The teaser showed what is currently being dubbed the top-end SKU from Intel’s Arc lineup for desktop. We got to see what potentially the OEM or reference design for Arc Alchemist desktop GPUs will look like, or at least what the mysterious “Limited Edition” GPU will look like. While we didn’t get any other information other than the visuals and a release window of summer 2022, it seems like Intel may have hid a little tease elsewhere in the show.
Arc Desktop GPU in disguise?
As mentioned at the start, Intel also showed off its new Arc Control software and one of the features in that software is the ability to monitor performance. This includes metrics like CPU and GPU utilization, temperatures, fan speeds, and clocks, among others. The performance monitoring tab was actually shown at the event for a brief moment, but the interesting bit is, some of the data seen in this tab does not align with any Arc mobile GPU announced that day. In particular, the GPU Power, GPU Clock, and VRAM Clock speed were the outliers that gave away the fact that we’re most likely looking at the specs of an Arc desktop GPU. And not just any Arc desktop GPU, but the highest-end flagship SKU in the lineup. The kind of numbers these metrics were reaching are next to impossible for laptop graphics, but that doesn’t completely negate the possibility of it. As you can see in the screenshots attached above, the GPU clock speed can be seen reaching 2200-2250Mhz with over 99% utilization, not only is that incredibly difficult for a laptop to achieve given the limited thermal headroom of a mobile chassis, that clock speed is currently speculated to be the boost clock of Intel’s upcoming flagship desktop GPU based on the ACM-G10 die, the Arc A780.
Further analysis
There are Geekbench scores for the Intel A350M that already exist which show the mobile GPU reaching a maximum frequency of 2200Mhz during the OpenCL test. In contrast, the base clock of the Arc A350M is only 1150Mhz which presents a massive discrepancy between the base and boost clocks. However, keep in mind that 2200Mhz is the peak clock speed. Intel’s Tom Peterson has already mentioned in an interview that the max boost clock for their GPUs is slightly lower than the peak boost clock they can achieve. In essence, the max boost clock from the factory is basically the most constrained boost clock that the GPU can reach, which means that the actual peak boost clock can go higher. This is similar to how NVIDIA dictates boost clocks for their GPUs as well. So, that kind of proves that the Arc A350M (the only GPU to launch at the event) is probably the GPU in those screenshots, right? Well, no, and the way we figure this out is by looking at the GPU power. We see that in the screenshot, the GPU seems to be running at 175W. Not only is that number in the most upper echelon of GPU TDPs, more importantly it does not match the TDP of the Arc A350M, which allows us to rule it out. Right now, only one GPU in the world can even fathom reaching that. And that GPU is the RTX 3080 Ti mobile which has a boost clock of 150W+ which still puts the 175W mark on the higher side for even that GPU. The upcoming Arc A770M laptop GPU, the flagship Arc A-Series mobile offering, is rumored to come with a 150W maximum GPU power so that proves we are not looking at another mobile SKU either. However, it’s important to keep in mind that laptops with both Intel CPUs and GPUs can take advantage of Intel Deep Link technologies, one of which is Dynamic Power Share. This feature dynamically adjusts and manages the power of the laptop to provide adequate power to the CPU and GPU by sharing the total power available between the two. That, in theory, could take the 150W maximum TGP of the Arc A770M to over 150W and match that 175W spec. Adding to this, 3DCenter has said that we are seeing only the raw GPU power when looking at that 175W TGP and that with the addition of other components like memory, the GPU could easily break the 200W barrier. On top of that, we are unsure if “GPU Power” does already include the memory or not so we can only hypothesis based on the memory values we are aware of.
(max) GPU Clock: 2250 MHz VRAM Clock: 1093 MHz = 17.5 Gbps GPU Power: 175W “GPU Power” looks like just GPU or GPU+Memory. The whole card will be draw ~210-230W. According to this, it should be DG2-512.https://t.co/nCToElfdxs pic.twitter.com/f0whNgW4at — 3DCenter.org (@3DCenter_org) April 1, 2022 Speaking of which, we can see that the “VRAM Clock” right below the GPU Clock reads 1093Mhz which indicates that the effective speed of the memory could reach up to 17.5Gbps. That would make it the fastest memory on a mobile GPU, that is, if it really is a mobile GPU. The fastest mobile GPU in existence can only reach 16Gbps and since memory is seldom overclocked on mobile GPUs, we can evaluate that the desktop version of this mobile GPU will at least match or be faster than the RTX 3070, which features a 14Gbps memory speed.
Time will tell
All of this could be an accident where Intel mistakenly revealed the specs of a desktop GPU in their showing for mobile GPUs, or it could all be entirely deliberate and Intel did this to spark interest and to get people talking. Even us speculating the possibility of this being either a mobile or desktop GPU is giving Intel the press and attention they need to get more eyes on their product. Whether we’re looking at the leaked Intel Arc A780 desktop graphics card or the Arc A770M flagship mobile GPU from Arc A-Series, it is safe to say that Intel is stirring up the competition successfully. With the Arc A350M laptops already here and Arc 5 & 7 to grace us soon, the Blue Team is slowly starting to materialize its discrete GPU lineup after so many years of conceptualizing. The real test for Intel will be their desktop GPUs which seem to have been hyped up to oblivion. Intel is looking to not only disrupt the market as the third major player in the GPU space, but it’s also looking to do so fiercely, by offering highly competitive rates. If their strategy will be similar to how AMD debuted Ryzen—competitive performance at a significantly lower price—then we are in for a treat.
