Ada now starts at 300 USD. Are you excited?
NVIDIA is all-in on AI, but consumer GPUs are still a thing. And while marketing for even this modest little RTX 4060 will still center around AI-powered technologies (DLSS, frame generation), it is still a conventional desktop graphics solution. As such, there is the raster performance question, particularly as it relates to AMD’s latest offering.
And yes, this is another 8GB product in 2023, and I do remember that the RTX 3060 had 12GB of VRAM. Call it a regression if you must, but NVIDIA will point out that their big L2 cache and DLSS + FG will all but alleviate your VRAM concerns. But let’s be honest. No product with 8GB of VRAM that costs $300 in 2023 is going to gain the Gamers Nexus or Hardware Unboxed seal of approval. YouTube skews negative, and they will probably have a field day with this.
In this written review, I will present some actual data from benchmarks I have just completed running over the past few days, and let you decide if this card is worth your hard-earned money. No drama, just the facts, ma’am. Beginning with these specs (which I’ll probably have to update since I am not 100% sure on a couple of them):
|RTX 4060||RTX 4060 Ti||RTX 4070||RTX 4070 Ti||RTX 4080|
|Architecture||Ada Lovelace||Ada Lovelace||Ada Lovelace||Ada Lovelace||Ada Lovelace|
|Tensor Cores||96 (4th Gen)||136 (4th Gen)||184 (4th Gen)||240 (4th Gen)||304 (4th Gen)|
|RT Cores||24 (3rd Gen)||34 (3rd Gen)||46 (3rd Gen)||60 (3rd Gen)||76 (3rd Gen)|
|Base Clock||1830 MHz||2310 MHz||1920 MHz||2310 MHz||2205 MHz|
|Boost Clock||2460 MHz||2535 MHz||2475 MHz||2610 MHz||2505 MHz|
|Memory||8GB GDDR6||8GB GDDR6||12GB GDDR6X||12GB GDDR6X||16GB GDDR6X|
|Memory Data Rate||17 Gbps||18 Gbps||21 Gbps||21 Gbps||21 Gbps|
|L2 Cache Size||24MB||32MB||36MB||48MB||64MB|
|Memory Bandwidth||272 GB/s||288 GB/s||504 GB/s||504 GB/s||716 GB/s|
|Die Size||146 mm^2||190 mm^2||295 mm^2||295 mm^2||379 mm^2|
|Process Tech||TSMC 4N NV Custom||TSMC 4N NV Custom||TSMC 4N NV Custom||TSMC 4N NV Custom||TSMC 4N NV Custom|
Just my opinion, but the $100 USD gap between the RTX 4060 and RTX 4060 Ti looks wrong. Or maybe it’s the gap in specs / capability between the two. Should they be this disparate? Isn’t there room for another SKU in between these two? Maybe I’m missing something.
The ASUS DUAL Card
The entry-level solution from ASUS, named after its dual-fan configuration, the DUAL is a compact unit that is still quite a bit larger than the PCB hiding beneath the cooler.
The most interesting aspect of this design to this editor? That 8-pin PCIe connector. Yes! It’s tough, it’s reliable, it’s on every PSU. You just can’t beat it.
Some Performance Benchmarks
|Processor||Intel Core i5-13600K (Stock, Power Limits Enforced)|
|Motherboard||MSI MPG Z790 EDGE WIFI DDR4
BIOS v1.60, Resizable BAR Enabled
|Memory||16GB (8GBx2) PNY XLR8 REV DDR4-3600 CL18|
|Storage||Samsung 980 1TB NVMe SSD|
|Power Supply||be quiet! Dark Power Pro 13 850W|
|CPU Cooler||be quiet! Pure Loop 2 FX 280mm AiO|
|Operating System||Windows 11 Pro, 22H2|
|Drivers||GeForce GRD 531.79 – 536.20 (RTX 4060 Press Driver)
Adrenalin 23.4.3 – 23.5.2
Here’s a look at four game benchmarks, beginning with Far Cry 6 – this time around with the HD textures turned off (having these on for 8GB and under cards is just asking for trouble).
Now, Far Cry 6 isn’t the most NVIDIA-optimized title around, so maybe that’s why the RX 7600 has the lead, right? Let’s look at some more titles:
I’m not sure what to make of these results. Clearly, the Radeon RX 7600 has improved with driver maturity since launch, and of course NVIDIA will continue to showcase performance using DLSS 3.0 with Frame Generation (which we will get to shortly).
For a sanity check, I went back and ran 3DMark Time Spy on all eight cards (three runs each, averaged), and here are the GPU scores:
The Radeon RX 7600 is just slightly ahead of the GeForce RTX 4060 in Time Spy, but still, I didn’t expect any advantage (let alone the larger disparity in the above game results) considering the pricing. AMD doesn’t have the same features as NVIDIA (remember, AMD’s FSR 3.0 with frame generation is still MIA), but they are winning the sub-$300 raster war this generation. However, we can’t forget that NVIDIA is all about that DLSS and ai-generated frames, so let’s check that out.
I’m sure what’s up with the lackluster DLSS + FG results from this RTX 4060, as the RTX 4060 Ti really shines in that area. Perhaps this is a game optimization or driver issue, and we will see more DLSS 3.0 uplift after launch.
Finally, a look at power draw. I decided to keep things simple and just show gen-on-gen performance here:
The RTX 4060 has a rated TGP of 115 watts, though it is clearly exceeding that here. Still, we are looking at a solid improvement in efficiency over the RTX 3060 – while offering a sizable performance advantage over this 30 Series counterpart.
If you are looking for a GPU at this price level, the only competition from this generation (forgetting the mountain of previous-gen cards in the pipeline) is AMD Radeon RX 7600, and the Intel Arc stuff I didn’t test as I was running behind schedule. If we are talking rasterization, AMD is winning the sub-$300 area this generation, but the 4060 is an RTX product, and that means ray tracing and DLSS.
Naturally, DLSS with the RTX 40 Series means frame generation, and frame generation means higher perceived FPS on top of the performance gains from DLSS. AMD is apparently trying to thwart NVIDIA’s DLSS advantage by blocking the feature on games in which they are partners (see Starfield drama), but many, many titles do support NVIDIA’s upscaling tech. And, honestly, if you’re buying a 4060 with 8GB of VRAM, you should use DLSS.
I just can’t help but think that if this product had been called the RTX 4050 Ti – even if it was launching at the same price (though that wouldn’t go over well at all) – there would still be room for a cut-down AD106 product between this and the RTX 4060 Ti 8GB. And that product could have been the RTX 4060. Product names and pricing might be all we talk about when we look back on this generation of NVIDIA GPUs, years from now.