Rumored specs for NVIDIA's upcoming "Ampere" RTX 3000 series graphics cards (RTX 3060, RTX 3070, RTX 3080 and RTX 3090)
- Thread starter Deleted member 3812
- Start date
- Rumor
- Status
It's also not realistic to draw that much power 24/7. My machine idles at <100W most of the time. It jumps to 400W while gaming maybe. Even at the ridiculous price of US$0.20/kW that's only $175/year idling.
God I hope it's not another i740. What a disappointing turd that was even in the discrete version.
I want a new GPU, I want to enjoy the summer at the same time I'd like to know what Nvidia and AMD had to offer.
Also all the rumours about Ampere consumption are a bit scary and also weird coming from Nvidia recent standard.
Looking at my 550 gold PSU, I hope I'll not have to buy a new PSU too.
Also all the rumours about Ampere consumption are a bit scary and also weird coming from Nvidia recent standard.
Looking at my 550 gold PSU, I hope I'll not have to buy a new PSU too.
Yeah, absolutely, it's not running full blast all the time. IMO we should all still try to be more sustainable.
Is Ampere expected to bing improvements to hardware video encoding/NVENC?
I found it surprisingly good with my RTX card compared with x264 or Intel QuickSync.
I found it surprisingly good with my RTX card compared with x264 or Intel QuickSync.
I fully expect Nvidia to keep the same prices from the 20 series.
Nvidia MSRP prices are BS though so expect to pay at least $100 - $200 over whatever they announce
Does the rumored 300+ wattage/tdp mean these cards are gonna run stupid hot? My 180w GTX 1080 already increases my room temperature a few degrees when under heavy load so I'm really worried the 3000 series is gonna make my room unbearable :/
Glad I opted for a 750w Platinum PSU on my build. It felt like overkill at the time, but who knows.
Should probably wait for the partner cards with extra cooling to come out then. Thats what i'll probably do.
Edit: reading comprehension is not my strong suit....
Last edited:
Same here but a 750W Gold. I went with it specifically for this reason: if wattage requirements go up I don't want to have to think about it.
The amount of heat generated will be the same no matter what kind of cooler is used. A better cooler just moves more heat from the card into the ambient environment.
That's not how that works. The card will still be using the same amount of power. A better cooler will only make the card run cooler
Another 750W Gold here, only because it was available. For 13 years I've had 600W-650W for my computers.
I also have gone with 600 W in the past but in my last build a year ago something made me feel as though 750W was a good idea.
I bought my Seasonic 750W platinum just because it was the first PSU that I found that was available at MSRP and was highly rated. There were some 500W options and non-modular options but I didn't want to deal with the wires. The PSU shortage right now is very real. Also it came with a 10-year warranty so I figured that the extra cost would be made up if I use it in at least one more build in the future.
I'm worried about fan noise if it's that high. My 2080 with a 3 slot HSF already can easily run 80C depending on my fan curve, and it's already noisy at that level.
absolutely correct, i was just correcting the statement that its not happening *at all*, and if you have an HEDT cpu like a threadripper gen 3 part, then yeah 250w will be the norm... during multithreaded productivity loads, gaming power draw on basically all of these cpus is like.... 50 - 90w its not even worth caring about in this context
I read Aqua's post wrong. I took it as Aqua was worried about the computer getting too hot and not the room.
Derp.
If $699 that's not bad. Pretty much the same core count as the 2080 Ti. So basically getting a much improved 2080 ti with better RT and faster memory (although losing 1gb), for quite a bit less money.
We'll see if they hit that price point. Hope the 3070 can hit $500 as well.
We'll see if they hit that price point. Hope the 3070 can hit $500 as well.
LOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOL in your dreams.
Mid range cards at launch have not been around 500 for a few gens now sadly :(
No retailer had them for that price.
20XX series was sold out all the time or priced up by a few 100 for months.
I dunno rumour is they are using two different nodes (7nm TSMC and 8nm Samsung) for gaming and pro cards. So it doesn't sound too implausible to me, but yeh maybe the word 'refresh' is wrong.
Would be funny a couple of years ago but with Navi 21 this may well be an option finally.
Which is also why people expecting the doubling of 2080 price or something are weird.
I hate that these insane graphics cards prices have been normalized. I don't think the price of the 2080 Ti was fine to begin with.
Yep, exactly. The real question is going to be the pricing, I'd prefer they lower the pricing across the board but if not they at least need to not raise the price. With Turing they had an excuse since they had monster size chips due to not having a node shrink.
That's fair. I was really more focused on the performance gains between generations. As long as it's close to the gains of Pascal, I'd mark that down as a success. It would be nice for pricing to come down especially with Nvidia expressing some disappointment with initial RTX sales.
The problem fof me is not the PSU failing, or being on the edge, which would be unusual cases anyway.
It would be the heat and thus noise generated from a high power card as is rumoured paired with a very hot and power hungry CPU (9900k or 10900K). I am past dealing with a noisey system and I'm past dealing with my room heating up after a gaming session, I thought this was all in the past!
There is a certain joy from gaming on a cool and quiet system.
I definitely agree that doubling the price would be suicide. But beyond that, I have zero expectations on pricing. You can't tell me Jensen Huang, and his niece, Lisa Su, haven't had conversations about this. The GPU market is a niche and hardcore market. Price wars don't lead to growing the pie. It may result in market share moving a couple percentage points here or there, but ultimately it destroys the margins for both without bringing in many new customers. I expect Nvidia will charge the most it thinks its customers are willing to pay, and AMD will undercut that pricing slightly.
AMD it's always "wait for x..." where x has no real competition against Nvidia's top end.
I guess it's what the death or Moore's law looks like. We're still getting good gains in performance, but we're going Back to the Future in terms of power draws (and their ability to heat a room up).
What I wouldn't give to know the new Xbox's power draws. Have they truly made headway here? Or are the new consoles going to be power monsters. Digital Foundry would never do it (since it would kill their relationship with MS), but it would have been amazing if they would have plugged their test machine into a Kill-a-Watt device. This would give us some indication as to whether AMD's new GPU architecture has improved any in this area.
It is not always like that , before series 900 AMD was a real competitor in this market. 5700 XT showed they can compete again.
It was the same thing about CPUS before Ryzen and see how it is now the market...
If they delivery a GPU 50% more powerfull than 2080 TI they are in the game, even 40%
I'm not sure but I think the progression of cards today it's more of a don't upgrade every generation, maybe every other generation. I don't see the 20 series to 30 series being substantial, however, if you are in the 10 series or lower the 30 series should be a substantial upgrade.
'Technically, Lisa Su's grandfather is Jen-Hsun Huang's uncle. They are not exactly niece and uncle, but close relatives.'
That was for the FE
Yeah but because Nvidia prices $100 above MSRP pretty much everyone else does too and most aftermarket cards are actually nicer than the founders edition as far as pcb and power delivery and some are 3 fans. Why would they not change the same price Nvidia is charging. 2070s launching at $500 even Nvidia's own model was effectively a price drop.
You couldn't find a $500 2070 for months after launch outside of maybe a blower model and even that I'm not sure on. Sometimes those aren't really even cheaper because sometimes people seek them out for sff cases. EVGA with their black series are pretty much the only vendor that stays close to the MSRP and those would sell out immediately for months.
The whole founders thing is skeezy in general. Just launch with one actual MSRP for your own actual product like everyone else does. Not hey MSRP is $499 but here is our $600 version board partners feel free to also charge $600! Like certainly board partners can charge more than MSRP but a reference card should be priced at the MSRP.
Why're you all flexing your 750W PSUs? Was some news revealed that would lead you to believe you'd need gigantic PSUs? I got a 650W and I was always told its oversized for single GPU rigs.
Thanks for the reply and especially thanks for pointing out what is very unlikely.
I guess its going to take a few additional years before GPU MCM or GPU Chiplets will become useful for games, and made easy to code for, for developers.
Having 2 or 4 separately manufactured pieces of silicon on a card, where they act as one huge GPU to developers is not an easy thing, and it's not hear yet, and probably won't be in 2022. Maybe 2024 ? These would also need to using HBM type memory, so save on power consumption.[/QUOTE]
Oh, we definitely shouldn't expect any of the two companies to gift their new GPUs for some peanuts or such. But considering that NV's top end will launch to a competition coming in a couple of months after that expecting them to increase the prices is equally unrealistic. This upcoming gen will be a big fight between the two which means that anything above $1000 is likely out - if only because either would be forced to introduce a much cheaper card with essentially the same performance (think 1080Ti vs GTX Titan and such). With NV opting for a supposedly cheaper Samsung 8nm process they seem to be fully aware of what's coming and are trying to protect their margins this way which means that the prices won't go up and have a good chance of actually coming down from where they were with Turing and RDNA1.
Navi 21 is top end though and it should launch this year so there's that.
It's... going to be more then some years I think. Chiplet design for GPUs will prompt for a total rebuild of how games are being programmed and the likely dumping of rasterization h/w in favor of something more natively parallel. Won't happen at the beginning of a new console gen which is for all intents and purposes very old school in its GPU approach. By the end of this gen though or at the start of the next one, when we'll be testing 3nm processes which are very likely to be the end of scaling on silicon - well, maybe. But that's 2025+ at best I think.
There will likely be some HPC / DL testing for chiplet designs though, we've already seen the research on that from several IHVs and it's very probable that we'll see some products designated for pure GPU compute markets much sooner than anything like this will come to gaming.
Depending on which card you go with in the Ampere lineup, your PSU might need to supply your GPU alone with between 300-350W, more if you overclock it.
This is the one area where I don't expect a major difference between Big Navi/RDNA 2 given how power hungry the 5000 RX series cards were, and I say that as someone who uses a 5700 XT in a mini build.
So maybe, just maybe, Chiplet design for GPU might be viable in time for PS6 and the equivalent Xbox assiming they are released Holiday 2027?
Or is even that too optimistic to think? (I know it's WAY to early to know).
You should be fine but I would have recommended waiting to see what NVIDIA and AMD have to offer in a few months time, if only to see how much better ray-tracing performance is on either company's cards.
depends on the resolution you game at. I have a 2070 super as well and, at 1440p, certain titles from 2016 and 2018 are kind of tough. barely getting 60fps in ass creed odyssey and going through areas in Deus Ex Mankind Divided where the fps drops to high 40s/low50s is kind of wild for a $560 2018/9 gpu imo
I mean I got it like a year ago (and then the first one died and I had to get a replacement, naturally) so I do think I did ok with the info I had at the time but I guess I'm getting another one in like, a year and a half, which blows
- Status