Nvidia GeForce RTX 4060 Review: Truly Mainstream at $299

Nvidia GeForce RTX 4060 drops entry price to $299 for Ada Lovelace architecture and RTX 40-series GPUs. It sits between the previous RTX 3060 and RTX 3050 in pricing, showing a potentially great value proposition, although there are always compromises to make as you move down the price-performance ladder. For those on a tight budget, he could be one of the best graphics cards if the performance is good enough.
There are some legitimate complaints about the AD107 GPU at the heart of this card being limited to a 128-bit memory interface, but the lower price point compared to the RTX 4060 Ti alleviates some of that burden. Still, the previous generation of his RTX 3060 had him with a 192-bit interface and 12 GB of memory, so this represents a clear setback in that area. It’s an important topic, so I’ll go into more detail on the next page.
We will be updating the GPU benchmark hierarchy later today as the embargo has ended. The conclusion is not so surprising. For most games, his new RTX 4060 easily outperforms his RTX 3060 from the previous generation. largely Catch the RTX 3060 Ti. Considering DLSS 3’s frame generation and dramatically improved efficiency, it makes sense to buy the RTX 4060 over the previous generation card. You won’t get new levels of performance, but you’ll get all of his latest Nvidia features and upgrades.
There are two main competitors with AMD.latest generation Radeon RX 7600 cuts Nvidia prices by up to $50 Now, in previous generations RX 6700 XT Pricing Starting at $309The price is basically on par with the RTX 4060 but with 50% more memory and potentially better overall performance. Depending on price and availability, the RTX 3060 Ti (and other 30-series GPUs) could also be an interesting option, but unless you’re willing to buy a used card, such a card won’t last long. I can’t wait.
Let’s take a look at the specs revealed over a month ago with the RTX 4060 Ti announcement. Nvidia is currently allowing reviews of cards with a MSRP of $299, but the more expensive models are barred until tomorrow. I just received the Asus RTX 4060 Dual OC model from Nvidia. It comes with a modest factory overclock, but is still priced at $299.
graphics card | RTX4060 | RTX 4060 Asus Dual OC | RTX4060Ti | RTX4070 | RTX3050 | RTX3060 | RTX3060Ti | RTX3070 | RX7600 | RX6700XT | Ark A770 16GB | Ark A750 |
---|---|---|---|---|---|---|---|---|---|---|---|---|
architecture | AD107 | AD107 | AD106 | AD104 | GA106 | GA106 | GA104 | GA104 | Navi 33 | Navi 22 | ACM-G10 | ACM-G10 |
process technology | TSMC 4N | TSMC 4N | TSMC 4N | TSMC 4N | Samsung 8N | Samsung 8N | Samsung 8N | Samsung 8N | TSMC N6 | TSMC N7 | TSMC N6 | TSMC N6 |
Transistor (billion) | 18.9 | 18.9 | 22.9 | 32 | 12 | 12 | 17.4 | 17.4 | 13.3 | 17.2 | 21.7 | 21.7 |
Die size (mm^2) | 158.7 | 158.7 | 187.8 | 294.5 | 276 | 276 | 392.5 | 392.5 | 204 | 336 | 406 | 406 |
SM/CU/Xe core | twenty four | twenty four | 34 | 46 | 20 | 28 | 38 | 46 | 32 | 40 | 32 | 28 |
GPU core (shader) | 3072 | 3072 | 4352 | 5888 | 2560 | 3584 | 4864 | 5888 | 2048 | 2560 | 4096 | 3584 |
Tensor / AI core | 96 | 96 | 136 | 184 | 80 | 112 | 152 | 184 | 64 | N/A | 512 | 448 |
Ray Tracing “Core” | twenty four | twenty four | 34 | 46 | 20 | 28 | 38 | 46 | 32 | 40 | 32 | 28 |
Boost clock (MHz) | 2460 | 2505 | 2535 | 2475 | 1777 | 1777 | 1665 | 1725 | 2625 | 2581 | 2100 | 2050 |
VRAM Speed (Gbps) | 17 | 17 | 18 | twenty one | 14 | 15 | 14 | 14 | 18 | 16 | 17.5 | 16 |
VRAM (GB) | 8 | 8 | 8 | 12 | 8 | 12 | 8 | 8 | 8 | 12 | 16 | 8 |
VRAM bus width | 128 | 128 | 128 | 192 | 128 | 192 | 256 | 256 | 128 | 192 | 256 | 256 |
L2 / Infinity Cache | twenty four | twenty four | 32 | 36 | 2 | 3 | Four | Four | 32 | 96 | 16 | 16 |
ROP | 48 | 48 | 48 | 64 | 48 | 48 | 80 | 96 | 64 | 64 | 128 | 128 |
TMU | 96 | 96 | 136 | 184 | 80 | 112 | 152 | 184 | 128 | 160 | 256 | 224 |
TFLOPS FP32 (Boost) | 15.1 | 15.4 | 22.1 | 29.1 | 9.1 | 12.7 | 16.2 | 20.3 | 21.5 | 13.2 | 17.2 | 14.7 |
TFLOPS FP16 (FP8) | 121 (242) | 123 (246) | 177 (353) | 233 (466) | 36 (73) | 51 (102) | 65 (130) | 81 (163) | 43 | 26.4 | 138 | 118 |
Bandwidth (GBps) | 272 | 272 | 288 | 504 | 224 | 360 | 448 | 448 | 288 | 384 | 560 | 512 |
TDP (Watts) | 115 | 115 | 160 | 200 | 130 | 170 | 200 | 220 | 165 | 230 | 225 | 225 |
release date | July 2023 | July 2023 | May 2023 | April 2023 | January 2022 | February 2021 | December 2020 | October 2020 | May 2023 | March 2021 | September 2022 | September 2022 |
Release price | $299 | $299 | $399 | $599 | $249 | $329 | $399 | $499 | $269 | $479 | $349 | $289 |
online price | $300 | $300 | $380 | $585 | $220 | $260 | $275 | $400 | $250 | $310 | $340 | $240 |
Scrolling to the right, the table above lists 12 GPUs representing the most useful comparisons for the RTX 4060, but the first column is the most relevant. The new GeForce RTX 4060 will use Nvidia’s AD107 GPU, the same chip that powers the RTX 4060 and 4050 laptop GPUs.
The RTX 4060 uses the entire AD107 chip and features 24 streaming multiprocessors (SMs) with 128 CUDA cores each. This brings the total number of shaders to 3,072. Any astute mathematician will note that this is less than his 3,584 shaders on his RTX 3060 of the previous generation. However, like the rest of the RTX 40 series lineup, the clock speeds are significantly higher at 2460 MHz compared to 1777 MHz on the 3060. As a result, peak compute performance is ultimately 19% higher.
Memory bandwidth is lower in raw throughput at 272 GB/s compared to 360 GB/s on the RTX 3060. However, the L2 cache has been inflated from 3MB on the 3060 to 24MB on the 4060, which Nvidia said has increased the effective bandwidth by 67% to 453 GB/s. The memory subsystem and its implications are detailed on the following pages.
One thing to note is that the RTX 4060 features an x8 PCIe interface, while the RTX 4060 Ti and newer use x16 link widths. It’s similar to the RX 7600 and his RTX 3050 from the previous generation, and cuts out additional PCIe lanes to keep the die size small. This isn’t much of an issue on most modern PCs, but if you plan to upgrade an older PC that only supports PCIe 3.0 to his RTX 4060, you’ll see a slight drop in performance compared to what the benchmarks show. may decline.
There are many options when compared to competitors based on relatively similar pricing. AMD has a new RX 7600 8GB card alongside the previous generation RX 6700 XT 12GB and RX 6700 10GB. From Intel we have the Arc A770 8GB and the Arc A750. Next, Nvidia will also have to contend with existing cards like the RTX 3060, RTX 3060 Ti and RTX 3070. There’s no doubt that Nvidia can match or beat AMD and Intel cards when it comes to ray tracing performance and AI workloads, but he could lose out to the 3060 Ti and beyond in the same task. Rasterization performance should be an even tougher battle for beginners.
We also include results for the RTX 2060, which launched in early 2019. While many gamers will skip his one or two generations in hardware, Nvidia (similar to AMD and the RX 7600) still touts his RTX 4060 as an excellent upgrade path for gamers. Use cards like GTX 1060, RTX 2060, RX 570/580/590. With all the complaints about the RTX 40 series and its high gen pricing, it’s nice to see Nvidia priced on par or even above that of his two previous generation GPUs. The RTX 2060 launched at $349 before dropping to $299. The RTX 3060, on the other hand, launched at $329, a price we’ve barely seen until the last few months.
The block diagram of the RTX 4060 / AD107 shows just how many elements Nvidia has cut to match mainstream pricing. Most other Ada chips have multiple NVDEC/NVENC blocks, but the AD107 only has one of each. As mentioned above, there are only 24 SMs in total, spread across 3 GPCs (Graphics Processing Clusters). Finally, Nvidia offers up to 8MB of L2 cache per 32-bit memory channel, while only his 6MB is enabled on the RTX 4060 for a total of 24MB. (Mobile RTX 4060 gets all 32MB.)
Like other Ada Lovelace chips, the RTX 4060 comes with Nvidia’s 4th generation Tensor Cores, 3rd generation RT cores, new and improved NVENC/NVDEC units for video encoding and decoding with AV1 support, and significantly more power. includes an optical flow accelerator (OFA). ). The latter is being used for DLSS 3, indicating that Nvidia has no intention of enabling frame generation on his Ampere and his earlier RTX GPUs.
Tensor Cores now support FP8 with sparsity. It is not clear how useful this will be for different workloads, but AI and deep learning can, at least in some cases, ensure a less precise numerical format to improve performance without significantly altering the quality of the results. is used for Ultimately it depends on the work being done, but it can be difficult to understand when to use FP8 vs FP16 and sparsity.
Of course, running AI models on low-cost mainstream cards like the RTX 4060 is not the main goal. Yes, stable diffusion works. I’ll show the test results later. Other AI models that fit in 8GB VRAM will run similarly. However, anyone serious about AI and machine learning will almost certainly want a GPU with more processing power and more VRAM.