AMD, Radeon RX 6600 XT and RX 6600 specs leaked

AMD, Radeon RX 6600 XT and RX 6600 specs leaked


A Chiphell forum user shared two screenshots from popular GPU-Z software that reportedly reveal the specs of AMD's much-rumored Radeon RX 6600 XT and Radeon RX 6600 graphics cards. GPU-Z recently added support for both unannounced products based on the RDNA 2 architecture, so the information should be accurate. The Radeon RX 6600 XT and RX 6600 will be the first cards based on the Navi 23 GPU and, according to the images, both should be limited to the PCIe 4.0 x8 interface, as has already happened previously with the Radeon RX 5500 XT and RX 5500.

Credit: Chiphell The Radeon RX 6600 XT will be equipped with 32 Compute Units (CU), equivalent to 2,048 Stream Processors (SP), as well as 32 Ray Accelerators for ray tracing. Furthermore, there should probably be 32MB of Infinity Cache on board. In terms of operating frequencies, the Radeon RX 6600 XT is expected to have a base clock of 1,692MHz and a boost clock of 2,684MHz. The subsystem vendor ID points to AMD, so this particular Radeon RX 6600 XT should have been made based on the design reference, but it could also be an engineering sample.

The Radeon RX 6600, on the other hand, may be limited to just 28 CU or 1,792 SP, which are accompanied by 28 Ray Accelerators. Unfortunately, GPU-Z was unable to detect clock speeds. Apparently, the Radeon RX 6600 XT and the Radeon RX 6600 share one thing in common: memories. The pair of graphics cards appear to be equipped with 8GB of GDDR6 at 16 Gb / s with an overall bandwidth of 256 GB / s, given the use of a 128-bit bus.

Credit: Chiphell Consumers won't have to worry about a possible raid by cryptocurrency miners, as the products don't perform well in Ethereum. In fact, according to the Chiphell user, the graphics cards provide, respectively, a hash rate of just 30 MH / s and 27MH / s. As for performance, the leaker said the Radeon RX 6600 XT and Radeon RX 6600 scored 9,439 and 7,085 points in the Time Spy benchmark, respectively, recording a 33.2% gap between the two. Sure, they won't be the best graphics cards on the market, but they should be more than enough for an entry level setup that aims for 1080p gaming.

Looking for a new PSU to power your next GPU? Corsair RM750X, 750W modular power supply, is available on Amazon.

AMD's Risk/Reward Looks Better Than It Has in a While

Markets have finally sold off to a point where a slew of my favorite growth tech companies -- though by no means all of them -- are sporting intriguing valuations. Advanced Micro Devices  (AMD) is a good case in point.

At Wednesday's closing price of $74.64, AMD's stock was down 19% on the year and at levels first hit last July, when the shares took off thanks to both a strong second-quarter report and news of Intel's (INTC)  7nm manufacturing process setback. That leaves the stock trading for 28 times a 2022 EPS consensus of $2.68 that I think could prove low, given that it's underpinned by expectations AMD, which guided last month for ~50% 2021 revenue growth, will only see 15% sales growth next year.

2021 and 2022 Growth Drivers

Between Intel's 7nm setback, the competitiveness of AMD's current PC and server CPU lines and the expected 2022 arrival next-gen CPU lines relying on AMD's Zen 4 microarchitecture and TSMC's (TSM) 5nm process node, AMD looks well positioned to add to its CPU share gains over the next 18-24 months. And along the way, the company's CPU sales should also benefit from rising cloud capex and a notebook market that's stronger than it has been in many years.

In addition, AMD -- much like Nvidia (NVDA) -- continues benefiting from a tsunami of demand for both game consoles and gaming GPUs. And though many investors are already betting on AMD to keep taking CPU share in 2021 and 2022, I think markets generally haven't factored in the possibility that AMD could take meaningful share next year in a high-end gaming GPU market that has been long dominated by Nvidia, thanks to its next-gen RDNA 3 architecture.

In an interview last fall, AMD exec Rick Bergman indicated that AMD is aiming for a performance-per-watt improvement for RDNA 3 that's comparable to the roughly 50% improvement provided by RDNA 2 gaming GPUs, which began shipping in November. Meanwhile, recent rumors suggest high-end RDNA 3 GPUs will be far more powerful than AMD's most powerful RDNA 2 GPU, the Radeon RX 6900 XT, thanks in part to the adoption of a modular, chiplet-based architecture.

Nvidia, to be sure, is also working on a next-gen gaming GPU architecture (it's codenamed Hopper) that's expected to deliver large performance gains. And for now at least, real-time ray tracing performance remains a strength for Nvidia's GPUs, as does their support for an AI-based image upscaling technology known as DLSS (AMD's counter to DLSS, known as FidelityFX Super Resolution, is expected to arrive soon).

But given how dominant Nvidia has been over the last decade in the high-end gaming GPU market, it's not hard to imagine RDNA 3's performance and power efficiency improvements helping turn 2022 into a stronger year for AMD's gaming GPU business than what markets are currently expecting.

What's Been Weighing on AMD's Stock

Admittedly, there are a few things that appear to have been weighing on AMD's shares besides broader selling pressure for chip stocks. The biggest of these:

  • Worries that ARM-architecture server CPUs from the likes of (AMZN) , Nvidia, private Ampere Computing and (reportedly) Microsoft (MSFT) will limit AMD's server CPU sales growth.
  • Concerns that Intel, which announced in March that it will rely on both TSMC (TSM) and internal manufacturing to launch advanced PC and server CPUs in 2023, will be more competitive going forward under new CEO Pat Gelsinger.
  • Worries that notebook and gaming hardware demand will cool off amid reopenings.
  • Concerns that AMD's pending, all-stock acquisition of top FPGA developer Xilinx (XLNX) makes its growth story a little murkier, given the steep price being paid for the company.
  • Each of these concerns is understandable. But collectively, I think they're a little overblown.

    With AMD having only achieved a double-digit server CPU share in mid-2020, there's still plenty of room for it to gain server CPU share even if (as I expect) ARM CPUs gain more traction. In addition, the software ecosystem for Intel and AMD's x86-architecture CPUs is still much more robust, which is a major reason why ARM server CPUs haven't yet seen massive adoption by traditional enterprises.

    I do think Intel, aided by both TSMC and its advanced packaging technologies, will start to be more competitive by 2023. But it's worth keeping in mind that AMD is one of TSMC's most strategically-important clients, has been executing quite well in terms of CPU development (and now has more R&D resources to pour into it) and was recently rumored to be planning to use TSMC's next-gen 3nm node for its Zen 5 CPUs, which are expected to arrive in 2023. Tougher competition from Intel doesn't mean that the clock is going to get turned back a decade.

    Notebook and gaming hardware demand will inevitably cool some from their current, white-hot levels. But the fact that there's currently a lot of unmet demand will help ease the pace of the slowdown -- Sony (SNE)  recently forecast the AMD-powered PlayStation 5 would remain supply-constrained through year's end. Also, in the case of the notebook market, AMD still has a lot of room to gain additional share from Intel and -- given its focus on premium notebook segments -- probably will be affected less than Intel by a slowdown in low-end consumer notebook sales, which have soared thanks to demand from remote workers and students.

    As for the Xilinx deal, while the price was a little high -- based on AMD's current stock price, the deal price is equal to 38 times Xilinx's EPS consensus for its March 2022 fiscal year -- it becomes a little more palatable when one takes into account both the $300 million worth of cost synergies that AMD expects from the deal and the various product synergies that should result from pairing AMD's CPUs with Xilinx's FPGAs and packaging technologies.

    Notably, Xilinx's stock is trading at a 9% discount to the implied AMD deal price, thanks to concerns about Chinese regulatory approval. But between recent Chinese approvals of chip deals, the lack of meaningful product overlap between AMD and Xilinx, and AMD's relationships with numerous Chinese OEMs and cloud giants, the odds of a deal ultimately being approved by its year-end target still look good, barring any significant new geopolitical pressures. As a result, those looking to obtain AMD shares at an additional discount might want to consider Xilinx's stock, provided they're willing to accept a measure of antitrust risk.

    And for those who would rather not deal with such a risk, the risk/reward for AMD's own stock looks more intriguing than it has in a while, even if the shares haven't quite reached bargain-basement levels. If the stock drops below $70, then I think the risk/reward really gets interesting.

    (AMD, NVDA, AMZN and MSFT are holdings in Jim Cramer's Action Alerts PLUS member club. Want to be alerted before Jim Cramer buys or sells these stocks? Learn more now.)

    Get an email alert each time I write an article for Real Money. Click the '+Follow' next to my byline to this article.

    Powered by Blogger.