Mobility NVIDIA GeForce RTX 3070 GPU Geekbench 5 OpenCL Scores Spotted – 8GB, 2560 CUDA Cores, 1300MHz – 27% Slower Than Desktop 3070, 23% Faster Than Mobile RTX 2070 Super
1 January 2021
2070S 2080S 3070 AMPERE Benchmark GB5 Geekbench GeForce LAPTOPS leak MAX-P MAX-Q mobile MOBILITY NOTEBOOKS Nvidia OPENCL rtx turing

Since the announcement and supposed launch of NVIDIA’s latest Ampere based RTX graphics cards consumers have been left dumbfounded by the lack of supply worldwide however hope may finally arrive with RTX 3000 mobile GPUs seemingly just around the corner.

Considering how laptops and mobile devices are “consoles with a free TV” according to NVIDIA CEO Jensen Huang.

 

Mobile devices with Ampere based componentry might be a little more obtainable than the dedicated hardware of which they’re based as NVIDIA are too busy selling their discrete RTX 3000 products directly to miners.

Now it appears that the announcement for the mobility range of the RTX 3000 series is here with a mobile GeForce RTX 3070 being spotted from Geekbench’s OpenCL test of the Geekbench 5 benchmark.

Once again being spotted by twitter user APISAK, the tweet in question shows two individual results which were clearly ran in succession for a mobility GeForce RTX 3070 on a HP Omen 15 laptop featuring an Intel Core i7-10870H.

Before getting into the actual scores it’s worth pointing out the rough specifications of this mobile 3070 given out by Geekbench itself.

This mobility GeForce RTX 3070 features a total of 40 SM’s which gives it a CUDA Core count of 2560 or 5120 FP32 bullshit used as advertising material, however with a maximum frequency reported at just 1.29GHz it’s pretty indicative of just how inefficient the Ampere architecture truly is on Samsung’s horrendous “8nm” process.

Given that the discrete desktop RTX 3000 GPUs operate above the efficiency sweet spot resulting in the scale of power consumption and frequency to fall off the face of the earth it’s understandable at just how much of a clockspeed nerf is required in order to tame Ampere, when comparing a 2080Ti to that of a 3080 especially.

Continuing on this mobile RTX 3070 features the same 8GB of GDDR6 memory as also found on its desktop counterpart which in tail will also drag over its 256-bit memory bus though whether or not the memory itself is operating at a lower frequency is uncertain.

However the main question arises from this frequency that has other tech urinalists in a bit of a tizzy. Would this RTX 3070 be considered a Max-Q or Max-P graphics card? It seems that everyone is waging their bets on this being the Max-Q option which is tailored towards a much lower thermal envelope which itself is customized to whatever the partners desire. There’s no real definite answer on the wattage Max-Q products are limited to.

For shits and giggles I’m going to go the opposite here, with a maximum frequency of 1.3GHz I’m going to state here that this mobile RTX 3070 is that of a Max-P which offer higher thermal freedom and therefore higher frequencies and better performance.

I honestly do not believe Ampere to be much more efficient than the previous Turing architecture. I still believe there’s still more frequency to be lobbed off in order to get these RTX 3000 GPUs to reach optimal <90W TDPs, honestly I was expecting base frequencies down into the sub 1GHz territory to achieve it.

“Show me the fucking performance already”

If you haven’t already managed to see both results pages for yourself up above which once again can be found here and here. This mobility RTX 3070 for all that it’s worth certainly does pack quite a punch.

 

Taking the best score of 110,839 from the GB5 OpenCL test shows that this mobile variant of the GeForce RTX 3070 isn’t a pushover, however compared to the traditional desktop variant the mobility GPU lags behind severely as desktop RTX 3070’s manage OpenCL scores anywhere from 135,000 to over 140,000. For reference I’ve taken this recent result with an OpenCL score of 140,833.

At it’s very best the mobility RTX 3070 falls short of its discrete cousin by a staggering 27.06%, if we take a modest score of let’s say 132,500 the performance difference is still very nearly 20-percent.

However it’s not all doom and gloom as comparing against its “predecessor” the 2070 Super I’m met with a plethora of results taken from mobile devices such as the Razer Blade 15, ASUS Zephyrus along with HP and Lenovo solutions, the GeForce RTX 2070 Super in such configurations is more than happy to achieve scores of above 90,000 or there about give or take around a couple thousand points each way.

Taking the score of 90,098 as a baseline this RTX 3070 Mobile only proves to be just 23.02% faster than a 2070 Super which isn’t quite a substantial improvement but it’s certainly better than nothing. All things considered it’s a fairly nice little mobile graphics card but it isn’t exactly “that good”, granted a desktop 2080 Super pretty much eeks it out barely by over 5% the Ampere GPU certainly puts on a good showing but it’s no miracle.

However the real big issue I have here would be the potential and lethality of AMD’s RX 6000 series in a mobile configuration considering that with discrete desktop GPUs alone AMD simply dominates NVIDIA when it comes to power efficiency this generation around. What with the aid of a vastly superior TSMC 7nm processing node under AMD’s thumb along with the architectural tweaks that comes with RDNA 2 such as its Infinity Cache. When AMD and partners start ushering out RDNA 2 on mobile it’ll more than certainly be a bloodbath with AMD taking the convincing charge.

That I can assure you.

blog comments powered by Disqus