4060 Ti info announced - first cards imminent

[AGTDenton]

Real prices are coming in for the 4060Ti on various stores.

So far no FE editions, and the cheapest custom card is £389.99 in stock (Scan) / now £379.99 @ OcUK

The next jump is to £409.99 (CCL/OcUK) but this will probably all change in the coming days.

OcUK have a preorder Gainward card at £374.99 which would make it the cheapest I've found so far, tho not in stock.

Palit have done a really nice short single fan card which is in stock @ OcUK, for £379.99
Would be interesting how that fairs in a small case.

[DanceswithUnix]

It will be very interesting to see how the 16GB version street prices line up. If the card manufacturers are buying just the GPU chip from Nvidia, then that will be a free for all on the increase in price over 8GB. OTOH, if Nvidia are selling a kit with vram (which I know has happened in the past) then the price on that might get skewed.

[kompukare]

It will be very interesting to see how the 16GB version street prices line up. If the card manufacturers are buying just the GPU chip from Nvidia, then that will be a free for all on the increase in price over 8GB. OTOH, if Nvidia are selling a kit with vram (which I know has happened in the past) then the price on that might get skewed.

Has anyone mentioned how much Jensen loves Apple? So I think it probably is a kit with a huge markup

[AGTDenton]

der8auer adds some clarification to his original video as well as a detailed comparrison with the 3060Ti which he was lacking in the previous video he did.
A very useful comparison of power consumption over 3 years.

[CAT-THE-FIFTH]

As expected it performs worse on a PCI-E 3.0 system,so is no different than an RTX3060TI.

But he assumes you play 4 hours a day of Cyberpunk for 365 days a year which most people won't do. But the issue is his figures are quite large(official power figures?):


TPU measure the power consumption of dGPUs directly using Cyberpunk 2077. 53W difference,which is below his 88W figure.

So if you game for 20hours a week(say during the weekend),and take two weeks off,that is 1000 hours a year. That is equivalent to 53KWh.

The new price cap is 30p/KWh so that works out or £15.90 a year. After three years it will be £48. Even if it was 40p/KWh that would be £64.

You can get an RTX3060TI for £300ish,and considered how much power I saved off mine,using custom curves I don't think even lower power consumption can save it. IIRC,I reduced power consumption by 30W~40W.Then you have had the RX6700XT for £285 after cashback too.

But all this power consumption bumpf is Nvidia marketing - nobody was mentioning how much less power the RX6000 series was drawing. The RX6600/RX6800 were the most efficient cards of their generation.

Anyone who spends nearly £400 on an 8GB dGPU in 2023,sorry to say this,needs a reality check.

Everyone likes lower power consumption,but if someone can't even run a 200W dGPU,it's most likely they also don't have a lot of money either.

You can get an RX6600 for under £200 now.



Its a shame TPU has forgotten about the RX6600,but it consumes about 30W less than an RX7600/RTX4060TI.

With undervolting,etc it can be pushed under 100W.

[Iota]

The RTX4060TI is literally an RTX3050 replacement.

Seems fairly accurate, it follows up the product stack as well.

According to Nvidia the ray tracing performance has nearly doubled compared to 3060ti, and tensor core performance has nearly tripled. If shader performance had gone up like that, then I'm sure I would give the card a thumbs up. But I'm still not convinced I care about RT and tensor performance.

It's not much use without rasterisation performance and the memory pipeline to feed it though.

I know naming isn't important as such but:

3050 - 106 die, 23.8% of full die cores, 276mm², 128-bit bus, 8GB GD6, PCIe4 x8

4060ti - 106 die, 23.6% of full die cores, 190mm², 128-bit bus, 8GB GD6, PCIe4 x8

I think we all know Nvidia is trying to sell the lower tiers at higher tier prices, just by renaming the card and trying to obfuscate by talking about RT performance gains and increased tensor cores etc.

It will be very interesting to see how the 16GB version street prices line up.

Is 16GB of vram going to make much difference with such a narrow memory pipeline though?? I guess we'll find out.

[DanceswithUnix]

Is 16GB of vram going to make much difference with such a narrow memory pipeline though?? I guess we'll find out.

It won't make any difference to the average fps at normal resolutions because 8GB can still cope with that. What it will help with is avoiding using that x8 PCIe bus as much to get data from system ram to vram, which should help the 1% low stutters and stop it falling off a cliff at higher resolutions.

[[GSV]Trig]

That De8auer video once again says that Nvidia claim the extra gpu cache will make up for only having 8GB of vram. Do Nvidia really say that? Or are they being unclear because that's just plain wrong.

The extra cache will help get over the 128bit bus to avoid fetching from vram, which it seems to do a passable job of right up to 4K resolutions where it collapses under the strain. But if data isn't in the undersized vram then that's a fetch from system ram and the cache isn't going to help that one single bit. Not a jot.

Was it AMD that said that the GPU cache would make up for the slower bus, not a lack of RAM?
Unless NVidia have got like an extra few Gb of cache on there to cover the lack of RAM and a slower bus, doesn't make sense...

Fair to say that the 40 series isn't going to leave the same mark as the 30 series did, the 3060Ti was a wicked card while I had it, but safe to say the 3080 will last me a good while yet with its extra few gigajiblets..

[DanceswithUnix]

Was it AMD that said that the GPU cache would make up for the slower bus, not a lack of RAM?
Unless NVidia have got like an extra few Gb of cache on there to cover the lack of RAM and a slower bus, doesn't make sense...

Fair to say that the 40 series isn't going to leave the same mark as the 30 series did, the 3060Ti was a wicked card while I had it, but safe to say the 3080 will last me a good while yet with its extra few gigajiblets..

It was a passing comment on a couple of videos, which sounded like it came from the press pack that companies always supply with review cards.

So yes, it doesn't make sense. But then review material will be provided by Nvidia marketing, not engineering

[Iota]

It won't make any difference to the average fps at normal resolutions because 8GB can still cope with that. What it will help with is avoiding using that x8 PCIe bus as much to get data from system ram to vram, which should help the 1% low stutters and stop it falling off a cliff at higher resolutions.

I guess it's a 1080p card really, 1440 is probably pushing it, 4K is a step too far. For the price though, it should cope with 4K. Heck I'd want it singing and dancing and doing the dishes for that money....

[Saracen999]

It was a passing comment on a couple of videos, which sounded like it came from the press pack that companies always supply with review cards.

So yes, it doesn't make sense. But then review material will be provided by Nvidia marketing, not engineering

I don't know about either nVidia (*) or current industry practices but it certainy used to be the case that nearly all communication would come from the presss office, but SOME was provided by engineering.

It's also the case that if you want to do a release day review, you need to be on the press office list. The issue is, what you make of that. For instance, I used to get review kit (with an NDA and a dated release) weeks or sometimes months in advance of public shipping. The logic was that if you wanted a review done by the time a keen public could actually buy the card, you HAD to get a review card (or whatever). It took advantage of the time, often substantial, that it took manufacturers to manufacture and fill the channel, prior to release.

Using "review" cards isn't necessarily a bad thing, but it's critical te the company plays fair. Nearly all used to, but I'm not so sure that's still true. At least part of the problem is that if I was reviewing for a major publication, they couldn't pressure me with threats - I'd just review someting else in that slot. There were occasions I'm awae of that a company might try to pressure a publication, only to be told, essentially, to take a hike. If they wanted their products reviewed, they needed to supply review samples. If not .... okay. Bye.

But now, so many "reviewers" are either small companies or even independent individuals, not multi-national publishers and it's potentiaaly much easier to cajore, threaten or blackmail them. Which is why a number of the really big reviewers, like LMG et.al. jump to the defence of small channels if tey get too obnoxious.

There are companies, though, known to hold grudges ... for examples, this week's WAN Show mentions a couple, so I won't.

In short, though - if you want to review a product in time for the review to do well, it comes from the press office, pretty much 100% of the time. Most are fair about it, but not all. Buying you "review" cards retail is not a good idea for a timely review, especially for high-demand product. If you do that, by the time you get it, spend a sensible amount of time checking it, do your tests, write it up, record a video and get it published, all the attention has moved on because everybody else did theirs several weeks ago. Just don't review products from companies that try to push you about.



(*) Or, don't have coments I want to put into print, even if I might have a private opinion.