News Posts matching #TU106

Return to Keyword Browsing

AMD Readies Larger 7nm "Navi 12" Silicon to Power Radeon RX 5800 Series?

AMD is developing a larger GPU based on its new "Navi" architecture to power a new high-end graphics card family, likely the Radeon RX 5800 series. The codename "Navi 12" is doing rounds on social media through familiar accounts that have high credibility with pre-launch news and rumors. The "Navi 10" silicon was designed to compete with NVIDIA's "TU106," as its "XT" and "Pro" variants outperform NVIDIA's original RTX 2060 and RTX 2070, forcing it to develop the RTX 20 Super series, by moving up specifications a notch.

Refreshing its $500 price-point was particularly costly for NVIDIA, as it was forced to tap into the 13.6 billion-transistor "TU104" silicon to carve out the RTX 2070 Super; while for the RTX 2060 Super, it had to spend 33 percent more on the memory chips. With the "Navi 12" silicon, AMD is probably looking to take a swing at NVIDIA's "TU104" silicon, which has been maxed out by the RTX 2080 Super, disrupting the company's $500-700 lineup once again, with its XT and Pro variants. There's also a remote possibility of "Navi 12" being an even bigger chip, targeting the "TU102."

NVIDIA Manufacturing Turing GPUs at Samsung Korea Fab, 11nm?

During our disassembly of the GeForce RTX 2060 Super, we noticed a shocking detail. The 12 nm "TU106" GPU on which it is based, has the marking "Korea." We know for a fact that TSMC does not have any fabs there. The only Korean semiconductor manufacturer capable of contract-manufacturing a piece of silicon as complex as a GPU, for a designer with the energy-efficiency OCD as NVIDIA, is Samsung.

What makes this interesting is that Samsung does not officially have a 12 nm FinFET process. It has 14 nm, and the 11LPP, a 11 nm nodelet, which the company designed to compete with TSMC 12 nm. It would hence be really interesting to hear from NVIDIA on whether they've scaled out the "TU106" to 14LPP, or down to 11LPP at Samsung. It's interesting to note that the shrink in transistor sizes in these nodelets doesn't affect die-sizes. We hence see no die-size difference between these Korea-marked chips, and those marked "Taiwan." We've reached out to NVIDIA for comment.

Update July 3rd: NVIDIA got back to us
NVIDIA
The answer is really simple and these markings are not new. Other Turing GPUs have had these markings in the past. The chip is made at TSMC, but packaged in various locations. This one was done in Korea, hence why his says "Korea".

On an unrelated note: We already use both TSMC and Samsung, and qualify each of them for every process node. We can't comment in any further detail on future plans, but both remain terrific partners.

EVGA RTX 2060 Super and RTX 2070 Super Pictured

Ahead of formal launch tomorrow (2nd July), followed by a 9th July market availability, pictures of EVGA's custom-design GeForce RTX 2060 Super and RTX 2070 Super graphics cards were leaked to the web by an early listing on Amazon. The RTX 2060 Super XC Ultra is a triple-slot monstrosity, which uses a thicker heatsink ventilated by a pair of spinners; while the RTX 2070 Super XC is the base variant of an entire stack of cards based on this chip. The boxes have clear "Super" branding in the SKU detail corners. This branding is also found on metal decals on the cooler shrouds, and printed on the back-plates.

The listing also confirms that the RTX 2060 Super features 8 GB of GDDR6 memory. This move was necessitated by AMD equipping its Radeon RX 5700 graphics card with 8 GB of 14 Gbps GDDR6 memory. The RTX 2060 Super makes full use of the 256-bit memory bus width of the "TU106" silicon. The RTX 2070 Super, on the other hand, is based on the larger "TU104" chip, and has 2,560 of the 3,072 CUDA cores present on the chip enabled. This SKU is designed to compete with the RX 5700 XT. We'll find pictures of custom-design RTX 2080 Super as we inch closer to its July 23 market availability date.

NVIDIA GeForce RTX 2060 Super Smiles for the Camera

Here are some of the first live pictures (not renders) of the upcoming GeForce RTX 2060 Super graphics card. As with the rest of the RTX 20 Super-series, this card features a reference board design resembling that of the original RTX 20-series, but with a chrome embellishment that accommodates the "Super" badge. The RTX 2060 Super is designed to compete with the upcoming Radeon RX 5700 at USD $399, or $50 more than the original RTX 2060. It's based on the "TU106" silicon, and is configured with 2,176 CUDA cores, but more importantly, a memory setup that's both 33 percent larger and faster than that of the original RTX 2060, made up of 8 GB of GDDR6 memory across a 256-bit wide bus, clocked at 14 Gbps. The card is expected to perform halfway between the RTX 2060 and RTX 2070.

NVIDIA's SUPER Tease Rumored to Translate Into an Entire Lineup Shift Upwards for Turing

NVIDIA's SUPER teaser hasn't crystallized into something physical as of now, but we know it's coming - NVIDIA themselves saw to it that our (singularly) collective minds would be buzzing about what that teaser meant, looking to steal some thunder from AMD's E3 showing. Now, that teaser seems to be coalescing into something amongst the industry: an entire lineup upgrade for Turing products, with NVIDIA pulling their chips up one rung of the performance chair across their entire lineup.

Apparently, NVIDIA will be looking to increase performance across the board, by shuffling their chips in a downward manner whilst keeping the current pricing structure. This means that NVIDIA's TU106 chip, which powered their RTX 2070 graphics card, will now be powering the RTX 2060 SUPER (with a reported core count of 2176 CUDA cores). The TU104 chip, which power the current RTX 2080, will in the meantime be powering the SUPER version of the RTX 2070 (a reported 2560 CUDA cores are expected to be onboard), and the TU102 chip which powered their top-of-the-line RTX 2080 Ti will be brought down to the RTX 2080 SUPER (specs place this at 8 GB GDDR6 VRAM and 3072 CUDA cores). This carves the way for an even more powerful SKU in the RTX 2080 Ti SUPER, which should be launched at a later date. Salty waters say the RTX 2080 Ti SUPER will feature and unlocked chip which could be allowed to convert up to 300 W into graphics horsepower, so that's something to keep an eye - and a power meter on - for sure. Less defined talks suggest that NVIDIA will be introducing an RTX 2070 Ti SUPER equivalent with a new chip as well.

NVIDIA RTX Logic Increases TPC Area by 22% Compared to Non-RTX Turing

Public perception on NVIDIA's new RTX series of graphics cards was sometimes marred by an impression of wrong resource allocation from NVIDIA. The argument went that NVIDIA had greatly increased chip area by adding RTX functionality (in both its Tensor ad RT cores) that could have been better used for increased performance gains in shader-based, non-raytracing workloads. While the merits of ray tracing oas it stands (in terms of uptake from developers) are certainly worthy of discussion, it seems that NVIDIA didn't dedicate that much more die area to their RTX functionality - at least not to the tone of public perception.

After analyzing full, high-res images of NVIDIA's TU106 and TU116 chips, reddit user @Qesa did some analysis on the TPC structure of NVIDIA's Turing chips, and arrived at the conclusion that the difference between NVIDIA's RTX-capable TU106 compared to their RTX-stripped TU116 amounts to a mere 1.95 mm² of additional logic per TPC - a 22% area increase. Of these, 1.25 mm² are reserved for the Tensor logic (which accelerates both DLSS and de-noising on ray-traced workloads), while only 0.7 mm² are being used for the RT cores.

EK Releases Vector RTX Series Blocks for ASUS ROG Strix Series Graphics Cards

EK Water Blocks, the Slovenia based water cooling gear manufacturer, is introducing its new generation water blocks for the popular ROG Strix GeForce RTX series graphics cards, based on Turing TU106, Turing TU104 and Turing TU102 graphics processor.

The inspiration for the new GPU block name "Vector" came from the sheer computing power of the graphics cards that are on the market today. Naming a water block "Full Cover" isn't enough these days, when the product is packed with unique features, such as these. The EK Vector Strix RTX water blocks are specially designed for multiple ROG Strix GeForce RTX Turing based graphics cards. The water block itself uses the signature EK single slot slim look, and it covers the entire PCB length. This sophisticated cooling solution will transform your powerful ROG graphics card into a minimalistic, elegant piece of hardware with accented RGB LED lighting. The block also features a unique aesthetic cover over the block Terminal which is designed to showcase the graphics card model via LEDs, visible from the side.

NVIDIA TU116 GPU Pictured Up Close: Noticeably Smaller than TU106

Here is the first picture of NVIDIA's 12 nm "TU116" silicon, which powers the upcoming GeForce GTX 1660 Ti graphics card. While the size of the package itself is identical to that of the "TU106" on which the RTX 2060 and RTX 2070 are based; the die of the TU116 is visibly smaller. This is because the chip physically lacks RT cores, and only has two-thirds the number of CUDA cores as the TU106, with 1,536 against the latter's 2,304. The die area, too, is about 2/3rds that of the TU106. The ASIC version of TU116 powering the GTX 1660 Ti is "TU116-400-A1."

VideoCardz scored not just pictures of the ASIC, but also the PCB of an MSI GTX 1660 Ti Ventus graphics card, which reveals something very interesting. The PCB has traces for eight memory chips, across a 256-bit wide memory bus, although only six of them are populated with memory chips, making up 6 GB over a 192-bit bus. The GPU's package substrate, too, is of the same size. It's likely that NVIDIA is using a common substrate, with an identical pin-map between the TU106 and TU116, so AIC partners could reduce PCB development costs.

GIGABYTE Outs GeForce RTX 2070 Gaming OC White Graphics Card

GIGABYTE extended its all-white trim to its third graphics card from the RTX 20-series, the RTX 2070 Gaming OC White (model: GV-N2070GAMINGOC WHITE-8GC). This follows similar trims for the RTX 2060 Gaming OC Pro White and RTX 2070 Gaming OC Pro White. As with the others, the card's USP is its mostly-white cooler shroud with chrome inserts, and a matching white metal back-plate. Contrasting them are a trio of 90 mm matte-black fans with chrome hub stickers, and the black PCB carried over from the original Gaming OC series.

As with the RTX 2070 Gaming OC, this card offers a factory-overclock of 1725 MHz boost, compared to 1620 MHz reference. The memory is untouched at 14 Gbps (GDDR6-effective). The cooler offers 0 dBA mode (idle fan-stop). The card draws power from a combination of 8-pin and 6-pin PCIe power connectors. Display outputs include three DisplayPort 1.4, one HDMI 2.0b, and a USB-C VirtualLink. Based on the 12 nm "TU106" silicon, the GeForce RTX 2070 offers 2,304 CUDA cores, 288 tensor cores, 36 RT cores, 144 TMUs, 64 ROPs, and a 256-bit wide GDDR6 memory interface, holding 8 GB of memory. GIGABYTE is offering a unique 4-year warranty with this card, if you register your purchase with them.

NVIDIA GeForce RTX 2060 Founders Edition Pictured, Tested

Here are some of the first pictures of NVIDIA's upcoming GeForce RTX 2060 Founders Edition graphics card. You'll know from our older report that there could be as many as six variants of the RTX 2060 based on memory size and type. The Founders Edition is based on the top-spec one with 6 GB of GDDR6 memory. The card looks similar in design to the RTX 2070 Founders Edition, which is probably because NVIDIA is reusing the reference-design PCB and cooling solution, minus two of the eight memory chips. The card continues to pull power from a single 8-pin PCIe power connector.

According to VideoCardz, NVIDIA could launch the RTX 2060 on the 15th of January, 2019. It could get an earlier unveiling by CEO Jen-Hsun Huang at NVIDIA's CES 2019 event, slated for January 7th. The top-spec RTX 2060 trim is based on the TU106-300 ASIC, configured with 1,920 CUDA cores, 120 TMUs, 48 ROPs, 240 tensor cores, and 30 RT cores. With an estimated FP32 compute performance of 6.5 TFLOP/s, the card is expected to perform on par with the GTX 1070 Ti from the previous generation in workloads that lack DXR. VideoCardz also posted performance numbers obtained from NVIDIA's Reviewer's Guide, that point to the same possibility.

NVIDIA GeForce RTX 2060 Could Launch Mid-January

NVIDIA could launch its "RTX for the masses" SKU, the GeForce RTX 2060, sometime mid-January, according to Andreas Schilling. It is also confirmed that the RTX 2060 will feature 6 GB of GDDR6 memory. Schilling confirmed no other specifications of the GPU, but posted official branding for the RTX 2060 SKU. Earlier leaks pin the RTX 2060 as being carved out from the 12 nm "TU106" silicon, with 1,920 CUDA cores, 120 TMUs, 48 ROPs, and a 192-bit wide GDDR6 memory interface, which at 14 Gbps produces a memory bandwidth of 336 GB/s. NVIDIA could target the crowd that wants DXR-enabled gaming at 1080p thru 1440p resolutions.

NVIDIA GeForce RTX 2070 and 2080 Mobile Could Make an Appearance at CES 2019

With NVIDIA's GeForce RTX 20-series having already released for desktops, it was only a matter of time until laptops got the RTX treatment as well. Current rumors are suggesting that Nvidia will officially launch their GeForce RTX 20-series mobility GPUs on January 6th at CES with the RTX 2070 and RTX 2070 Max-Q taking center stage. An embargo date of January 26th has also been set, with NVIDIA delaying their final release drivers until then. Meaning final performance results for the new mobile GPUs won't be available until after the embargo date, which should coincide with the general availability of RTX 20-series equipped laptops.

Along with the RTX 2070 and 2070 Max-Q mobility parts, the flagship RTX 2080 Max-Q which isn't expected at the show, is still in the works, with its TU104M 1eab device ID having been leaked earlier. The rest of the GeForce 20-series mobility GPUs are likely to use the GTX moniker if NVIDIA's desktop lineup is anything to go by; however, that is merely speculation at this point.

Palit Announces GeForce RTX 2070 Series Graphics Cards

Palit Microsystems Ltd, the leading graphics card manufacturer, releases the new NVIDIA Turing architecture GeForce RTX series in Palit GeForce product line-up, GeForce RTX 2070 Dual. NVIDIA GeForce RTX delivers the ultimate PC gaming experience. Powered by the new NVIDIA Turing GPU architecture and the revolutionary RTX platform, RTX graphics cards bring together real-time ray tracing, artificial intelligence, and programmable shading. This is a whole new way to experience games.

GeForce RTX 2070 is based on the NVIDIA Turing architecture TU106 GPU armed with 2304 CUDA cores and equipped with 8GB of GDDR6 memory running at an effective clock rate of 14 GHz on a 256-bit bus. Palit GeForce RTX 2070 is designed to deliver 6x more performance than previous-generation video card and brings the power of real-time ray tracing and AI to games. The Palit GeForce RTX 2070 Dual is targeting at MSRP $499 which is available in the market now.

NVIDIA Readies TU104-based GeForce RTX 2070 Ti

Update: Gigabyte themselves have come out of the gates dismissing this as a typo on their part, which is disappointing, if not unexpected, considering that there is no real reason for NVIDIA to launch a new SKU to a market virtually absent of competition. Increased tiers of graphics card just give more options for the consumer, and why give an option that might drive customers away from more expensive graphics card options?

NVIDIA designed the $500 GeForce RTX 2070 based on its third largest silicon based on "Turing," the TU106. Reviews posted late Tuesday summarize the RTX 2070 to offer roughly the the same performance level as the GTX 1080 from the previous generation, at the same price. Generation-to-generation, the RTX 2070 offers roughly 30% more performance than the GTX 1070, but at 30% higher price, in stark contrast to the GTX 1070 offering 65% more performance than the GTX 970, at just 25% more price. NVIDIA's RTX varnish is still nowhere in sight. That said, NVIDIA is not on solid-ground with the RTX 2070, and there's a vast price gap between the RTX 2070 and the $800 RTX 2080. GIGABYTE all but confirmed the existence of an SKU in between.

NVIDIA GeForce RTX 20-series Mobility mGPU Lineup Revealed

NVIDIA is giving finishing touches to its first GeForce RTX 20-series Mobility GPUs for notebooks, based on the "Turing" architecture, with product launches expected from Q1-2019. The company could debut the series with a high-end part first, the GeForce RTX 2080 Mobility Max-Q. The rest of the lineup includes the RTX 2070 Mobility Max-Q, RTX 2060 Ti Mobility, RTX 2060 Mobility, RTX 2050 Ti Mobility, and RTX 2050 Mobility. What's interesting about this list is that NVIDIA is limiting the Max-Q design to its top-tier RTX 2080 Mobility and RTX 2070 Mobility parts.

Max-Q is an all-encompassing laptop thermal-design methodology, which allows gaming notebook designers to come up with thinner notebooks with higher performance. One of the key aspects is special Max-Q ready variants of the GPUs, which are probably binned to run the coolest, and least voltages. With a device ID 1eab, the RTX 2080 Mobility Max-Q is based on the TU104M chip, while other SKUs could be carved from the TU106M or a chip even smaller. It's being reported that with this generation, NVIDIA is playing a more active role in helping its partners engineer their Max-Q notebooks, and helping them meet NVIDIA's strict Z-height minimums.

NVIDIA Announces Availability of GeForce RTX 2070 Graphics Card - Cheapest Raytracing on October 17th

NVIDIA today announced official availability dates for what will forever be engraved in history as "the cheapest Turing" option - which contrary to what that might lead you to expect, isn't cheap at all. NVIDIA's RTX 2070 graphics card will be available starting October 17th, bringing the benefits of raytracing acceleration to a much lower price-point than the already-launched RTX 2080 and RTX 2080 Ti graphics cards.

That said, the RTX 2070 will still retail for $499 - a full $120 higher than NVIDIA's last-gen GTX 1070, and that's not counting what's being less-than-amicably called the "NVIDIA tax", which brings Founders' Editions pricing of the graphics cards up to $599, and allows AIB to increase pricing of their own designs up to that level - or higher. It's not a cheap option - especially considering how the RTX 2070 is now being built in the TU106 silicon, a smaller counterpart to the full-fledged TU104, and in contrast to the previous GTX 1070, which was built from the same chip as the GTX 1080).

ASUS Announces Its NVIDIA RTX 2070 Graphics Card Lineup

ASUS has revealed their entire lineup, interpreting NVIDIA's RTX 2070 silicon (and its TU106 die, a first - remember that **70-class cards previously featured cut-down versions of the full NVIDIA chip). There aren't many surprised here - ASUS has already spent enough R&D in previous years so as to only need to minimally iterate on their designs for each new generation.

The ROG Strix graphics cards sit at the top of the heap, featuring the company's DirectCU III cooling tech (triple fan) in a 2.5-slot design. RGB lighting and dual BIOS support are par of the course by now, as are some of the other features - backplate and metal brace included. Connectivity-wise there are 2x HDMI 2.0b ports, 2x DisplayPort 1.4 outputs, and 1x USB Type-C port for VirtuaLink. The graphics card draws power from the 6-pin and 8-pin PCIe power connectors and will be available in three versions (Gaming OC, Gaming Advanced, and Gaming) according to factory overclocks.

NVIDIA Segregates Turing GPUs; Factory Overclocking Forbidden on the Cheaper Variant

While working on GPU-Z support for NVIDIA's RTX 20-series graphics cards, we noticed something curious. Each GPU model has not one, but two device IDs assigned to it. A device ID is a unique identification that tells Windows which specific device is installed, so it can select and load the relevant driver software. It also tells the driver, which commands to send to the chip, as they vary between generations. Last but not least, the device ID can be used to enable or lock certain features, for example in the professional space. Two device IDs per GPU is very unusual. For example, all GTX 1080 Ti cards, whether reference or custom design, are marked as 1B06. Titan Xp on the other hand, which uses the same physical GPU, is marked as 1B02. NVIDIA has always used just one ID per SKU, no matter if custom-design, reference or Founders Edition.

We reached out to industry sources and confirmed that for Turing, NVIDIA is creating two device IDs per GPU to correspond to two different ASIC codes per GPU model (for example, TU102-300 and TU102-300-A for the RTX 2080 Ti). The Turing -300 variant is designated to be used on cards targeting the MSRP price point, while the 300-A variant is for use on custom-design, overclocked cards. Both are the same physical chip, just separated by binning, and pricing, which means NVIDIA pretests all GPUs and sorts them by properties such as overclocking potential, power efficiency, etc.
Return to Keyword Browsing