Nvidia’s presentation at CES 2025 was a bumper one, because the tech big unveiled the entire particulars in regards to the upcoming launch of the Blackwell structure 50 collection.
Amongst all of that juicy new information was loads of spectacular efficiency guarantees, and these new playing cards are undoubtedly a marked step up from their equivalents within the earlier era.
The revamped structure is geared in the direction of permitting AI to make important enhancements to the general expertise, prompting CEO Jensen Huang to recommend that Blackwell is “probably the most important pc graphics innovation since we launched programmable shading 25 years in the past.”
That is all very thrilling, but it surely was one other assertion from Huang throughout his keynote presentation that drew all the consideration.
When speaking in regards to the upcoming launch of the 5070 – the funds choice in comparison with the 5080/5090 – Huang urged that those that purchase the cardboard can be getting “RTX 4090 efficiency at $549.”
A daring declare, however is it true? Effectively, after trying on the at present out there knowledge: sure and no.
What’s underneath the hood? 4090 vs 5070
NVIDIA
To begin this dialog as objectively as doable, a side-by-side comparability of the 2 playing cards doesn’t mirror notably effectively on the 5070. Although the Blackwell structure is objectively higher than the Ada Lovelace foundations of the 4090 (with help for Show Port 2.1), that’s largely the place the outright {hardware} benefits finish.
The 4090 has a monstrous 24GB of GDDR6X reminiscence in comparison with the 5070’s 12GB of GDDR7. By way of parallel processing, the 4090 additionally vastly outstrips its rival, packing 16,384 CUDA cores, whereas the 5070 has simply 6,144. It additionally has double the bus width and nearly double the bandwidth, and over thrice as many TMUs and ROPs, leaving the 5070 with just about no tangible benefits on paper.
So, what is that this debate all about? Effectively, it comes right down to a heady mixture of AI, true efficiency vs perceived efficiency, and a touch of obfuscation on Nvidia’s half.
DLSS 4 is doing the heavy lifting
NVIDIA
Jensen Huang’s claims are primarily based on the theoretical output supplied by DLSS 4, a significant step up from the already pioneering DLSS 3.5 that the earlier era of playing cards makes use of. DLSS 4 has a brand new Multi Body Technology approach that generates 3 further frames for each historically rendered one.
Clearly, the distinction this makes to the person is important, dialing them up from 30 FPS to 120 FPS with the expertise energetic. The addition of neural rendering provides much more to that tapestry, permitting them to moderately declare parity with the 4090.
Equally, Nvidia’s CEO additionally put the 5070’s efficiency at twice the extent of its equal within the earlier era. With all being equal, enjoying a recreation with DLSS 4 on the 5070 will give roughly twice the efficiency of the identical setup with a 4070 on DLSS 3.5.
The most important concern right here is that many video games don’t help DLSS 4 at this stage, rendering these efficiency positive aspects moot. With out DLSS 4, you’re solely a efficiency enhance of round 30%.
Whether or not you like the funds providing of the 5070 or the titanic output of the 4090 doubtless depends upon your perspective, however don’t be fooled into considering these playing cards are one and the identical.
On the face of it, the 5070 has the potential to be the perfect funds card of all time throughout the context of its quick opponents, but it surely doesn’t have the {hardware} underneath the hood to compete with the 4090 in a straight shootout.
In any case, the 5090 is the brand new monster on the scene, so when you’ve got a few thousand {dollars} to burn, then that’s the one to go along with.
