An nVidia RTX 4060, 4070, 4080, 4090, A6000, A3000, Others Chart

They are still quite expensive, and take up much more power than the 4xxx series.

The 3090 (non-ti) is still over $1200, and loses to the 4080.

I’m seeing them on FB Marketplace for around $750.

I thought the same thing, about the 3xxx series. The memory is nice, but the 4xxx series had shockingly big jump. The 4070 Ti is now faster at rendering.

With 48GB of memory and nearly on par with the 4090, the A6000 remains a great x-mass stocking-stuffer : )

I will updated the chart, using used Founders Edition prices from ebay.

While the 4xxx did jump over the 3xxx, the fact that the card has 24GB of memory is quite provocative.

In a way, it’s the 24GB 4070ti/4080 nVidia will never make. We will see nVidia update their Quado products, about the time of the 4090Ti release.

Your Santa must be vastly superior to my Santa!

I think in real world usage, you’ll not notice the performance difference all that much between the 4070 ti and the 3090, but you WILL notice all the extra RAM of the 3090/3090TI. @AlanMattano and I have been doing a big render stage and texturing project, and I can tell you the 24 gigs is MUCH needed. when you really get into it.

2 Likes

Thank you. Yes I have tried it. I haven’t tested for speed, but my laptop can you Optix, so it should be faster.

The interim problem is: the lighting, perhaps the Sun is quite brighter in the WIP version, so I would have interoperability between the versions.

If it was me, my ranked choices would be:

  1. A lightly used 3090 or 3090ti.
  2. 4060 Ti 16GB
  3. 4080
2 Likes

Sorry, they are NOT saying good things about the 4060. The 3090ti is tempting because of the RAM, but it takes a lot of power for a 4070-classed card. The 4080 seems like the cat’s meow : )

1 Like

I’m building a machine for my friend’s son - I managed to snag a returned 4080 from Best Buy at $1,000 - came with a full warranty, wasn’t even opened as best as I can tell, because the power cord was still sealed. It’s worth keeping an eye on BB regularly. For instance, this shows up for me right now:

This is the card we got - another open box on this one too:

Open box, $959 is pretty good! Of course you need to be close-ish to a Best Buy - they won’t ship cards to your home, and the returned unit has to be in your general area - close enough that they’ll send it to your local store for pickup.

Heck it’s raining returned cards at BB:

$911.99!

I am sorry, but I put a typo in above. I am also, but I can’t let it stand, because it might affect someone’s buying choice.

There are rumors of a 4070/4070ti based from binned 4080 cores. If the 4070 Ti had 16GB, that would be a better financial choice for me, right now.

I saw a cheaper 4080, but they wised up.

I might as well state this: I hate the new power connectors on the cards. There was no reason to change what worked. All of this nonsense because people can’t read the number on a power supply box–so they have to come out with a key’ed system, with the idea to sell new power supplies. Then the FN wires melt because they expect too few resistive contacts to conduct too many feisty electrons. Idiots.

It does seem that the general response to the 40 series varies cards between “meh”, and “this is awful”.

If I were to switch to nvidia, I’d certainly avoid the 4080, and just save a bit more and go for a 4090.

It seems the 4070 and the 4060 non-Ti versions have certainly had the least bad feedback (meh) for thier price. But by those points, the choice widens to AMD I think, if you use a renderer that will use AMDs acceleration or hardware.

We don’t even get the nice flash discounts in Europe like the US. Not anywhere near as often anyway.

1 Like

The thing I like about the 4080, especially the Founder’s Edition was: the card runs cool, because it’s the same cooling solution used for the 4090. Also…

As a customer, nVidia are you reading this: I don’t trust your rotten little connector to endure 450 watts. There was no reason to reinvent the wheel. No, we aren’t all going to throw away our power supplies, such as my Corsair HX850, and buy a new one. You see, there was already a system, whereas you plugged in the amount of 12v pairs, mounted in a connectors, and you plugged in as many connectors as was needed, such as 3 of 8 pin connectors, or 2 of 6-pin connectors. They even had 6/8 pin connectors. It was all worked out, already.

I agree, just going the little extra $300 gets 25% performance, but more importantly: 8GB more vRAM.

The thing I am afraid of: I am using rackmount cases, with power going from 320 to 450 watts is a big difference in heat. While the 4090 would not have to run as long, there likely will be times whereas the card is rendering 24 hours at a time, such if I want to do a 4k animation.

Also, if I game with the thing, am I really going to want to be in a 10’x12’ room, with a 450 watt heat source–in summer? Perhaps I can frame-limit a RTX 4090 to make it more civilized. And just wait until the 4090TI comes out with it’s 600-watt power draw. Then there will be the Quadro part, which will likely have more memory, such as 48GB.

In the fall, AMD is expected to roll out some new cards, too. It is my hope that they will challenge nVidia. My wish: AMD works on its drivers, and comes out with a card that isn’t overlocked incandescent hot before it can compete.

If you are nervous about the 12VHPWR cable, the specification has just been upgraded to 12V-2x6. This introduces changes from PCI-SIG, aimed at resolving the connection pull-out that affected this type of connector, and typically the 4090 burn out. Of course, I can’t envisage nvidia or any partners telling us which serials this affects; as there is so much stock of nearly all 40 series cards.

nvidia have seemingly no particular motivation to produce a 4090Ti at the moment, maybe later this year. Like AMD, they are sort of dropping gamers at the moment in favour of the new hip thing; AI. This seems to have swept up general productivity activities too; as the push will be just get the AI to do it. Especially as such AI production lends itself to Apple-like closed systems on the cloud. Maybe when the hype calms down…

The same connectors we were using for years, worked.

I have an additional problem: nVidia puts their power connectors perpendicular to the top edge of a card, which is a real detriment to rackmount case use, well, and mid towers, too. Without even looking at it, I can just see the new stupid connector–not bending the long way, because it’s long, whereas the cables with two short connectors, could make bend from either axis.

Because there are no true blower cards, I might get an FE, which seems to have one fan blowing, and the other mixing, and create a duct to route the mixing heat out of the back of the case. It’s going to be one of those Cardboard Aided Design until I get it roughed out. Perhaps it would be a better aluminum sheet metal project than a ABS one.

We need to rework the ATX spec for video cards and modern CPUs. How the CPU and GPU fan interfaces with the case–should be part of the spec. Currently, we have CPUs kind of facing the holes in the back of the case–and the GPUs fans at right angles.

The result would be a little like Dell used to do with their CPU fans being on the backplane–with a green duct moving the air from the heatsink. Now, imagine it with about 100mm-120mm fans. So, if I did it, if you looked at the back of the video card, you would see a fan staring at you, with 4 connectors going up the side, or along the bottom. Such a system could only have 2 video cards. Well, they are talking 4 slots for the 4090ti, anyway. The biggest problem in a modern system is that with all the port connectors running along the bottom edge of the board, there’s little room for anything else. I admired the Cheeze-Grater Mac for it’s cooling design; the new one is ugly.

Most unfortunate that the EVGA 4090 was never mass produced, as they moved the power connector to the back/opposite face to the slot.

I suppose if you wanted, there are cablemod 90-degree adaptors for the power connector.

In addition, if you wait for, or find a 12V-2x6, I think the cable has a minimum length before bending defined, so it may may it more difficult for your situation.

The only card (and I think the thinnest other than a China-only MSI 4090 blower) I could think of was the MSI Ventus 3X, which had no RGB, and offered some space between for air flow if you had two of them. Not even sure if you can still get them.

If I change to any 4090 config from my current card, I will have to change case too.

@Brenda,

Have you seen the Nvidia RTX A4000 SFF Ada-generation card announced in March? Small form factor (rackmount?), low power (70W) biggish memory (20GB).

Specs here: NVIDIA RTX 4000 SFF Ada Generation | NVIDIA

Downsides: Performance may be similar to an RTX 3070 and only price I’ve seen is circa $1500 at Amazon.

Regards
Jeremy

1 Like

I did a similar comparison for our tiny render farm using Octane Render, which we had a while ago.

[EDIT: I am updating the image with less information and newer prices. Also split the 4060ti into models with 8 and 16 GB since there is a big price difference at the moment)

That’s what it looks like. You basically compare OBR (Octane Bench Result) with Price, to see how you can maximise your investment. The most important factor is Price/OBR. You can see that the RTX 4060 is quite good, but has a low total OBR. You would need 4 of them to match the top performer the 4090. 4x4060 actually costs less than the 4090, but of course its more complex and brings with it other costs (bigger PSU, bigger mainboard, etc).

Usually the 70 series (so 3070, 4070, etc.) is the best balance of power and price in my opinion. You only need 2 to match the top performer and its still cheaper. You can start with 1 and add another later when the price has come down more.

Nowadays actually Memory has quickly become a really important factor. Nvidia is only now realizing this and releasing a 4060 with 16GB. Unfortunately that is why a lot of the older GPUs are not worth considering any more, even though they are often still plenty powerful.

The columns on the right especially are for checking whether it is more sensible to have multiple GPUs and how much the price of for example a used GPU can be to be “competitive”.

The prices are in Franc and are just the prices I can find currently. I have not updated the older models recently. I also only use the gaming GPUs and not the professional A series ones.

4 Likes

6,144 is between an RTX 4070 and 4070 Ti’s core count, but they often adjust the clockspeed, well whatever binning demands : )

19.2 TFLOPS is lower though than a 4070 Ti @ 29.15 TFLOPS
The memory size at 20GB is nice and fluffy.

I like the color coding. Nice chart. Better than mine. : )

The sad thing is the RTX 3090 TI, so much memory at 24GB, RTX 4070 performance though. Nice used price @ $850 U.S.

If they don’t come out with a 16GB RTX 4070, soon, I might have to get a 4080. I like the idea of the 4090’s oomph, but I don’t know if I would like to live with it.

It has been suggested that this has been part of the marketing tactic so far. The value proposition and poor choices for RAM, added with the poor bandwidth of the 60s seems to basically force users up the chain to the 4070 (which has missing RAM)… then to the 4080 (terrible value)… then to the 4090 (yikes!), to get something completely overkill, but actually does something the user wants.

I came to exactly that conclusion when I started shopping this year. I just kept ending up at 4090, because of the strategically bad combinations; if I wanted a 40-series card. Fortunately, Vray made the decision for me that I didn’t need the expensive GPU at all… yet.