Items tagged with hbm2e
NVIDIA kicked off its 7-nanometer Ampere party with the A100, a massively powerful machine learning GPU that began shipping to customers a little over a year ago. The company would later flesh out its machine learning lineup with an 80GB model, though up until now, it has only been offered in the SXM4 form factor. That is about to change with the introduction of a PCI Express variant. The A100 is the biggest and baddest (in a good way) version of Ampere. NVIDIA packed 54 billion transistors onto a die size measuring 826mm2. It has a total of 6,912 FP32 CUDA cores, 432 Tensor cores, and 108 streaming multiprocessors. FP32 performance checks in at a staggering 19.5 teraFLOPS, which is only half...
Read more...
Earlier this morning, Micron let it slip that NVIDIA's upcoming GeForce RTX 3090 Ampere graphics cards will ship with 12GB of GDDR6X memory. According to Micron's documentation, the GeForce RTX 3090 will be capable of achieving 1TB/sec of memory bandwidth. However, that's not all that Micron had in store via its Tech Brief, which is inexplicably still hosted on its server. Also mentioned is HBMnext, and Micron says that it is "fully involved in the ongoing JEDEC standardization" for the memory. Unfortunately, the company is rather light on details, only adding that it will "enter the market toward the end of 2022". It's not clear if HBMnext is a complete rethink of Micron's...
Read more...
One thing we know for sure is that Big Navi is coming. AMD has not kept this a secret, even going so far as to say it will be a halo product. How it will fare against Ampere remains to be seen, but make no mistake, AMD has every intention of competing at the high end of the graphics market. The question is, will Big Navi arrive with HBM2e or GDDR6 memory? Almost definitely the latter, though the rumor mill is undecided on the matter. AMD has not shied away from HBM memory in the past, which offers gobs of bandwidth. However, it is generally more expensive than GDDR memory. Combined with potential supply issues, it makes sense to choose GDDR6 for AMD's next-gen GPUs, as NVIDIA will presumably...
Read more...
Anticipation is running high for AMD's second-generation Radeon DNA (RDNA 2) graphics architecture, the one that will underpin the so called 'Big Navi' GPU. Will it dethrone NVIDIA's competing Turing lineup, or be competitive with upcoming Ampere parts? Those questions will be answered at a later date. In the meantime, a fresh leak on Twitter outlines potential specifications for a Big Navi card, and they're a doozy. Before we get to that, let's talk about the evolution of Navi. The current versions (Navi 10 and Navi 14) are what power AMD's Radeon 5000 series. They are based on RDNA 1.0 and leverage a 7-nanometer manufacturing process. The boldest iteration so far is the Radeon RX 5700 XT with...
Read more...
The name "Flashbolt" sounds like it could be a comic book superhero, or a spell you cast when running up against some ruffians in a role playing game. Neither of those things are what Samsung had in mind when launching Flashbolt, which is actually its third-generation high bandwidth memory 2E (HBM2E) designed for high performance computing (HPC) applications. Samsung's unveiling comes hot on the heels of JEDEC publishing an updated revision of the HBM2 standard called JESD235C. The updated standard pushes speeds to 3.2Gbps per pin, paving the way for 410GB/s of memory bandwidth per stack. And so it goes with Samsung's new Flashbolt solution, which adheres to those specs. "With the introduction...
Read more...
SK Hynix is not the first company in the world to develop an HBM2E DRAM product—that distinction belongs to Samsung, which announced HBM2E DRAM of its own back in May—but the company is retaining bragging rights nonetheless. That's because SK Hynix's new HBM2E DRAM sports the industry's highest bandwidth. According to SK Hynix, its new HBM2E product can top 460 gigabytes per second (GB/s), based on 3.6 gigabits per second (Gbps) performance per pin with 1,024 data I/Os. That is a 50 percent jump over HBM2. It also scales to twice the capacity, paving the way for 16GB solutions by way of stacking eight 16Gb chips in a dense package. SK Hynix sees its HBM2E DRAM being used in a variety...
Read more...
