A new test found in the SiSoft Sandra benchmark database indicates that the high-end GPU from the Intel Arc Alchemist family, with 4,096 cores, should deliver slightly higher performance than the GeForce RTX 3070 Ti, the green team’s advanced mid-range solution. The result reinforces old rumors, which already indicated that Arc cards would not compete with the best AMD and Nvidia have, but would be competitive enough to attract enthusiasts.
Intel Arc Alchemist beats RTX 3070 Ti in leaked test
The result was found by the renowned leaker TUM_APISAK and refers to the most powerful GPU of the Arc family, equipped with DG2-512 chip. The solution features 512 Vector Engines (or Execution Units) for a total of 4,096 cores, clocks of 2.1 GHz and 16 GB of GDDR6 VRAM on a 256-bit interface, thus offering enough performance to surpass the RTX 3070 Ti. , according to the listing.
In general computing tests, Intel Arc reached 9017.52 Mpix/s, a result 7.7% better than the one presented by Nvidia’s competitor, which scores 8369.51 Mpix/s. Looking at more specific numbers reveals that the two GPUs are competitive and exhibit strengths and weaknesses, yet the comparison favors the GeForce family solution.
Want to stay on top of the best tech news of the day? Access and subscribe to our new youtube channel, Kenyannews News. Every day a summary of the main news from the tech world for you!
The highlight goes to the calculation of algorithms in Single-float point (FP32), the most common type in games, a situation in which Arc Alchemist is surpassed by almost 30%. On the other hand, in the Double-float (FP64) and Quad-float point (FP128) algorithms, intended for high-performance computing, the Intel board manages to open up an astronomical advantage of up to 400%.
This may mean that Intel’s GPU should still be inferior to the RTX 3070 Ti in games, but be an interesting option for professionals who rely on FP64 and FP128 calculations, which shows the company’s focus on this segment and points to a possible reason for the cancellation. of the Xe-HP architecture, aimed at this audience. It is possible that the strong advantage in more complex algorithms can somehow be converted to games, but it is still too early to say.
Intel gaming GPUs to debut soon
Based on the Xe-HPG microarchitecture, the Intel Arc Alchemist family reimagines the brand’s GPUs with new structures designed to allow for good scalability, with the cores now organized in Vector Engines, formerly known as Execution Units (EUs). The junction of 16 Vector Engines, allied to the company’s AI cores, form the so-called Xe-Core, the basic unit scaled by Intel to form the different GPUs of the family.
The high-end solution features 32 Xe-Cores, with 512 Vector Engines and 4,096 cores, but there will be other configurations, including variants with 16 and 8 Xe-Cores, in models for desktops and notebooks. During the family reveal, Intel had announced the intention to launch the GPUs in the first quarter, but it is possible that plans have changed – the official website of the cards changed the forecast to only “2022”.
