The new generation RDNA 4 architecture introduced this year has produced only a handful of Radeon GPUs so far. Another (only the sixth to date) desktop gaming graphics card is coming now. While it is based on existing die, it stands out with reduced power consumption and thus higher energy efficiency, which could be appealing for quiet and compact Mini-ITX systems (or when your power bill shoots up due to an AI data center close by).
Without any prior leaks or reports of preparations, AMD has quietly released a new graphics card model—the Radeon RX 9060 XT LP. It is largely based on the configuration of the Radeon RX 9060 XT, which is nothing new, but the difference lies in the reduced power consumption—and that is rare these days. The “LP” stands for “Low Power,” according to the description or motto of the card on AMD’s website (“Immersive Experiences at Lower Power”).
The Radeon RX 9060 XT LP uses the full configuration of the Navi 44 chip with 2048 shaders, 32 Ray Accelerators, and 64 AI Accelerators (32 CUs of the RDNA 4 architecture). AMD already lists specifications on its website, but they are not complete, so exact clock speeds are unknown. However, it is stated that the card delivers a raw compute performance of 25 TFLOPS in FP32 compute (and 400 TFLOPS / 800 TOPS on the AI units when using FP8 and INT4).
By comparison, the standard Radeon RX 9060 XT (with clocks of 2530 MHz as the “Game Clock” and 3130 MHz as “Boost”) is listed with 25.6 TFLOPS in FP32 and 410 / 821 TFLOPS / TOPS in AI. These theoretical compute performance figures are always derived from the official boost clock, so a simple calculation shows that the low-power Radeon RX 9060 XT LP has a reference boost clock of exactly 3050 MHz. We do not know the game clock, but real-world performance will in any case depend on real-world sustained GPU clock. It is possible that the difference versus the standard RX 9060 XT will be larger than what the mere 2.5% difference in boost clocks would suggest.
Power consumption is reduced a bit more noticeably—under reference parameters, the RX 9060 XT LP is said to have a total board power of 140 W, whereas the standard XT version released in the summer has a base TDP of 160 W (meaning the new model should consume 12.5% less power). The card has an official recommended PSU wattage of 450 W (which assumes a relatively low-power CPU), but the same value is already specified for the standard RX 9060 XT with its reference 160 W power draw.
In practice, however, most cards are sold factory-overclocked, and such models also come with increased power consumption (and consequently higher PSU requirements). In direct comparisons, the Radeon RX 9060 XT LP could therefore turn out significantly more efficient than overclocked RX 9060 XT models—assuming manufacturers stick to the recommended 140 W limit for the LP card.
Fortunately, the card retains 16 GB of GDDR6 memory with a bandwidth of 320 GB/s (effective clock 20 GHz), so there has been no cutback due to the current mess with memory shortages and high prices caused by the AI craze. As with the XT model, memory bandwidth will be helped by a 32 MB Infinity Cache.
Otherwise, the “LP” card should support all technologies associated with the RDNA 4 architecture—for example support for DisplayPort 2.1a and HDMI 2.1b for monitors, AV1 video encoding and decoding, and FSR 4 “Redstone” technology in games. Aside from its operating characteristics and slightly lower performance, it should therefore be fully equivalent to the more powerful but less efficient Radeon RX 9060 XT 16 GB.
- Tip: Better, more capable than expected: RDNA 4 architecture deep dive
- Tip: AMD FSR Redstone is official: using AI for higher image quality
- Tip: FSR “Redstone” Frame Generation: How’s It Running for You?
Price and availability? Unknown
The launch of the card happened quietly, without any announcement to speak of. The company has not disclosed when the card will go on sale or what its MSRP will be. It is possible that it will be an OEM model that you would find in prebuilt systems (although an 8 GB card would sound more likely at the moment due to rising memory prices). However, AMD has published the specifications of the model on its international website in English, so this time it should not be a card intended specifically for the Chinese market.
Sources: AMD, VideoCardz
English translation and edit by Jozef Dudáš
⠀
