Rumor: AMD allegedly to release seven NAVI GPUs but faces efficiency issues
Click here to post a comment for Rumor: AMD allegedly to release seven NAVI GPUs but faces efficiency issues on our message forum
RzrTrek
Another round of frying pans, why am I not surprised... but I won't throw a tantrum as I've grown used to it with my HD 6970 and RX 580, but it's very disappointing news.
Mesab67
This is a long overdue common sense injection to a rapidly growing hype train. Reading the various updates over the last months it's not too hard to see a pattern of silence with Navi. A recent update suggest we're still on a GCN base with, crucially, it's known delivery characteristics. While reduction + fine-tuning will go only so far I'd be pretty confident that they'll be priced competitively. These same updates also suggest a new architecture is in the works - I think most of us hoped Navi was that new architecture. Companies should monitor and correct public misinformation a damn sight sooner in these cases where a negative impact is guaranteed.
I'm still really not keen on supporting Nvidia after the RTX price gouging fiasco - that raised benchmark is still there. Remember also the recently released but 'dead on arrival' 1650 (beaten by AMD's 3yr old RX570 (i.e. 470) on both performance and price). Greed and underhandedness do not sit well with any company and I'll repeat, they're in a enviable position where they do not need to display both. Instead of a handful of individuals thinking about lining their own pockets, buying new leather jackets, they should think about correcting the company's image. Damaged it is.
FrostNixon
The 3080 xt sounds like a deal if the price is like this.
Evildead666
This is what I was expecting, but on the lower end of the scale.
Like one of the worst outcomes.
Its 7nm. Yes its expensive process, but yields are good, just perf/watt is nowhere to be seen.
They can and will sell all they will produce, because they can drop the price margins to the max, up the TDP, and just sell what is needed.
The APUs will take a lot longer, so will the mobile chips, which could push back their release date until the perf is good, if its possible.
Maybe they'll be lucky because they wont clock so high ?
I can understand the engineers want done.
They want to get going with the next new GPU tech for the 3-5 years to come, not polishing the last of the current generation.
I'd guess that the chiplets won't be that far down the line (famous last words).
The IO die is common among CPUs and GPUs, the tech and process well known.
Cost and yields will be even better, when they get chiplet gpus out. You'd think the chiplets would be easier to design and test, would't you without a whole load of IO stuff on-die ?
I'm actually waiting on a good AMD gpu to build a complete new rig.
Looks like i'll be waiting a bit longer.
Kaarme
The very first rumours of Navi said it's going to be mainstream oriented, and the scarce info from AMD seemed to support that guess. Seeing how you sport a Vega 64, your chances of finding a new GPU, with lots of more muscles, among the Navi models was thus low from the beginning if we heed any of the speculation whatsoever. I'd say even if all the worst rumours turn out true, for someone like me with a dusty old video card, Navi could still be quite plausible. I just hope they won't be limited to 32 ROPs within my price range.
screwtech02
This appears promising, as long as there is no anal-fleecing/price gouging when they are released.... Enough of that on NVIDIA side now as it is...
Evildead666
I went from a 290 to the Vega but I was somewhere hoping that i could get a new CPU, and maybe add a Navi if it was equal or better than Vega64, but was a lot less power, and cheaper.
Then i could wait a year or so for the next big AMD card. with chiplets 🙂
7nm is supply constrained as well, so i suppose that might work in AMDs favour if the CPUs are in high demand...
schmidtbag
I think the TDP numbers seem pretty normal (especially considering how AMD seems to over-volt everything) but something seems a bit off with those performance numbers. Ideally, most of the "Performance" column would be shifted down 1 row, with the 3090XT being equivalent to the RTX 2080.
lol yeah... I think I'm more bothered by the naming scheme than the TDP, price point, or rumored performance. Sure, Nvidia's 1600 series (as a product name) makes no sense at all, but at least it isn't a blatant ripoff. AMD seriously needs to stop trying to one-up their competitors in product names.
Yakk
AMD just concentrating on high efficiency mainstream pc parts to complement their gaming consoles should be more than good, don't waste factory overclocking GPUs past their peak efficiency ratings just for benchmarks. Also helps that AMD can basically dictate the pace of game development with studios.
jbscotchman
If that chart is correct, then this will be good competition for Nvidia.
dfsdfs1112
Why no one registers if there is HDMI 2.1 ?????????????????????????
schmidtbag
Probably because at this rate HDMI 2.1 is implied, so if they didn't include that, they'd really start pissing people off. I think we'd all be shocked if the 3070+ didn't include HDMI 2.1.
dfsdfs1112
They have promised PCI-E 4.0
Without HDMI 2.1 it would be a bad joke
slicer
If NAVI is somewhat more powerhungry than previously calculated, then I bet that is only the case for the Navi20 series (high end). Same was with Polaris. If RX480/RX580 are operating in its efficient voltage range, then it is super competitive, but as soon as you try to raise the gpu clock, then it goes out of balance (+30% more power for only 15% more performance or even worse). I bet this is the case for higher end NAVI's and thats why they are delayed.
But lower end Navi cards will be very competitive, thanks to the gddr6 usage now. No HBM2 memory, which already added like 160$. So now think of Vega64 performance with 8GB gddr6 memory (-80$ compared to HBM). But at the power is down almost 45% and price is 250 euros!
I would buy it!
SniperX
Yeah, this makes no sense. Why would Navi 20 (225W TDP) have efficiency and heat issues when Vega 20 (300W TDP) have none such issues, considering they both using the same 7N process and have the same amount of compute units? If true, then Navi is using way more power than the suggested TDP.
vMax1965
I have to admit I find this leak very odd indeed and as to price gouging fiasco with the RTX 2080, the same can be said for the launch price of the Vega 7 at $699! As to the other current GPU's, AMD can only compete on price. What concerns me is the pricing as I just cannot see any business sense why AMD would be willing to lose money. They will price according to the competition as proved by Vega 7 already. And it is important to understand that AMD is a business and have to drive profitability especially on the GPU side where they need to up there margins and there customers will pay the price so long as they are getting parity or better against Nvidia.
The real worry for AMD is Nvidia 7nm! AMD need a decent launch in 2019 as the GTX 1080Ti at 16nm can hold it's own against the 7nm Vega 7 right now! Nvidia's 7nm launch will be primed to drive not only performance but efficiency when the Turing respin drops to 7nm. The Hype really worries me as we do need a strong AMD GPU to compete with Nvidia but sometimes I read these stories and I just wonder how bad the let down will be if they keep hyping to the clouds..
Reddoguk
I like the 3080 XT. Good bang for buck if those prices are right.
schmidtbag
You mean Radeon 7 (as opposed to Vega 7)? It's launch price was so high because it's basically just a binned workstation GPU with a lot of very expensive memory. I'm sure the price point wasn't due to matching Nvidia's performance, but because they needed to cover the cost of the HBM. I'm confident that if they gave it 8GB of GDDR5/6, it would've been $200-300 cheaper (at which point, it would have probably sold pretty well since it'd basically be a 2080 without RTX).
EspHack
430$ for a 2080 equivalent seems mighty good, maybe not as crazy as some think, vega VII is like 300$ worth of vram so this thing using GDDR6 and being a bit more efficient should do the trick, provided AMD expects no profit
Nima V
AMD will never abandon GCN architecture. these days most games are optimized for consoles aka GCN architecture and because of that AMD is enjoying a huge advantage over Nvidia specially on games that run on low level APIs (DX12, Vulkan). by leaving GCN, AMD must give up on this free privilege which they will never do. just check out how older games from last gen consoles or games that are not on consoles run on GCN compare to Nvidia equivalent GPUs. most of the times they can't even compete.
