Wednesday, June 21st 2023
Geekbench Leak Suggests NVIDIA GeForce RTX 4060 Nearly 20% Faster than RTX 3060
NVIDIA is launching its lower end GeForce RTX 4060 graphics card series next week, but has kept schtum about the smaller Ada Lovelace AD107 GPU's performance level. This more budget-friendly offering (MSRP $299) is rumored to have 3,072 CUDA cores, 24 RT cores, 96 Tensor cores, 96 TMUs, and 32 ROPs. It will likely sport 8 GB of GDDR6 memory across a 128-bit wide memory bus. Benchleaks has discovered the first set of test results via a database leak, and posted these details on social media earlier today. Two Geekbench 6 runs were conducted on a test system comprised of an Intel Core i5-13600K CPU, ASUS Z790 ROG APEX motherboard, DDR5-6000 memory and the aforementioned GeForce card.
The GPU Compute test utilizing the Vulkan API resulted in a score of 99419, and another using OpenCL achieved 105630. We are looking at a single sample here, so expect variations when other units get tested in Geekbench prior to the June 29 launch. The RTX 4060 is about 12% faster (in Vulkan) than its direct predecessor—RTX 3060. The gap widens with its Open CL performance, where it offers an almost 20% jump over the older card. The RTX 3060 Ti presents around 3-5% faster performance over the RTX 4060. We hope to see actual in-game benchmarking carried out soon.
Sources:
Tom's Hardware, VideoCardz, BenchLeaks Tweet
The GPU Compute test utilizing the Vulkan API resulted in a score of 99419, and another using OpenCL achieved 105630. We are looking at a single sample here, so expect variations when other units get tested in Geekbench prior to the June 29 launch. The RTX 4060 is about 12% faster (in Vulkan) than its direct predecessor—RTX 3060. The gap widens with its Open CL performance, where it offers an almost 20% jump over the older card. The RTX 3060 Ti presents around 3-5% faster performance over the RTX 4060. We hope to see actual in-game benchmarking carried out soon.
30 Comments on Geekbench Leak Suggests NVIDIA GeForce RTX 4060 Nearly 20% Faster than RTX 3060
Performance isn't the worst part it's nvidia decision to reduce the vram amount is just stupid. This same card with 12GB while not overly exciting would have been a much better buy.
It would be nice if those mid-end GPUs under $300 had at least 10GB of Vram...
Besides, until we see actual game tests Geekbench is to be taken with a grain of salt on the performance side. It will be interesting to see how this thing runs in the wild at the estimated price point (If its true) of $300.
suckerslimited edition :nutkick:You can scream until you are blue in the face about utilization and how 8GB is still plenty, but once you are done clutching that pearl you will realize the day of the 8GB GPU is long over, much like 4GB cards before them, and no $150+ GPU should have that kind of frame buffer. FFS the 580 had a standard 8GB framebuffer in 2017. 6 years ago.
Any company that pushes 8 gb in 2023 is greedy beyond belief. Period.
While I think in 2023 people should not be buying 8GB cards if someone loves running games at 1080p medium and just cares about fps then 8GB is fine.
The memory bandwidth is such an issue and the fact that card manufacturers the last gen and this one seem to be removing features valuable in Linux have not been helpful at all. In fact some of the devs for different emulators have come out and said the last gen cards trounce this gen cards for emulation because the memory bandwidth is that bad unless you go up to the 4080 or higher.
And to the guy that said under $300 card. This is a $300 card. $299 is $300. Don't care what anyone says. 99 percent of people pay sales tax so it is going to be over $300 regardless. $279 would be under $300 but still in some states tax would push it over.
In my view not just the naming/pricing is hilarious, but more likely the last years of TDP gaining. This 4060 step back to the 1060 120W TDP level. The only problem is that it should be more likely a 4050/4050Ti rather than 4060. I am still not satisfied and overall disspaointed with the firms hipocrisy, because of sacrifice efficiency on the altar of performance.
DLSS can't save this garbage, last gen at this money is the way to go unless you blindly believe the marketing bollox.
There is an ignore button
I am just saying we need to be real when we talk about these cards. Ray Tracing is cool but its still only for cards 4070ti and higher that can make is somewhat useable in the few games that truly support it. We should not be trying to sell that as a feature that pushes it over the edge on the lower end cards as the performance degradation even with DLSS is just too high and not noticeable enough at 1080p.
And while I think 8GB cards should be in the sub 200 usd range the reality is that both AMD/Nvidia think 250-300 usd is the actual market for them well with Nvidia up to 400 usd lol which the 4060ti should really be DOA hopefully nobody buys that shite.
My belief is that Nvidia at least was so sold and enamored with sales with mining that they developed the 4000 series for miners and not gamers. They thought they would barely improve performance but would do it in a way that would cost less energy and be more energy efficient. So they wanted people to double dip and upgrade the cards based on power draw and not really any actual increases in performance. Ignoring the gamers while trying to spin DLSS and all this software garbage that doesn't matter to over 90 percent of pc gamers.
My informed guess is that its probably 10-14% faster on average at 1080p and actually loses to the 3060 12GB at 1440p and 4k.