AMD R9 Nano compared to a pencil!

[ZetZet]

as of now we don't know what the power delivery is like and that is what overclocking is pushing it past rated limits. people do this all the time with almost every GPU on the market so why can't we do it on the nano?

 

and that is why I was stating I want to know. 

We do know. It has one 8 pin. 

 

 

So 225W max. Fury X uses two 8pins. So 375W. And it eats around 300W at full load. http://www.guru3d.com/articles_pages/amd_radeon_r9_fury_x_review,10.html

[Megahurt]

PcPer showed theirs on their podcast wednesday, and again yesterday during the logitech event. I expect it to perform the same as a fury, and amd probably doesn't care about it being expensive as they already have problems with supply, it will most likely drop in price in the coming months.

[The Benjamins]

We do know. It has one 8 pin. 

 

 

So 225W max. Fury X uses two 8pins. So 375W. And it eats around 300W at full load. http://www.guru3d.com/articles_pages/amd_radeon_r9_fury_x_review,10.html

 

the FuryX PLUGS are rated for 375W but the power delivery is rated at 500W. a portion of overclocking it getting the power delivery AND plugs to run higher then spec. the high overclock on 980ti range from 300-400W way past its rating on plugs and power delivery. 

 

 

FuryX power delivery can handle 400 amps at 1.2V which is 480W

http://www.pcworld.com/article/2937335/components-graphics/behold-the-beast-full-amd-radeon-r9-fury-x-tech-specs-and-design-details-revealed.html

[KeltonDSMer]

Well I first I considered that and thought it was a bad choice as well but since, I came up with an idea why they're doing that: They just don't have enough production to supply chips for both reference and non-reference. They wanna get their reference design out first once they get better yields they will probably let em do whatever they want.

 

If you think about it is also why the Fury nonx does allow non reference design: they can take all the not so great yields from the full Fiji and run those to the Fury nonx

 

Hmm, I like the thinking you have going on there. That is a better hypothesis than anything I could come up with. Maybe Fiji's 2nd itteration as a 490x, or whatever they end up calling it, will show us what all those cores can really do (maybe with 8GB HBM2).

[ZetZet]

the FuryX PLUGS are rated for 375W but the power delivery is rated at 500W. a portion of overclocking it getting the power delivery AND plugs to run higher then spec. the high overclock on 980ti range from 300-400W way past its rating on plugs and power delivery. 

So maybe it will overclock close to fury x (doubt it), which is still pointless unless the price drops like 300$ below fury x since you will need a water block to cool it.

[MEC-777]

Why are some expecting this to cost less compared to the Fury x (or even Fury, wtf)? It is a full spec Fiji core, of course it will cost more than the Fury. By that logic, a small 980 should cost less than a 970!

 

Your logic is a little flawed. A small 980 that does not perform as good as a big 980 should cost less than a 980, not 970. We expect the Nano to cost less than the Fury X because it lacks the power delivery and cooling capability necessary to match and sustain Fury X performance levels. 

 

If it could match Fury X performance like the mITX 970's can match full-sized 970 performance - then I would say it's worthy of the same price tag or a slight premium. But the fact is; (based on what we know thus far) it can't and probably won't. 

 

EVGA goes ahead and charges up to an extra $300 for a top binned 980ti, no problem there. AMD bins some full Fiji cores for certain characteristics, sells them for the same price as Fury x in a cute little package,  and that is crummy for some reason?

 

I thought all that mattered these days was perf/watt... now that AMD has likely narrowed the gap with their underclocked mini Fiji, no one seems to give a shit.

This is not the same situation. EVGA uses those top binned 980Ti chips in their highest performance cards. Not for small form factor cards. 

 

What AMD is doing is releasing a slower card for the same price as a faster one. How much slower remains to be seen just yet and I don't trust benchmarks from AMD, alone.

[Misanthrope]

 

Your logic is a little flawed. A small 980 that does not perform as good as a big 980 should cost less than a 980, not 970. We expect the Nano to cost less than the Fury X because it lacks the power delivery and cooling capability necessary to match and sustain Fury X performance levels. 

 

If it could match Fury X performance like the mITX 970's can match full-sized 970 performance - then I would say it's worthy of the same price tag or a slight premium. But the fact is; (based on what we know thus far) it can't and probably won't. 

 

This is not the same situation. EVGA uses those top binned 980Ti chips in their highest performance cards. Not for small form factor cards. 

 

What AMD is doing is releasing a slower card for the same price as a faster one. How much slower remains to be seen just yet and I don't trust benchmarks from AMD, alone.

 

Less power consumption is not always an indication of a worst chip: Maxwell generally uses a lot less power than Kepler yet they were priced at a higher level because they still performed better. It can potentially be a more expensive or better yielded Fiji chip to be able to scale down dramatically on power and still perform within reason. Intel does this with their architecture with slower chips using a lot less power being more expensive than higher performing chips at a higher TDP. Nvidia does it as well as I just said. But if AMD does it, well damn if they do (omg 290x for them winters!) and damn if they don't (Extremely more efficient chip? OMG SO EXPENSIVE WHY ISN'T IT CHEAPER THAN THE THING WE WERE JUST SAYING USES TOO MUCH POWER!)

[The Benjamins]

So maybe it will overclock close to fury x (doubt it), which is still pointless unless the price drops like 300$ below fury x since you will need a water block to cool it.

 

and that is what i want to know, we also don't know if the FuryX or Nano gets the better binned chips.

[MEC-777]

Can we all please stop calling it "Fury non-x" please? It's just "Fury". 

 

It's unnecessary. If people assume you're talking about the Fury X, it's their own fault for not reading more carefully.

[Misanthrope]

Can we all please stop calling it "Fury non-x" please? It's just "Fury". 

 

It's unnecessary. If people assume you're talking about the Fury X, it's their own fault for not reading more carefully.

 

Yes we actually could, but I do it more as a perpetual reminded of how fucking stupid AMD naming schemes are.

[The Benjamins]

Can we all please stop calling it "Fury non-x" please? It's just "Fury". 

 

It's unnecessary. If people assume you're talking about the Fury X, it's their own fault for not reading more carefully.

 

can we call the nano " FurX with a smaller PCB and air cooled and lower TDP" you know to shorten it.

[ZetZet]

can we call the nano " FurX with a smaller PCB and air cooled and lower TDP" you know to shorten it.

fury cripple x

[Misanthrope]

fury cripple x

 

[cesrai]

Meh.

[MEC-777]

Less power consumption is not always an indication of a worst chip: Maxwell generally uses a lot less power than Kepler yet they were priced at a higher level because they still performed better. It can potentially be a more expensive or better yielded Fiji chip to be able to scale down dramatically on power and still perform within reason. Intel does this with their architecture with slower chips using a lot less power being more expensive than higher performing chips at a higher TDP. Nvidia does it as well as I just said. But if AMD does it, well damn if they do (omg 290x for them winters!) and damn if they don't (Extremely more efficient chip? OMG SO EXPENSIVE WHY ISN'T IT CHEAPER THAN THE THING WE WERE JUST SAYING USES TOO MUCH POWER!)

 

Maxwell and Kepler are two different generations of chips. The Fury X and Nano are the exact same chip, right down to the core count. So it's not really the same comparison. If you're comparing the full Kepler to the full Maxwell, then you have to compare Hawaii to Fiji. In that context, both managed efficiency improvements, but Nvidia came out on top in this last round. There was so much hype built around Fiji and the fact is, it just wasn't quite enough. It's great that AMD made these improvements - most would agree, but it goes to show what's most important in the end which is outright performance.

 

For the record, I'm not saying the Nano should cost less than the Hawaii XT (390x). I don't think it should. But I am suggesting it should cost less than the Fury X. Maybe closer to the price of the Fury or a bit less. Depends on where the performance actually falls. AMD is claiming the Nano is slightly better than the 390x so I think pricing it close to the Fury is reasonable ($500-550). You pay a little premium for the form factor but not so much that it's unreasonable for the performance. 

[MEC-777]

Yes we actually could, but I do it more as a perpetual reminded of how fucking stupid AMD naming schemes are.

 

It wouldn't be a problem if everyone would just use the proper names.  

 

It's one letter. Not that hard to understand.

[Sauron]

I'd buy it for $500 max.

 

that's the price of a 980 and this should wipe the floor with those - the fury does and this one is supposed to be faster iirc. If the fury were 500 bucks and this 550, then we'd be talking.

[AlwaysFSX]

AMD needs to sell this for less than $650...

I know they're trying to make profit but that's too much.

AMD has to charge this much otherwise they'll never recoup any money they've set on fire.

[KeltonDSMer]

 

Your logic is a little flawed. A small 980 that does not perform as good as a big 980 should cost less than a 980, not 970. We expect the Nano to cost less than the Fury X because it lacks the power delivery and cooling capability necessary to match and sustain Fury X performance levels. 

 

If it could match Fury X performance like the mITX 970's can match full-sized 970 performance - then I would say it's worthy of the same price tag or a slight premium. But the fact is; (based on what we know thus far) it can't and probably won't. 

 

This is not the same situation. EVGA uses those top binned 980Ti chips in their highest performance cards. Not for small form factor cards. 

 

What AMD is doing is releasing a slower card for the same price as a faster one. How much slower remains to be seen just yet and I don't trust benchmarks from AMD, alone.

 

I hear what you are saying. My "by that logic, a small 980 should cost less than a 970" statement was simply in response to a few comments suggesting the Nano should cost less than a Fury; I find that ridiculous because Nano has fully unlocked shaders. A full fat chip is always going to be more expensive than one that has been saved from defect by being cut down, especially considering how freaking huge Fiji is and the inevitable reduction in yields that come along with that.

 

I think the EVGA comparison is fair. Both companies are binning chips for specific characteristics, and one is charging for it. Of course the Nano is slower than Fury x, I get that. I am arguing we are trading clockspeed for a chip binned for efficiency and for form factor, but not paying extra for it. The ASIC quallity binning is pretty much exactly what AMD is doing, but they are using the low leakage chips for efficiency purposes.

 

Again, non of this would really matter if AMD simply allowed AIBPs to do what they do. We have no options for air cooled Fury x cards and so if you want all 4096 Fiji cores, the two options are a CLC card or Nano with 100W less to work with. If a "regualr" air cooled Fury x was available in a few different versions from various board partners, people might be able to appreciate what the Nano has to offer. Because of lack of choice, we are left arguing over the pros/cons of a niche card in scenarios it was never designed for. Nano is not priced according to where it fits in overall performance, but is judged that way because there are no other options.

[MahtXL]

I'd buy it for $500 max.

well hows it stack against the msi r9 390x gaming g. because thats a 600 dollar card up here in the north.

[Misanthrope]

Maxwell and Kepler are two different generations of chips. The Fury X and Nano are the exact same chip, right down to the core count. So it's not really the same comparison. 

 

It's not the comparison being made which is Hawaii or Grenada vs Fiji. That's AMD's previous generation. Why would you or anyone compare same generation cards? Maxwell is pretty damn efficient but the flagship cards still run at 250 watt TDP

[LucidMew]

IMO, the reason why it costs so much is because AMD is binning for power efficiency, just like Intel is doing with a 2C4T, 4.5W chip that barely boosts to 2.7GHz that costs nearly $300 http://www.pcper.com/reviews/Processors/Intel-Launches-Skylake-6th-Generation-Core-Processors

[MEC-777]

I hear what you are saying. My "by that logic, a small 980 should cost less than a 970" statement was simply in response to a few comments suggesting the Nano should cost less than a Fury; I find that ridiculous because Nano has fully unlocked shaders. A full fat chip is always going to be more expensive than one that has been saved from defect by being cut down, especially considering how freaking huge Fiji is and the inevitable reduction in yields that come along with that.

 

I think the EVGA comparison is fair. Both companies are binning chips for specific characteristics, and one is charging for it. Of course the Nano is slower than Fury x, I get that. I am arguing we are trading clockspeed for a chip binned for efficiency and for form factor, but not paying extra for it. The ASIC quallity binning is pretty much exactly what AMD is doing, but they are using the low leakage chips for efficiency purposes.

 

Again, non of this would really matter if AMD simply allowed AIBPs to do what they do. We have no options for air cooled Fury x cards and so if you want all 4096 Fiji cores, the two options are a CLC card or Nano with 100W less to work with. If a "regualr" air cooled Fury x was available in a few different versions from various board partners, people might be able to appreciate what the Nano has to offer. Because of lack of choice, we are left arguing over the pros/cons of a niche card in scenarios it was never designed for. Nano is not priced according to where it fits in overall performance, but is judged that way because there are no other options.

 

Generally speaking, people don't buy GPUs based on their number of shaders etc. They buy based on performance and where they sit in terms of price to performance. Nobody is going to buy the Nano specifically because it has the full 4096 cores, they will buy it because it is simply the fastest card available in this form factor. You got it right in your last sentence.

 

I don't have a problem with it being priced with a premium, but IMHO I think they're aiming a little too high in this case and it may do more harm than good. It doesn't matter if it's the full Fiji or not. What matters is how well it performs and the fact that it fits a certain niche market.

 

It's not the comparison being made which is Hawaii or Grenada vs Fiji. That's AMD's previous generation. Why would you or anyone compare same generation cards? Maxwell is pretty damn efficient but the flagship cards still run at 250 watt TDP

 

My response was to the person I quoted, who was comparing Kepler to Maxwell against Fiji XT (Fury X) to the "binned" Fiji XT (Nano). I agree the proper comparison is as you've just mentioned, which is what I pointed out earlier.

[Notional]

Can we all please stop calling it "Fury non-x" please? It's just "Fury". 

 

It's unnecessary. If people assume you're talking about the Fury X, it's their own fault for not reading more carefully.

 

Nah we should start calling the 980 the 980 non ti. Just to spread the stupid

 

That being said, a R9 Nano is the highest performing, smallest card on the market. Of course that comes at a price premium.

[ICantThinkOfAnyGoodName]

Yes we actually could, but I do it more as a perpetual reminded of how fucking stupid AMD naming schemes are.

How is AMDs naming scheme stupid for having a Fury X and Fury when Nvidia has a Titan X, Titan Z and a Titan?