GREAT JOB RAJA! – Intel ARC Alchemist DG2-512 Spotted on Geekbench OpenCL Benchmark – EMBARASSING Performance
20 February 2022
ARCTIC SOUND Benchmarks DG2 GEEKBENCH 5 Gen 12 Intel Intel ARC Alchemist OPENCL Raja Koduri

It would seem as if Raja Koduri is cooking up another excellent curry of failure judging from leaked OpenCL performance figures on Geekbench with an engineering sample of Intel’s long awaited “DG2” graphics card.

Raja Koduri a very talented individual when it comes to organizing the most outrageous “launch parties”, when it comes to architectural design however is where Raja unfortunately flounders as he recites the same mantra since getting back to basics after leaving the Apple corporation.

Overhype, delay, overhype, underdeliver.

Such was the unfortunate fate for AMD’s “VEGA” architecture which despite a hefty initial delay still launched as a mere “Beta” test of a product with plethora of architectural features never coming to fruition.

And hilariously enough this has always been the case for Intel’s second or third attempt, depending on what you would define as a discrete graphics card for the mega-kike corporation to actually enter the market with their “ARC Alchemist” or Intel Xe “DG2” if you’d prefer.

From humble beginnings or more rather from unhumble beginnings considering this is Raja Koduri we’re talking about here, my contact at the Intel corporation did inform me that performance targets for Intel’s “first” discrete high-performance graphics card were in actual fact to be compared against the likes of the NVIDIA GeForce RTX 2070 SUPER and the AMD Radeon RX 5700 XT.

Needless to say a few short years ago as I was informed of such information that expectations for Intel’s “DG2” or ARC Alchemist were somewhat modest at least at the time, but by today’s standards such performance would be beyond laughable, signs of a DOA product.

And yet somehow, after two long years and numerous processing node advancements from not only Intel themselves but also the Taiwanese mafia known as TSMC performance characteristics for Intel’s DG2 apparently hasn’t changed all that much seen from a leaked Geekbench 5 submission utilizing Intel’s ARC Alchemist graphics card equipped with 512 Execution Units to boot, the full chip.

Now I’ve said it before and I’ll continue to say that Geekbench is entirely irrelevant, its mobile trash benchmarks are merely used by fangays alike from Apple to Intel CPU’s so such performance scores aren’t entirely indicative of actual performance but as if I’d ever pass up the opportunity to ridicule a company as heinous as (((Intel))) as tech urinalist publications either obscure or ignore the truth.

Intel’s DG2-512 graphics card is seen here on Geekbench 5 paired with a rather unusual Intel Core i5-9600K with the GPU in question reporting a maximum frequency of 2.4GHz which isn’t necessarily bad, but it’s the scores we’re interested in of which the soon to be ARC Alchemist flagship has managed to produce an OpenCL score of 85,448 points.

85,448 points by comparison to products from other manufacturers is an embarrassment, but as we’ve already seen from Intel’s “Iris Xe” DG1 piss poor performance is par for the course.

GPUOverall Scorevs. DG2-512
NVIDIA GeForce RTX 3080 Ti198808232.66%
NVIDIA GeForce RTX 3080180826211.62%
AMD Radeon RX 6800 XT156479183.12%
NVIDIA GeForce RTX 3070 Ti146191171.08%
NVIDIA GeForce RTX 2080 Ti138945162.60%
NVIDIA GeForce RTX 3070132175154.68%
NVIDIA GeForce RTX 3060 Ti117116137.06%
NVIDIA GeForce RTX 2080104833122.68%
AMD Radeon RX 6700 XT101380118.64%
NVIDIA GeForce RTX 2070 SUPER97283113.85%
NVIDIA GeForce RTX 306095145111.34%
NVIDIA GeForce RTX 207085689100.28%
Intel ARC Alchemist DG2-512 ES85448NaN%
NVIDIA GeForce RTX 2060 SUPER8511799.61%
AMD Radeon RX 6600 XT8153795.42%
AMD Radeon RX 5700 XT8107494.88%

When it comes to Geekbench 5’s overall score for OpenCL, NVIDIA GeForce graphics cards tend to simply dominate and that’s not really an exception for today whether or not Intel actually prioritize OpenCL as the aforementioned DG2-512 produces similar score figures as the NVIDIA GeForce RTX 2070 and 2060 Super.

To say this is underwhelming would be an understatement, with 16GB of GDDR6 memory the DG2-512 sample is merely obliterated by mid-range offerings from previous generation components.

When compared with its original intended target, the NVIDIA GeForce RTX 2070 Super, the flagship DG2 falls short by over 12% overall, being beaten out by the likes of the new “entry level” NVIDIA GeForce RTX 3060 graphics card which comes with 12GB of GDDR6 memory.

Radeon always falls short when it comes to OpenCL but regardless of that even the likes of the Radeon RX 6700 XT is simply too much for the Intel DG2 to bare as it exceeds both the RTX 3060 and 2070 Super in terms of its overall figures.

Less you start to compare with the sort of graphics cards I’d actually expect the likes of the 512 EU DG2 to compete against, such as the NVIDIA GeForce RTX 3060 Ti or if Raja could actually get his act together maybe even potentially the RTX 3070 which simply walk away by comparison to the embarrassing figures set by the ARC engineering sample as the DG2-512 trails behind them both by 27% and 35.35% respectively.

There is a silver lining for Intel however curtesy of Twitter user @326powah who compared individual test results between the DG2-512 sample versus scores taken from the NVIDIA GeForce RTX 3070 and AMD Radeon RX 6700 XT where the Intel GPU is convincingly beaten across the entire board but on a couple instances does it actually attempt to put up a fight which of course would skew overall results in favor of anything but Intel.

Geekbench is probably the furthest thing imaginable from actual “gaming” performance, it’s still a garbage mobile oriented benchmark at the best of times, but considering how performance expectations for Intel’s “first” GPU sort have always been a bit bleak, they’ve had a couple more years since then to get their act together or at the very least harness the advantages that have been made with fabrication processing nodes since then.

While we still wait for actual performance figures that truly matter I still reserve serious doubts over the performance capabilities of ARC Alchemist considering Intel would have had to manage serious leaps in performance from initial targets just to match the current capabilities from the likes of current generation mid-range components, and this is supposedly Intel’s flagship SKU we’re talking about here.

Just imagine if Intel weren’t able to come up with a 40% increase over initial expectations years later and the end result falls below the likes of the RTX 3060 Ti by a considerable margin, Intel’s flagship would be giving consumers performance that easily could’ve been had from 2018, and yet to this very day they’re oh so seclusive when it comes to showcasing the product whatever despite its launch window drawing closer.

I guess that’s probably why Intel are hyping up their own regressive upscaling solution, XeSS because if the performance of their ARC Alchemist GPUs are to follow by above’s example you’ll be needing to utilize XeSS from such a low resolution to actually make much use out of your Graphics Card with such lackluster performance.

For Intel’s flagship DG2 to fall short from the likes of the NVIDIA GeForce RTX 3060 Ti in everything barring the price tag would have devastating effects on the outlook of the DIY PC market as it’s certainly getting creamed by NVIDIA and AMD without much effort currently as Intel’s ARC will undoubtedly be another VEGA.

blog comments powered by Disqus