Category Archives: Amd

AMD’s 12- and 24-core Threadripper 2 CPUs arrive October 29th

You're in luck if you were enticed by AMD's second-generation Threadripper CPUs but felt that going all-out on the 32-core model was a little too over-the-top. The chip designer has announced that the 12-core Threadripper 2920X and 24-core 2970WX wi…

Nvidia RTX 2080 and 2080 Ti review: A tale of two very expensive graphics cards

Sam Machkovech

Specs at a glance: Nvidia GeForce RTX 2080 Ti Founders Edition
CUDA CORES 4352
TEXTURE UNITS 272
ROPS 88
CORE CLOCK 1,350MHz
BOOST CLOCK 1,635MHz
MEMORY BUS WIDTH 352 bits
MEMORY BANDWIDTH 616GB/s
MEMORY SIZE 11GB GDDR6
Outputs 3x DisplayPort 1.4, 1x HDMI 2.0b, 1x USB Type-C (VirtualLink VR)
Release date September 20, 2018
PRICE Founders Edition (as reviewed): $1,199. Partner cards priced at: $1,169.

Like any piece of expensive technology, a top-of-the-line graphics card comes with all manner of lingo and abbreviation. You’ll need a glossary to wade through the stuff inside (processors, CUDA cores, ROPs), the speeds measured (memory bandwidth, boost clocks, TeraFLOPS), and the results you want from a good card (anti-aliasing, frame rates, higher resolutions).

Thanks to Nvidia’s newest products, the RTX 2080 and RTX 2080 Ti, that required glossary is only getting bigger.

Read 61 remaining paragraphs | Comments

Blackmagic external GPU review: A very Apple graphics solution

Samuel Axon

Is the eGPU the future of Mac graphics? Apple wants you to think so, and, for the first time, it has worked with an outside firm (Blackmagic Design) to produce a sanctioned eGPU solution for the Mac. Simply titled the Blackmagic Design eGPU, it’s available for $699 exclusively through the Apple Store.

It includes a Radeon Pro 580 GPU with 8GB of GDDR5 memory, two Thunderbolt 3 ports, four USB-3 ports, and one HDMI 2.0 port—and it’s only compatible with Macs with Thunderbolt 3 ports. When Apple first mentioned this product, it claimed that it’s easier to set up, quieter, and smaller than most existing external GPU enclosures.

Read 30 remaining paragraphs | Comments

AMD’s next-gen CPUs, GPUs will all be built on TSMC’s 7nm process

Article intro image

Enlarge / An AMD Ryzen, built on GlobalFoundries’ 14nm process. (credit: Fritzchens Fritz)

In a big change of alignments, AMD has announced that its next generation of CPUs and GPUs will be manufactured by TSMC, not GlobalFoundries.

GlobalFoundries was spun off from AMD in 2009. The once-integrated chip company split into two parts: GlobalFoundries took all the manufacturing facilities, leaving AMD as a fabless chip company. This gave AMD a big infusion of cash at a time it needed it, and it allowed GlobalFoundries to build chips for a wide range of customers. This close relationship has positioned GloFo as AMD’s preferred manufacturer, though TSMC and Samsung have offered alternative facilities. AMD’s current Ryzen, Threadripper, and Epyc chips are all built by GloFo on its 14nm and 12nm processes. AMD uses TSMC for GPUs and the custom chips for Sony’s and Microsoft’s consoles.

But GlobalFoundries has ceased its development of its next-generation 7nm process. Instead, the company will continue to develop its 14nm and 12nm processes and will focus on tuning them for radios, memory, and low power, making them a better fit for new high-growth markets such as highly integrated systems-on-chips and new 5G cellular components. That direction, however, comes at the expense of traditional high-performance processors.

Read 5 remaining paragraphs | Comments

AMD prevails in patent fight with Vizio over TV graphics

AMD's legal blitz against TV makers over graphics patents has claimed a major casualty. The US International Trade Commission has issued a final determination that Vizio and chip supplier Sigma Designs violated an AMD patent for a parallel pipeline…

The blockchain bonanza is over for graphics card makers

Enlarge / A Philadelphia cryptocurrency miner snapped this shot of his rig for us earlier this year. (credit: Matthew Freilich)

For almost a year, cryptocurrency miners have snapped up all the graphics cards they could get their hands on. That was a financial windfall for Nvidia and AMD, the leading makers of consumer graphics cards. Both reported soaring profits their last two quarters.

But on Thursday, Nvidia reported its financial results for its second fiscal quarter, which ended on July 29. The results were pretty good overall, with strong demands for Nvidia products for AI and data center applications. However, cryptocurrency-related demand has cratered.

“Our revenue outlook had anticipated cryptocurrency-specific products declining to approximately $100 million,” said Nvidia CFO Colette Kress. “Actual crypto-specific product revenue was $18 million.”

Read 6 remaining paragraphs | Comments

AMD’s Radeon Pro WX8200 is for content creators on a budget

AMD has been pushing Intel on the CPU high-end with new, more budget-minded offerings like the 32-core Ryzen Threadripper, and now it's doing the same to NVIDIA. Just ahead of Siggraph 2018, AMD launched the Radeon Pro WX 8200, a card aimed not at ga…

Second-generation Threadripper goes on sale: $1,799 for 32 cores

Enlarge (credit: AMD)

Announced back in June, AMD’s second-generation Threadripper processors are now available for preorder. The top-end part, the Threadripper 2990WX, is a 32-core, 64-thread processor with a total of 64MB of level 3 cache, running at a base clock of 3.0GHz, boosting as high as 4.2GHz. It ships on August 13 and is selling for $1,799.

Model Cores/Threads Clock base/boost/GHz Level 3 cache/MB TDP/W Price Availability
2990WX 32/64 3.0/4.2 64 250 $1,799 August 13
2970WX 24/48 3.0/4.2 64 250 $1,299 October
2950X 16/32 3.5/4.4 32 180 $899 August 31
2920X 12/24 3.5/4.3 32 180 $649 October

The basic building blocks of the second-generation Threadrippers are the same as the first-generation parts. Threadripper processors are multi-chip modules (MCMs) containing multiple dies and Infinity Fabric interconnects. AMD calls the basic building block of each chip a Core Complex (CCX), which has four cores, eight threads, and 8MB of level 3 cache. Each chip contains two CCXes. The first round of Threadrippers had four chips, with two of them active and two inactive, for a total of 16 cores and 32 threads. The new second-generation parts announced today make all four chips active, bringing the counts up to 32 cores and 64 threads.

The new chips increase the total power to 250W compared to 180W for the first-gen parts. However, AMD has said that the new processors will work in existing motherboards using the X399 chipset. Motherboards that can’t deliver substantially more than 180W will see limits to overclocking and turbo boosting, but they should nonetheless work correctly. New boards built for the second-gen parts should offer a bit more headroom.

Read 3 remaining paragraphs | Comments

AMD’s wild 32-core Ryzen Threadripper chip goes on sale

There's good news for graphics pros: AMD's 32-core, 64-thread Ryzen Threadripper 2990WX chip unveiled in June is now up for pre-order. The 2nd-generation chip is built using AMD's 12-nanometer Zen+ core architecture, and works with existing X399 AMD…

New Spectre attack enables secrets to be leaked over a network

Enlarge (credit: Pete)

When the Spectre and Meltdown attacks were disclosed earlier this year, the initial exploits required an attacker to be able to run code of their choosing on a victim system. This made browsers vulnerable, as suitably crafted JavaScript could be used to perform Spectre attacks. Cloud hosts were susceptible, too. But outside these situations, the impact seemed relatively limited.

That impact is now a little larger. Researchers from Graz University of Technology including one of the original Meltdown discoverers, Daniel Gruss, have described NetSpectre: a fully remote attack based on Spectre. With NetSpectre, an attacker can remotely read the memory of a victim system without running any code on that system.

All the variants of the Spectre attacks follow a common set of principles. Each processor has an architectural behavior (the documented behavior that describes how the instructions work and that programmers depend on to write their programs) and a microarchitectural behavior (the way an actual implementation of the architecture behaves). These can diverge in subtle ways. For example, architecturally, a program that loads a value from a particular address in memory will wait until the address is known before trying to perform the load. Microarchitecturally, however, the processor might try to speculatively guess at the address so that it can start loading the value from memory (which is slow) even before it’s absolutely certain of which address it should use.

Read 10 remaining paragraphs | Comments