new iMac Pro, Apple is working on a completely redesigned, next-generation Mac Pro architected for pro customers who need the highest-end, high-throughput system in a modular design, as well as a new high-end pro display. "We're thrilled to give developers and customers a sneak peek at iMac Pro. This will be our fastest and most powerful Mac ever, which brings workstation-class computing to iMac for the first time," said John Ternus, Apple's vice president of Hardware Engineering. "We reengineered the whole system and designed an entirely new thermal architecture to pack extraordinary performance into the elegant, quiet iMac enclosure our customers love - iMac Pro is a huge step forward and there's never been anything like it."

AMD "Vega" Die-shot Revealed

AMD released the die-shot of its "Vega 10" ASIC. The multi-chip module (MCM) of the GPU die and two HBM2 memory stacks, sitting on a silicon interposer, is slightly smaller than the "Fiji" ASIC, as it features just two memory stacks. AMD didn't label the components of the GPU die, but we can make out 8 shader engines, holding 8 GCN compute units (CUs), each. This is unlike "Fiji," where the GPU holds four engines with 16 CUs, each. This would mean that each group of 8 CUs has its own dedicated geometry processor and rasterizer.

Since each CU holds 64 FP32 stream processors, we arrive at the total stream processor count of 4,096. Unlike "Hawaii," these stream-processors are FP16-capable, so simple compute tasks are handled at double the throughput. We also make out 32 render back-ends, double that of "Fiji" and "Hawaii," which could indicate 128 ROPs. The "Vega 10" ASIC features a 2048-bit HBM2 memory interface.

Update: AMD stated that the die-shot appeared in one of its marketing slides, but may not be real.

AMD Confirms Radeon RX Vega Soft-launch at Computex

AMD Radeon Technologies Group (RTG) head Raja Koduri, responding to questions on a Reddit AMA (ask me anything) session, confirmed that while the company will launch the consumer-graphics variant of "Vega," the Radeon RX Vega graphics card, at its 2017 Computex event, availability of the card won't follow immediately after, making it a soft-launch. "We'll be showing Radeon RX Vega off at Computex, but it won't be on store shelves that week. We know how eager you are to get your hands on Radeon RX Vega, and we're working extremely hard to bring you a graphics card that you'll be incredibly proud to own," Koduri said.

The first consumer graphics card based on the "Vega 10" ASIC will be the Radeon RX Vega Frontier Edition. This card will be armed with 8 GB of HBM2 memory spread across two 16 Gbit HBM2 8-Hi stacks, with its combined memory bandwidth around 480 GB/s. From the words of Koduri, we can deduce that AMD is still finding the right clocks to make Vega Frontier Edition a competitive product. Koduri confirmed that there will be faster/bigger implementations of Vega. "Consumer RX will be much better optimized for all the top gaming titles and flavors of RX Vega will actually be faster than Frontier version," he said. In the meantime, check out some groovy concept renders of RX Vega reference board by VideoCardz. Our money is on the one below.

Source: Reddit

AMD Announces Radeon Vega Frontier Edition - Not for Gamers

Where is Vega? When is it launching? On AMD's Financial Analyst Day 2017, Raja Koduri spoke about the speculation in the past few weeks, and brought us an answer: Radeon Vega Frontier Edition is the first iteration of Vega, aimed at data scientists, immersion engineers and product designers. It will be released in the second half of June for AMD's "pioneers". The wording, that Vega Frontier Edition will be released in the second half of June, makes it so that AMD still technically releases Vega in the 2H 2017... It's just not the consumer, gaming Vega version of the chip. This could unfortunately signify an after-June release time-frame for consumer GPUs based on the Vega micro-architecture.

This news comes as a disappointment to all gamers who have been hoping for Vega for gaming, because it reminds of what happened with dual Fiji. A promising design which ended up unsuitable for gaming and was thus marketed for content creators as Radeon Pro Duo, with little success. But there is still hope: it just looks like we really will have to wait for Computex 2017 to see some measure of details on Vega's gaming prowess.

SK Hynix Updates Memory Catalog to Feature GDDR6 and HBM2

South Korean DRAM and NAND flash giant SK Hynix updated its product catalog to feature its latest GDDR6 memory, besides HBM2. The company had April announced its first GDDR6 memory products. The first GDDR6 memory chips by SK Hynix come in 8 Gb (1 gigabyte) densities, and data-rates of 14 Gbps and 12 Gbps, with DRAM voltages of 1.35V. The company is giving away small quantities of these chips for product development, mass production will commence soon, and bulk availability is slated for Q4-2017. This would mean actual products implementing these chips could be available only by very-late Q4 2017, or Q1-2018.

A graphics card with 14 Gbps GDDR6 memory across a 256-bit memory bus (8 chips) features 448 GB/s of memory bandwidth. A card with 384-bit (12 chips), should have 672 GB/s at its disposal. Likewise, the 12 Gbps memory chips offer 384 GB/s in 256-bit (8-chip) setups, and 576 GB/s in 384-bit (12-chip) setups. Meanwhile, SK Hynix also updated its HBM2 catalog to feature a 32 Gb (4 gigabyte) HBM2 stack, with a clock speed of 1.60 Gbps. The 2.00 Gbps stack which featured in the Q4-2016 version of this catalog is no longer available. At 1.60 Gbps, a GPU with four stacks has 819.2 GB/s of memory bandwidth. A chip with two stacks, such as the purported "Vega 10" prototype that has made several media appearances, hence has 409.6 GB/s.

Source: SK Hynix

AMD Vega Makes an Appearance on CompuBench

An AMD RX Vega video card has apparently made its way towards CompuBench. Granted, the no-name AMD graphics card could be an Instinct accelerator instead of AMD's consumer-oriented RX Vega graphics cards. However, the card did appear on CompuBench under the 6864:00 device ID, which had already appeared under a Vega Linux patch issued by the company. granted, this doesn't necessarily make it a consumer graphics product, so we'll have to look into this with some reservations.

NVIDIA Announces Its Volta-based Tesla V100

Today at its GTC keynote, NVIDIA CEO Jensen Huang took the wraps on some of the features on their upcoming V100 accelerator, the Volta-based accelerator for the professional market that will likely pave the way to the company's next-generation 2000 series GeForce graphics cards. If NVIDIA goes on with its product carvings and naming scheme for the next-generation Volta architecture, we can expect to see this processor on the company's next-generation GTX 2080 Ti. Running the nitty-gritty details (like the new Tensor processing approach) on this piece would be impossible, but there are some things we know already from this presentation.

This chip is a beast of a processor: it packs 21 billion transistors (up from 15,3 billion found on the P100); it's built on TSMC's 12 nm FF process (evolving from Pascal's 16 nm FF); and measures a staggering 815 mm² (from the P100's 610 mm².) This is such a considerable leap in die-area that we can only speculate on how yields will be for this monstrous chip, especially considering the novelty of the 12 nm process that it's going to leverage. But now, the most interesting details from a gaming perspective are the 5,120 CUDA cores powering the V100 out of a total possible 5,376 in the whole chip design, which NVIDIA will likely leave for their Titan Xv. These are divided in 84 Volta Streaming Multiprocessor Units with each carrying 64 CUDA cores (84 x 64 = 5,376, from which NVIDIA is cutting 4 Volta Streaming Multiprocessor Units for yields, most likely, which accounts for the announced 5,120.) Even in this cut-down configuration, we're looking at a staggering 42% higher pure CUDA core-count than the P100's. The new V100 will offer up to 15 FP 32 TFLOPS, and will still leverage a 16 GB HBM2 implementation delivering up to 900 GB/s bandwidth (up from the P100's 721 GB/s). No details on clock speed or TDP as of yet, but we already have enough details to enable a lengthy discussion... Wouldn't you agree?

AMD Vega May Launch with Less Than 20,000 Units Available

Fresh from the rumor-mill comes a report that low HBM2 availability may cripple the Vega launch that is expected to happen in the next few weeks, if a report from TweakTown is to be believed. As far as sources, there isn't much other than TweakTown's news report and their article claiming they had been told this by an "exclusive industry source." Apply your usual grain of salt here vigilant reader, but its certainly interesting speculation, if nothing else. It may turn out to be FUD, or it may turn out to be truth. Only the coming weeks will reveal the truth.

Source: TweakTown

AMD Works on At Least Three Radeon RX Vega SKUs, Slowest Faster than GTX 1070?

AMD could be working on at least three SKUs based on its upcoming "Vega 10" silicon to make up its Radeon RX Vega series. Leaked 3DMark validations point to a device ID that's third in a series of possible device IDs of graphics cards based on the "Vega 10" silicon, the 687F:C1, 687F:C2, and 687F:C3. All three SKUs feature 8 GB of HBM2 memory, and according to leaked 3DMark TimeSpy scores, the "slowest" SKU is faster than NVIDIA GeForce GTX 1070. The fastest SKU is in the same league as the GTX 1080 Ti.

The three SKUs could differ with core-configuration and clock speeds. AMD carved four SKUs out of its "Fiji" silicon, the liquid-cooled R9 Fury X, the air-cooled R9 Fury (with 12.5% fewer shaders), the SFF-friendly R9 Nano (full core-config, but aggressive power-management), and the halo dual-GPU Radeon Pro Duo (1st gen). AMD could take a similar approach with "Vega 10." AMD is expected to launch its Radeon RX Vega series within Q2-2017.

Source: Guru3D Forums

AMD "Vega 10" Bears Core-Config Similarities to "Fiji"

A Linux patch for AMD's GPU drivers reveals that its upcoming "Vega 10" graphics processor bears numeric core-configuration similarities to the "Fiji" silicon which drives the enthusiast-segment Radeon R9 Fury series graphics cards. The patch bears configuration values which tell the software how to utilize the resources on the GPU, by spelling them out. The entry "gfx.config.max_shader_engines = 4," for example, indicates that "Vega 10" features four shader engines, like "Fiji." Another entry "Adev-> gfx.config.max_cu_per_sh = 16" signifies the number of GCN compute units (CUs) per shader engine. Assuming the number of stream processors per CU hasn't changed from 64 in the "Vega" architecture, we're looking at a total stream processor count of 4,096. This could also put the TMU count at 256.

At earlier reveals of the "Vega 10" package, you notice a large, somewhat square GPU die neighboring two smaller rectangular memory stack dies, which together sit on a shiny structure, which is the silicon interposer. The presence of just two memory stack dies sparked speculation that "Vega 10" features a narrower 2048-bit memory interface compared to the 4096-bit of "Fiji," but since the memory itself is newer-generation HBM2, which ticks at higher clocks, AMD could run them at double the memory clock as "Fiji" to arrive at the same 512 GB/s bandwidth. The 4,096 stream processors of "Vega 10" are two generations ahead of the ones on "Fiji," which together with 14 nm process-level improvements, could run at much higher GPU clocks, making AMD get back into the high-end graphics segment.
Sources: aceCrasher (Reddit), ComputerBase.de
Return to Keyword Browsing