Fury Nano, biggest bang in smallest package.. FPS/inch baby!

[raghu78]

A Nano priced the same as a Fiji Pro doesn't make sense to me, even for an efficiency premium. Size is less relevant on pricing than it should be...even my uATX N200 case houses a behemoth 7870 that rivals triple-fan coolers in length.

The Nano interests me most for the power envelope. I CAN run a 390X or R9 Fury on my setup but I don't really want to tax my power supply that much.

It's a touchy price segment. The way I see it, a Nano would have to fall into the 450-500 range to be competitive. I think it'll end up being $479 but lower would be nice. Any higher than $500 and I'll just end up with a 980 instead.

the Fury with 3584 sp and 1 Ghz core clock was just 7% slower than Fury X which had a 50 Mhz higher clock and 512 sp more. ASUS Fury Strix with a good power efficient design was also quite good at perf/watt though not as good as Maxwell.

http://www.techpowerup.com/reviews/ASUS/R9_Fury_Strix/31.html
http://www.techpowerup.com/reviews/ASUS/R9_Fury_Strix/29.html

So I can definitely see a 3072 sp with 925 Mhz and binning for lower leakage to easily hit 175W TDP and perform around R9 390X. With a 6 inch form factor and a 175W TDP all we need to see is how are the GPU temps and fan noise. If AMD can keep core temps below 75c and VRM temps below 95c and also fan noise comparable to Fury custom cards, then the R9 Nano would be an impressive card. A price around USD 450 would be very good price.

 

[Silverforce11]

You take Fury X, you give it -50% power limit in the CC overdrive, it's now ~175W TDP and about 11% slower.

Look at the Toms article when they did exactly that.

power consumption decreases from 267W to 170W when the power limit is set to -50 percent

http://www.tomshardware.com/reviews/amd-radeon-r9-fury-x-power-pump-efficiency,4215.html

FPS drop from 39.7 to 36 in Thief maxed 4K.

They didn't test more games, but from my own -50% & -25% power settings, the results are almost identical across all games.

Basically pretty easy for AMD to make a power efficient Nano. They already have all the tools available for them, all they need is just a bios with a lower TDP limit like ASUS Fury and GCN will adjust its clocks dynamically to match the target.

 

[Pinstripe]

You take Fury X, you give it -50% power limit in the CC overdrive, it's now ~175W TDP and about 11% slower.

Look at the Toms article when they did exactly that.



http://www.tomshardware.com/reviews/amd-radeon-r9-fury-x-power-pump-efficiency,4215.html

FPS drop from 39.7 to 36 in Thief maxed 4K.

They didn't test more games, but from my own -50% & -25% power settings, the results are almost identical across all games.

Basically pretty easy for AMD to make a power efficient Nano. They already have all the tools available for them, all they need is just a bios with a lower TDP limit like ASUS Fury and GCN will adjust its clocks dynamically to match the target.

Slight OT, but I always wanted to know what's the more efficient way to reduce TDP without crippling game performance too much: Clock rate or Power Limit? Are these two settings directly interlinked? Say, I have a R9 290 and want to reduce clock rate from 977Mhz down to 850Mhz to limit some power consumption, or is it better to keep the clock rate and instead reduce Power Limit by -20% to sustain acceptable framerates?

 

[Silverforce11]

Slight OT, but I always wanted to know what's the more efficient way to reduce TDP without crippling game performance too much: Clock rate or Power Limit? Are these two settings directly interlinked? Say, I have a R9 290 and want to reduce clock rate from 977Mhz down to 850Mhz to limit some power consumption, or is it better to keep the clock rate and instead reduce Power Limit by -20% to sustain acceptable framerates?

Just go via power limit, GCN is dynamic and will try to reach the highest clocks possible whilst fitting the TDP limit.

Basically if you set -50% power limit, GCN loses around 11% performance while power usage drops ~40-50%.

I did it for mining and found it works so well, I do it now for general gaming.

More info:
http://forums.anandtech.com/showpost.php?p=37596583&postcount=136

 

[Glo.]

I think Fury Nano is Full Fiji chip with 175W TDP bios cap.

What that means is that the GPU will over or downclock the core depending on thermal and power situation. What that means is we can see Full Fiji at 1000 or 900 MHz depending on game, work or power supply.

 

[tviceman]

A Nano priced the same as a Fiji Pro doesn't make sense to me, even for an efficiency premium. Size is less relevant on pricing than it should be...even my uATX N200 case houses a behemoth 7870 that rivals triple-fan coolers in length.

The Nano interests me most for the power envelope. I CAN run a 390X or R9 Fury on my setup but I don't really want to tax my power supply that much.

It's a touchy price segment. The way I see it, a Nano would have to fall into the 450-500 range to be competitive. I think it'll end up being $479 but lower would be nice. Any higher than $500 and I'll just end up with a 980 instead.

It's tough to see from either point of view. Ask yourself this: how can AMD sell a full fat Fiji at ~87% speed and 75 less watts power consumption for $150 less without undermining the sales of the more expensive full fat Fiji? People aren't going to spend 33% more to get ~12% more performance.

Nvidia bins GM204 chips and sells the chips that can run at lower voltages as mobile GPU's with a price premium. Do you think AMD is going to bin full functional golden Fiji dies and sell them at a significant discount? That makes no sense whatsoever to me.

 

[Headfoot]

Slight OT, but I always wanted to know what's the more efficient way to reduce TDP without crippling game performance too much: Clock rate or Power Limit? Are these two settings directly interlinked? Say, I have a R9 290 and want to reduce clock rate from 977Mhz down to 850Mhz to limit some power consumption, or is it better to keep the clock rate and instead reduce Power Limit by -20% to sustain acceptable framerates?

The best way is to undervolt as much as possible without dropping clock rate. You keep the same speed but drop power consumption. Then use Power Limit because it will limit power to a set TDP while attempting to maximize clocks

 

[B-Riz]

Just go via power limit, GCN is dynamic and will try to reach the highest clocks possible whilst fitting the TDP limit.

Basically if you set -50% power limit, GCN loses around 11% performance while power usage drops ~40-50%.

I did it for mining and found it works so well, I do it now for general gaming.

More info:
http://forums.anandtech.com/showpost.php?p=37596583&postcount=136

+1

I do this also with my 290.

For the bit of gaming I do lately (Diablo 3), there are no ill effects.

Last time I played Battlefield 4, it was still smooth n nice, so I do think the 290 / 290x was over-volted from the get go.

 

[Madpacket]

You take Fury X, you give it -50% power limit in the CC overdrive, it's now ~175W TDP and about 11% slower.

Look at the Toms article when they did exactly that.



http://www.tomshardware.com/reviews/amd-radeon-r9-fury-x-power-pump-efficiency,4215.html

FPS drop from 39.7 to 36 in Thief maxed 4K.

They didn't test more games, but from my own -50% & -25% power settings, the results are almost identical across all games.

Basically pretty easy for AMD to make a power efficient Nano. They already have all the tools available for them, all they need is just a bios with a lower TDP limit like ASUS Fury and GCN will adjust its clocks dynamically to match the target.

Not sure why you're reading the Toms results this way. Look at the minimum framerates, not the average. Anything set to less than -20% power limit is very detrimental to the framerates here and as stated in the article:

"Even though the power consumption decreases from 267W to 170W when the power limit is set to -50 percent, the resulting frame rates just aren’t in the playable range any more"

The best way to minimize power consumption is to use binned chips, set lower voltage targets for them and run them at a reduced clock rates. It'll be interesting if they achieve this with the Nano but it looks totally possible if they're aiming for 290X speeds.

 

[Elixer]

Was in BB earlier today, and the stock people were saying they should be getting in "new AMD products" on Sept. 1st for the launch.
I assume this is the Nano?

I always read that the Nano would launch this month though. Hmm.

 

[tential]

Was in BB earlier today, and the stock people were saying they should be getting in "new AMD products" on Sept. 1st for the launch.
I assume this is the Nano?

I always read that the Nano would launch this month though. Hmm.

Me too.
It could still launch this month. Retail on the 1st but launched/announced now.

 

[Silverforce11]

Not sure why you're reading the Toms results this way. Look at the minimum framerates, not the average. Anything set to less than -20% power limit is very detrimental to the framerates here and as stated in the article:

"Even though the power consumption decreases from 267W to 170W when the power limit is set to -50 percent, the resulting frame rates just aren’t in the playable range any more"

The best way to minimize power consumption is to use binned chips, set lower voltage targets for them and run them at a reduced clock rates. It'll be interesting if they achieve this with the Nano but it looks totally possible if they're aiming for 290X speeds.

I don't know how long the min fps drop was, one-off or all the time? No data. My own testing show min fps and avg fps drop at the same rate, ~11%. There's no massive clock rate drops, it hoovers around 880Mhz vs 1Ghz.

The only data that's solid from Toms is that avg perf drop by ~11% and power drop by ~100W.

 

[bystander36]

The chart did seem to show minimums dropping off at a consistent rate.

 

[Shehriazad]

Let's not forget that there would likely be some difference between just power target of 50% and a chip that is actually made (on a firmware/binning level) to run at 175w.

I still expect a lot from this chip...and I stand by the statement that it could end up being AMDs most important product in their current lineup...if availability and price are a match.

 

[Silverforce11]

The chart did seem to show minimums dropping off at a consistent rate.

That's not consistent, that's a large drop off a cliff.

A time vs fps chart would show the issue, or it could be a one-off during a loading scene. That's lacking actual data required to interpret the results properly.

In all the games I've played, it behaves like a downclocked GPU should, max/avg/min fps is lower by a similar margin. But I don't have Fury X to test, just R290s and R290X. I also don't have Thief.

It could be Fiji behave oddly with a power limit with lower TDP but reviews of the ASUS Fury with its 213W bios (versus 275W default) did not show adverse min fps results, so I will have to conclude that the Thief result Tom's found, was an anomaly.

 

[guskline]

Sounds like a VERY interesting Card. However, need to see it in the field and tested. Let's hope Joe Macri chooses his praise "carefully"

 

[RussianSensation]

It's tough to see from either point of view. Ask yourself this: how can AMD sell a full fat Fiji at ~87% speed and 75 less watts power consumption for $150 less without undermining the sales of the more expensive full fat Fiji? People aren't going to spend 33% more to get ~12% more performance.

For the last 9 months, by far the vast majority of PC gamers did buy GTX980 over R9 290X for what 21% more performance for 80-90% higher cost? Talk about a horrible value but most of this forum hardly endorsed the 290X. In that context, your sentence of "33% higher price for 12% more performance" is a horrible deal suddenly seems like is a smoking deal, but I guess we are talking about AMD cards, not AMD vs. NV, right?

I am still trying to figure out how any R9 390/390X/Fury/980/Fury X make any sense at all when XFX R9 290X is $260 with lifetime warranty in the US. I know these prices are unattainable in other parts of the world but the US is a huge market for GPUs.

AMD might go wild and price Nano at $499 initially to milk the early adopter mini-ITX niche market. It seems they are now more interested in chasing profits than retaining market share which may or may not work for them.

I can't think of many reasons to buy any single GPU in the $300-600 range (unless one needs HDMI 2.0 or hardware decode/encode of GTX960) when R9 290X is $260. Just buy that and resell it in 18-20 months for 16nm HBM2. No matter what perf/watt Nano hits, it'll be completely outdated the minute high-end Pascal chips drop in 2016, same with Arctic Islands.

This is the strangest GPU market I've seen in a long time where R9 290/290X/970 and 980TI are the only segments that make sense. The $100-200 segment is full of crippled GPUs like 750/750Ti/960 or inferior perf/watt + somewhat outdated feature set ones like R9 270/270X/280/285//280X. The entire range of 980/Fury/Fury X don't make any sense to me when 290X is $260. I will chuck Nano into the same "overpriced" segment if it's priced at $450. So what that it uses 175W of power, it's still an awful deal compared to a $260 R9 290X. It'll take 10+ years to make up the difference in power usage. On top of that, XFX R9 290X is one of the quietest cards!

AMD is hardly going to win many customers because the type of buyers who buy AMD aren't going to be lining up in droves to pay $150-200 for barely more performance to save 100W of power.

I will continue with my view that this Fury/Fury X AMD generation is a stop-gap, no more. I can at least endorse GTX980Ti since it has 25-30% OCing headroom.

AMD could have hit it out of the park by replacing R9 290 with Fury as a spiritual successor priced at $399-429 but instead they jacked up the price to $550! At that point might as well pay $100 extra for the 980Ti.

The Nano is going to be super niche. This is AMD's last chance to make something big happen before back-to-school/Q4 2015. They should price it at $399 to shake up the market, and drop prices on R9 390/390X. If the Nano is priced too closely to the 980/Fury, it might be better to just buy the 980/Fury. With the 980 one gets the flexibility of great perf/watt and overclocking while the Fury's PowerTune can be reduced to hit perf/watt closer to the Nano but when one needs the extra power, just crank it up. I'll be pleasantly surprised if I am wrong and the market for $400+ mini ITX cards is far larger than I thought. I suppose if someone wants to built a 35W Skylake-T + Nano, it could make for a very small and yet powerful rig.

 

[Silverforce11]

RS, I thought you once said chasing marketshare by lowering price and margin didn't work too well for them. Because regardless of how good a deal is, the majority still go with NV anyway.

What makes you think NV users will switch and buy Fury if it was priced $100 less?

The R290X custom models were ~10% behind 980 and at ~half the price. NV users still bought 980s. It was faster than 970 and runs cool & quiet, often found for cheaper than the 970, and NV users still bought the 970.

Not to mention the custom R290s which were significantly cheaper and offered 95% of the R290X performance.

I did not think the 390/X would sell well at its jacked up price but I'm wrong there. It's selling very well from the etailers I've talked to.

Perception has changed. 390/X are quiet, cool running, offering very competitive performance for less. There's no stigma eventhough they are power hungry, they don't run hot or loud which suggests people hated on the R290/X not because of power consumption, but the combination of hot, hungry, loud.

Fury/X are sold out as well. There's no need to lower the price. AMD thinks there's a premium ($) to be had for water cooling. It seems the market agrees else nobody would touch Fury X when custom 980Ti are better at 1440p & below.

They don't need to lower the price until they can't sell them fast enough.

Nano would be killer at $399 but completely unrealistic. AMD wants a premium for form factor with that level of performance on offer at an excellent perf/w level. I think they will be right, the market will reward them because Nano would make for a killer mITX setup.

 

[atticus14]

The nano is the holy grail for small steam boxes that actually still have power. I can only think of 14nm/HBM2 cards though, when probably all cards could potentially have a nano version. It'll be interesting to if the small case market really takes off then. If i had plenty of disposable income I would build one or two just because it's fun to see so much power in a little box.

 

[railven]

RS, I thought you once said chasing marketshare by lowering price and margin didn't work too well for them. Because regardless of how good a deal is, the majority still go with NV anyway.

What makes you think NV users will switch and buy Fury if it was priced $100 less?

The R290X custom models were ~10% behind 980 and at ~half the price. NV users still bought 980s. It was faster than 970 and runs cool & quiet, often found for cheaper than the 970, and NV users still bought the 970.

Not to mention the custom R290s which were significantly cheaper and offered 95% of the R290X performance.

I did not think the 390/X would sell well at its jacked up price but I'm wrong there. It's selling very well from the etailers I've talked to.

Perception has changed. 390/X are quiet, cool running, offering very competitive performance for less. There's no stigma eventhough they are power hungry, they don't run hot or loud which suggests people hated on the R290/X not because of power consumption, but the combination of hot, hungry, loud.

Fury/X are sold out as well. There's no need to lower the price. AMD thinks there's a premium ($) to be had for water cooling. It seems the market agrees else nobody would touch Fury X when custom 980Ti are better at 1440p & below.

They don't need to lower the price until they can't sell them fast enough.

Nano would be killer at $399 but completely unrealistic. AMD wants a premium for form factor with that level of performance on offer at an excellent perf/w level. I think they will be right, the market will reward them because Nano would make for a killer mITX setup.

Completely agree. Why I'm confused when Russian jumps on that soap box. For years the NV user has ignored the cheaper AMD cards. Hell, the last time we saw a real good market share battle was when HD 7970 was basically coming with up to 6 games and was also cheaper than the slower GTX 680.

This round with there literally no NV premium on the top cards (Fury X vs 980 Ti) I think AMD shot itself in the foot. If historically NV users would pay huge premiums for NV cards what serpent got on Lisa Su's shoulder and convinced her they can take on NV at equal price points? They are not going to win any of NV's users with that.

On the lower side, some AMD users holding on to HD 7Ks probably upgraded seeing AMD finally deliver a viable product. I'm sure most NV users already upgraded to GTX 970/980 and aren't swapping sides.

The R300 series is just way to late in my opinion. If they only made a few tweaks but were able to instantly reverse the negative stigma of R290, they should have done this months ago.

 

[3DVagabond]

Completely agree. Why I'm confused when Russian jumps on that soap box. For years the NV user has ignored the cheaper AMD cards. Hell, the last time we saw a real good market share battle was when HD 7970 was basically coming with up to 6 games and was also cheaper than the slower GTX 680.

This round with there literally no NV premium on the top cards (Fury X vs 980 Ti) I think AMD shot itself in the foot. If historically NV users would pay huge premiums for NV cards what serpent got on Lisa Su's shoulder and convinced her they can take on NV at equal price points? They are not going to win any of NV's users with that.

On the lower side, some AMD users holding on to HD 7Ks probably upgraded seeing AMD finally deliver a viable product. I'm sure most NV users already upgraded to GTX 970/980 and aren't swapping sides.

The R300 series is just way to late in my opinion. If they only made a few tweaks but were able to instantly reverse the negative stigma of R290, they should have done this months ago.

I haven't seen sales numbers for the 300 series and Fury. If you have, please share. I know Fury is still hard to find in stock, so there's no reason to think they need to reduce price.

 

[bystander36]

Completely agree. Why I'm confused when Russian jumps on that soap box. For years the NV user has ignored the cheaper AMD cards. Hell, the last time we saw a real good market share battle was when HD 7970 was basically coming with up to 6 games and was also cheaper than the slower GTX 680.

This round with there literally no NV premium on the top cards (Fury X vs 980 Ti) I think AMD shot itself in the foot. If historically NV users would pay huge premiums for NV cards what serpent got on Lisa Su's shoulder and convinced her they can take on NV at equal price points? They are not going to win any of NV's users with that.

On the lower side, some AMD users holding on to HD 7Ks probably upgraded seeing AMD finally deliver a viable product. I'm sure most NV users already upgraded to GTX 970/980 and aren't swapping sides.

The R300 series is just way to late in my opinion. If they only made a few tweaks but were able to instantly reverse the negative stigma of R290, they should have done this months ago.

When the 5870 came out, AMD did grab some share. When the 7970 first came out, it also gained some share. Both were superior to anything Nvidia had at the time. The Nvidia finally got the 480 out, and took the performance crown and started to take back shares, though they had a noise issue later fixed on the 460 and moved to the 580. The 680 came out and was faster and cheaper than the 7970 as well.

When AMD did have the faster cards, they did gain market share, even with the highest priced cards. I think they expected Fury X to compete better with the 980ti, or that the 980ti wouldn't have been released so soon. Unfortunately, it didn't live up to their expectations, and things aren't going quite to plan.

There are still fans who will stick with Nvidia, but with the single GPU performance lead, AMD has gained ground in the past.

 

[flopper]

I haven't seen sales numbers for the 300 series and Fury. If you have, please share. I know Fury is still hard to find in stock, so there's no reason to think they need to reduce price.

seems the 390 is the bang for the buck card so

 

[Stuka87]

When the 5870 came out, AMD did grab some share. When the 7970 first came out, it also gained some share. Both were superior to anything Nvidia had at the time. The Nvidia finally got the 480 out, and took the performance crown and started to take back shares, though they had a noise issue later fixed on the 460 and moved to the 580. The 680 came out and was faster and cheaper than the 7970 as well.

When AMD did have the faster cards, they did gain market share, even with the highest priced cards. I think they expected Fury X to compete better with the 980ti, or that the 980ti wouldn't have been released so soon. Unfortunately, it didn't live up to their expectations, and things aren't going quite to plan.

There are still fans who will stick with Nvidia, but with the single GPU performance lead, AMD has gained ground in the past.

The 7970 is now far faster than a 680 though thanks to all the driver improvements that GCN has got.

So it could be argued that one or two years from now, the Fury may be faster than a 980Ti. Especially if nVidia continues as they have with neglecting older architectures.

Unfortunately few review sites go back and show just how well older cards do in comparison to each other compared to launch day drivers.

 

[bystander36]

The 7970 is now far faster than a 680 though thanks to all the driver improvements that GCN has got.

So it could be argued that one or two years from now, the Fury may be faster than a 980Ti. Especially if nVidia continues as they have with neglecting older architectures.

Unfortunately few review sites go back and show just how well older cards do in comparison to each other compared to launch day drivers.

That last part definitely doesn't help.

It may also be that Nvidia got to the peak of their arch sooner with drivers, and are not able to improve much any more, where as AMD could have had less optimized drivers at the start. I'm also pretty certain AMD just guessed better on what future games would need, or got lucky.

Fury X could have some of the same kind of success. Who knows, DX12 may prefer AMD's arch over Nvidia's.