Tag:

AMDs

AMD’s Instinct MI355X accelerator will reportedly consume 1,400 watts

by admin June 12, 2025

Mark Papermaster, chief technology officer of AMD, formally introduced the company’s Instinct MI355X accelerators for AI and HPC at ISC 2025 — revealing massive performance improvements for AI inference, but also pointing to nearly doubled power consumption of the new flagship GPU compared to its predecessor from 2023, reports ComputerBase.

AMD’s CDNA 4 enters the scene

AMD’s Instinct MI350X-series GPUs are based on the CDNA 4 architecture that introduces support for FP4 and FP6 precision formats alongside FP8 and FP16. These lower-precision formats have grown in relevance in AI workloads, particularly for inference. AMD positions its Instinct MI350X processors primarily for inference, which makes sense as scale out world size of MI350X continues to be limited to eight GPUs, which reduces their competitive capabilities compared to Nvidia’s Blackwell GPUs. Still Pegatron is readying a 128-way MI350X machine.

AMD’s Instinct MI350X family of AI and HPC GPUs consists of two models: the default Instinct MI350X module with a 1000W power consumption designed for air cooling as well as the higher-performance Instinct MI355X that will consume up to 1400W and will be designed primarily for direct liquid cooling (even though AMD believes that some of its clients will be able to use air cooling with the MI355X).

Faster GPUs incoming

Papermaster expressed confidence that the industry will continue to develop even more powerful CPUs and accelerators for supercomputers to achieve zettascale performance in about a decade from now. However, that performance will come at the cost of a steep increase of power consumption, which is why a supercomputer offering a ZetaFLOPS performance could consume 500 MW of power — half of what a nuclear power plant can produce.

At ISC 2025, AMD presented data showing that top supercomputers have consistently followed a trajectory where compute performance doubles roughly every 1.2 years. The graph covered performance from 1990 to the present, demonstrating peak system GFLOPs. Early growth was driven by CPU-only systems, but from around 2005, a shift to heterogeneous architectures — mixing CPUs with GPUs and accelerators — took over. Now, in what AMD calls ‘AI Acceleration Era,’ systems like El Capitan and Frontier are pushing beyond 1 ExaFLOP, continuing the exponential growth trend with increasingly AI-specialized hardware.

But performance comes at a cost of power consumption. To maintain performance growth, memory bandwidth and power scaling have become urgent challenges. AMD’s slide indicated that GPU memory bandwidth must more than double every two years to preserve the ratio of bandwidth per FLOPS. This has required increasing the number of HBM stacks per GPU, which in turn results in larger and more power-hungry GPUs and modules.

Indeed, power consumption of accelerators for supercomputers is increasing rapidly. While AMD’s Instinct MI300X introduced in mid-2023 consumed 750W peak, the Instinct MI355X, set to be formally unveiled this week, will feature a peak power consumption of 1,400W. Papermaster envisions 1,600W accelerators in 2026 – 2027 and then 2,000W processors later this decade. By contrast, AMD’s peers from Nvidia seem to be even more ambitious when it comes to power consumption as their Rubin Ultra GPUs featuring four reticle-sized compute chiplets are expected to consume as much as 3,600W.

The good news is that in addition to increased power consumption, supercomputers and accelerators have also been gaining performance efficiency rapidly. Another one of AMD’s ISC 2025 keynote slides illustrated that performance efficiency increased from about 3.2 GFLOPS/W in 2010 to approximately 52 GFLOPS/W by the time exascale systems like Frontier arrived.

Looking ahead, maintaining this pace of performance scaling will require doubling energy efficiency every 2.2 years. A projected zettascale system delivering 1,000× exaflop-class performance would need around 500 MW of power at an efficiency level of 2,140 GFLOPs/W (a 41-fold increase from today). Without such gains, future supercomputers could demand gigawatt-scale energy — comparable to an entire nuclear power plant, making them way too expensive to operate.

AMD believes that to increase the performance of supercomputers dramatically a decade from now, not only it will need to make a number of architectural breakthroughs, but the industry will have to keep pace with compute capabilities to provide adequate memory bandwidth. Still, using nuclear reactors to power supercomputers seems in the 2030s seems to be a more and more realistic possibility.

Follow Tom’s Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.

Source link

June 12, 2025 0 comments

The AMD Radeon RX 9060 XT graphics card on a red gradient background

Product Reviews

A new report indicates Intel’s latest Battlemage GPUs are a total failure and AMD’s gaming graphics market share fell to just 8% but overall graphics cards sales are up

by admin June 6, 2025

AMD’s share of the add-in graphics card market fell to a fairly pitiful 8% in the first quarter of 2025 according to Jon Peddie Research (JPR), a specialist in tracking PC hardware sales. Meanwhile, JPR puts Intel’s graphics market share at essentially zero, with Nvidia hoovering up the remaining 92%. Ouch.

Before we all escalate the threat level to PC gaming to DEFCON 1 and panic, it should be noted that JPR’s figures cover the period up to the end of March. AMD’s new Radeon RX 9070 and 9070 XT were only released on March 6. So, that’s only three and a bit weeks of RDNA 4 sales in the figures.

In other words, we’ll have to wait for the Q2 numbers to get a full picture of how much impact AMD’s new GPUs have had. Indeed, we’ll arguably have to wait even longer than that, since it’s the RX 9060 XT, which only went on sale yesterday, that’s meant to be the real volume play for AMD and that won’t have a substantial impact on market share figures until the Q3 data is out.

Overall, GPU sales were up 8.5% in Q1 2025 compared to Q4 2024. JPR says that Q1 is usually flat or down compared to the previous quarter, but that, “In Q1’25, AMD and Nvidia introduced new AIBs, which stimulated the market.”

If we’ll have to wait a little longer to see how the RX 9070 and 9060 GPUs fare for AMD, the verdict on Intel’s Arc B570 and B580 GPUs is surely in and it not good news, not good news at all.

The blip that was 1% Intel market share has vanished from the latest GPU sales figures. (Image credit: JPR)

JPR reckons Intel’s market share fell by 1.2% in the latest quarter, essentially putting it on 0% and not appearing in the data for Q1 2025. If there was already doubt over Intel’s commitment to the add-in graphics card market, this apparent total failure of Intel’s second-gen Battlemage generation of GPUs is extremely troubling.

Battlemage GPUs went on sale at the end of 2024. So, in theory Q1 2025 was a full quarter for sales of Intel’s latest graphics hardware. Obviously sales of the B570 and B580 haven’t actually been zero. But if JPR is to be believed, the volumes are so small as to effectively push Intel out of its data.

If even half a percent of sales were going to Intel, JPR would no doubt have included that in its data. So, we’re talking less than one in 200 GPUs sales being Intel cards, and quite possibly an order of magnitude or more worse than that.

It really is a grim outlook for Intel graphics and puts rumours of more powerful Battlemage GPUs into context. With sales this low, is it remotely realistic to expect Intel to wheel out that purported Nvidia RTX 4070 / 5070 killer? It doesn’t seem terribly plausible in the light of this latest data.

Still, the numbers we’re really looking forward to will be Q3 2025, when both of AMD’s new RDNA 4 GPUs will have had at least a full quarter on sale. That data probably won’t be out until the end of the year. So, check back in December and cross all your fingers and toes until then.

Source link

June 6, 2025 0 comments

Product Reviews

The best CPU choice for high end gaming, AMD’s 9800XD, drops below MSRP

by admin May 27, 2025

It’s only been around for the last 6 months, but the Ryzen 7 9800X3D has cemented itself as one of the best gaming CPUs on the market. With its hefty stacked 3D V-Cache, it has been the in-demand CPU for gaming since its launch and has previously been very hard to obtain. Now, with more available stock, we’ve seen the processor not only become readily available but also undergo its first price cut below the original MSRP.

Available at Amazon, you can now pick up the AMD Ryzen 7 9800X3D processor for $472. It’s not an absolutely massive saving, but it’s cheaper, and hopefully a sign that computer components may weather the storm in terms of pricing due to tariffs and other economic influences.

You can find more details of the performance of the Ryzen 7 processor in our detailed review of the 9800X3D. With an 8% performance gain on the previous generation 7800X3D, and easily outperforms the competition from Intel in gaming benchmarks. The 9800X3D is unlocked for overclocking and has minimal cooling requirements, making it the best choice for a high-end gaming rig.

The 9800X3D features eight cores and 16 threads, with a 120W TDP and is unlocked for overclocking. There is even integrated Radeon graphics, but you would of course want to pair this processor with a powerful graphics card for the best gaming experience.

Follow Tom’s Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.

Source link

May 27, 2025 0 comments

Game Updates

AMD’s New RX 9060 XT Graphics Card Will Launch In 8GB And 16GB VRAM Versions

by admin May 24, 2025

AMD announced its latest graphics card, the Radeon RX 9060 X, at Computex 2025 earlier this week. This is the third GPU in AMD’s 9000 series lineup that also includes the RX 9070 and 9070 XT, and its slated to launch June 5. While it’s positioned as the entry-level option in the RX 9000 series, the RX 9060 XT runs on the same RNDA 4 architecture as the more powerful RX 9070 and 9070 XT and includes FSR 4 resolution upscaling and frame generation, as well as DisplayPort 2.1a and HDMI 2.1b support.

AMD is producing two versions of the RX 9060 XT: one with 8GB VRAM and another with 16GB VRAM. Both models will feature 32 RDNA 4 CUs (computation units), up to 3.13GHz clock speed, and between 150W and 182W power draw, depending on the model’s VRAM and clock speed. While those are solid specs for an entry-level card, things get a bit more confusing from there.

Power and pricing

AMD lists the MSRP as $300 for the 8GB model and $350 for the 16GB model. However, the company is not releasing RX 9060 XT reference cards itself. Instead, third-party manufacturers like Asus, Acer, Gigabyte, and others are launching their own versions of the RX 9060 XT starting June 5. As such, performance, power draw, and pricing will likely differ between the various RX 9060 XT models (especially for overclocked models), but at least some of these GPUs will stick to the reference clock speeds and therefore MSRP (assuming the ongoing tariffs do not impact those prices).

Asus is launching three RX 9060 XT models

As for performance, the RX 9060 XT 16GB is positioned as the replacement for AMD’s RX 7600 XT as its flagship entry-level GPU, making it the primary competitor to Nvidia’s $300 GeForce RTX 5060 and RTX 5060 Ti (which costs either $380 for the 8GB version or $429 for the 16GB version). According to AMD’s Computex 2025 presentation, the company’s internal testing found the RX 9060 XT is roughly six percent faster than the RTX 5060 Ti on 40 games running at 1440p–and at roughly $80 MSRP cheaper. However, those tests compare the RX 9060 XT 16GB to the GeForce RTX 5060 Ti 8GB, rather than the RTX 5060 16GB. The RX 9060 8GB model’s performance is a bigger question, as the smaller amount of VRAM could lead to significant performance dips.

On X (formerly Twitter), AMD’s Chief Architect of Gaming Solutions and Gaming Marketing, Frank Azor, explained that most players use a 1080p display, for which 8GB of VRAM is more than enough. While technically true, it doesn’t alleviate the concerns over confusing consumers who may see the above comparisons of the RX 9060 XT 16GB’s 6% performance boost over the RTX 5060 Ti 8GB and assume that applies to all RX 9060 XT models. And again, the actual performance will likely differ between the various SKUs launching June 5, which could further compound the issue.

Majority of gamers are still playing at 1080p and have no use for more than 8GB of memory. Most played games WW are mostly esports games. We wouldn’t build it if there wasn’t a market for it. If 8GB isn’t right for you then there’s 16GB. Same GPU, no compromise, just memory…

— Frank Azor (@AzorFrank) May 22, 2025

Where to find the AMD RX 9060 XT

Those potential concerns and unanswered questions aside, the RX 9060 XT nevertheless looks like a solid entry-level choice for those already invested in AMD’s hardware. As mentioned above, multiple manufacturers are releasing RX 9060 XT GPUs starting June 5, and while none are currently listed for preorder, you can learn more about the models announced from Asus, Acer, and Gigabyte, at each company’s online store. Other brands like PowerColor also showed off upcoming RX 9060 XT GPUs on the Computex 2025 show floor. We’ll be sure to update this post once the GPUs are available.

Source link

May 24, 2025 0 comments

Gaming Gear

AMD’s Radeon RX 9060 XT Could Do Budget GPUs Better Than Nvidia

by admin May 21, 2025

In the battle of the low-end, 60-class graphics cards, AMD wants to see if it can pull off the same sucker punch of price and performance it gave Nvidia during the launch of its mid-range GPUs. The graphics card maker offered the first, sparse details on its Radeon RX 9060 XT graphics processors late Tuesday at Computex. The card may offer enough power for your PC to hit solid gaming performance at 1440p resolution, similar to the $450 Nvidia GeForce RTX 5060 Ti, on cheaper gaming rigs. The real inflection point of this latest card will be whether you can actually buy it for its base price.

The Radeon RX 9060 XT is the step down in GPU performance from the RX 9070 that AMD launched back in March. It’s based on the same RDNA 4 microarchitecture of the mid-range cards, but with 32 of the company’s latest compute units compared to the 56 on the higher-end card. The GPU comes with two options: one with 8 GB and another with 16 GB of GDDR6 VRAM. The version with more memory will be better for your rig long-term, especially if you plan to hook your PC up to a 1440p monitor and run the latest, more graphically intensive games.

AMD did not offer us the full range of specs, which makes it hard to pin down just where this GPU will land in terms of raw performance compared to Nvidia’s latest cards. While the number of RDNA 4 compute units—the core clusters on AMD cards that process the thousands of calculations necessary for graphically intensive tasks—offers a vague impression of performance compared to the RX 9070, AMD didn’t provide any charts to compare FPS between games. The GPU runs on a 3.13GHz boost clock and has between 150W and 182W of board power compared to the 2.54 GHz clock and 304W board power on the company’s Radeon RX 9070 XT.

Without a price tag, it’s impossible to judge how much of a step down the latest card is compared to the RX 9070. AMD didn’t offer any word on a non-XT variant, either. The card will require a PCIe 5.0 x16 interface, the same as its other cards. AMD doesn’t craft its own GPUs and instead relies on AIC (add-in card) makers to produce its cards. We’ll update this article if AMD announces details on price or availability during its Computex keynote.

The crown jewel of AMD’s current lineup of graphics cards is the RX 9070 XT. AMD made headlines when it set the suggested sale price of the GPU at $600, only $50 more than the 9070, but it packs enough performance to get playable framerates out of multiple intensive games at 4K with a fair amount of ray tracing settings turned up. Unfortunately, because of a combination of tariffs and stock woes, the 9070 XT ended up priced at over $800 and as high as $1,000 at some online retailers.

We’ve seen prices fluctuate regularly over the past several months, but a near 20% price inflation to what should be a mid-range card is simply too much to stomach. However, the lower-end GPUs are faring better. The RTX 5060 Ti MSRP is set at $450, and the lowest price we’ve seen so far is $480. The $300 RTX 5060 is sitting closer to $320 from some AIC makers like Gigabyte. A fair number of Nvidia’s lowest-end GPUs are currently listed as “Out of Stock” or “Coming Soon” on sites like Newegg and Best Buy. Those buying a lower-end GPU are more price sensitive than people who can drop $2,000 on an RTX 5090 without blinking. AMD has even more impetus to set a price people can afford, and make sure it can keep costs level when the card finally hits store shelves.

Source link

May 21, 2025 0 comments