What we think about it:
JacqueMorrison - "Can’t have negative reviews impact the launch date if you don’t send cards to reviewers. taps forehead"
rtplor - "4060 reviews were not that inspiring and optimistic."
FlyerFocus - "Oh, how I love my $NVDA stock."
ThehemanCringercat - "Made a few enquiries today among the old distribution folk I used to work with. No importer has any, nor has any due dates. Best I got was "Tyne Wares thinks it might have a few palettes [almost nothing] first week of August" and very few of them had any retailer orders in after initial stock."
MetricT - "How dependent is LLM training on speed vs RAM? Would this make a good card to tinker with 65b LLaMA's?"
SolidPoint - "Way too many llamas"
D1stRU3T0R - "You might be able to load the 7B, but you usually need 2GB RAM for every 1B parameters, and this is for inference only For training you will need more."
aliveform - "Yes, but this doesn't give more fps"
OneMetalMan - ""I...need.....just....one....more""
dubvision - "Who the heck buys 4060 for Houdini lmao. These are more gaming cards cuz they aren't very strong for that much beyond gaming. Even for basic 3D animation rendering with Ray tracing etc people will usually opt for a 4070 or 4080."
dubvision - "All those links seem to be discussing 4070 and up like I said. Why would you get a 4060? 4060 seems like it's exactly only good enough for gaming like 3060 was."
dubvision - "No cards either. It's a paper launch."
