Sapphire Fury Nitro mini review (vs. 390 Nitro)

[Krzych]

Sapphire Fury Nitro mini review (vs. 390 Nitro)

 

Recently I had my hands on Sapphire R9 390 Nitro and Sapphire R9 Fury Nitro and since I like benchmarking and testing I made few tests and will post them here, maybe someone will find it useful.

 

Introduction

 

I tested both cards for overclocking, temperatures, power draw, undervolting, performance in benchmarks and for games only in Witcher 3 since this is not any kind of huge review, this is why I called it mini review, it is just kind of amateur comparison from someone who is bored and likes benchmarking

 

More information about specification and performance can be found elsewhere, for example in serious reviews.

 

Specific information about entire testing platform can be found in my signature. 

 

Important for temperatures, case is Fractal Design Define R5 with 5x slow RPM Silentium PC Sigma Pro 140mm fans set to 5V - 400 RPM (2x front intake, bottom intake, side intake, rear exhaust)

 

 

 

Core Clocks, Memory Clocks and Fan Speed

 

 

Maximum overclock I was able to get from R9 390 was 1090/1650. Going any higher didn't caused crashes, but there were some visual artifacts. Also, the card needed massive +100mV voltage boost to be able to overclock at all.

 

Things were looking better with Fury and I was able to get 1120/500 with slight +18mV overvoltage. Going any higher resulted in crashes, even with maximum possible +96mV overvoltage. Overclocking HBM memory caused artifacts.

 

As for fan speeds and noise, I don't heave any device for measuring noise, but I set fixed fan speed to both cards that I think is quiet and is not exceeding the sound of my 400 RPM case fans, so this kind of sound won't affect gaming experience if you play without headphones. Fan speed is set to be equal in terms of noise on both cards.

 

Performance

 

(in-game performance was measured with FRAPS)

 

You can see that Fury takes very significant lead over 390, which is expected because those are two different price segments.

 

Temperatures and Power Consumption

 

Power draw measurements are made for entire setup including monitor and speakers, so you can cut ~25-30W from those measurements.

 

 

Here Fury again takes significant lead. On paper both cards have TDP at 275W and their coolers are looking very similar, but the reality shows completely different thing.

 

On stock Fury is 14 degrees cooler than 390 and 17C cooler after overclocking both cards. VRM temperature measurements show the same thing and differences almost reach 20 degrees.

 

On stock settings, Fury is also much more power efficient card, delivering 25% more performance with lower power draw.

 

Power consumption after overclocking shows enormous difference between the two, but each card overclocks differently, so those scores cannot be taken as constant for all cards. This time both didn't overclock too well, but R9 390 needed significantly more voltage to stabilize itself on 1090 clock and this is why we see such huge power consumption. At the same time, Fury needed only +18mV, which increased power draw only by a bit. Setting +96mV overvoltage to Fury would result in the same enormous 450W+ power draw.

 

I have also downclocked Fury to 390 performance to see the difference in power draw and temperatures. I don't know how much sense such comparison makes, probably not much, but anyways, there it is:

 

 

 

Value

 

 

Based on prices in Germany from where I got my Fury, value is exactly the same for both cards. Normally higher end cards have less value than lower end ones, but releases of new generation GPUs resulted in some very interesting discounts for higher end models of previous generation. However, such discounts are not available in all countries, for example in Poland R9 Fury costs exact the same as on release. But some of retailers in Germany and UK decided to discount this card and it can be found for 349 EUR or 299 GBP, compared to 279 EUR or 250 GBP for R9 390.

 

Undervolting R9 Fury

 

After determining overclocking limitations on my Fury I decided to go opposite way and see if I can lower the voltage on the card and make it run at lower temperatures and power draw. 

 

 

Results are quite amazing. I was able to undervolt my card on stock clocks by as much as -90mV without causing instability or performance decreasement.

 

 

As you can see differences in performance are withing margin of air. Card passed multiple stability tests on undervolted settings.

 

This gave some amazing scores for temperatures, power draw and performance per watt:

 

 

Undervolting the card resulted in 10 degrees temperature decrease on core and 15C on VRM compared to stock voltage.

 

Power draw dropped by 51W which resulted in significant performance per watt increase.

 

Conclusion

 

Overall, my results are far from what I expected, especially from very well reviewed and respected Sapphire R9 390 Nitro.

 

The cheaper card was much worse than I expected it to be, especially in terms if cooling performance. To maintain good noise levels I had to let the card go as hot as 85C which left no overclocking potential. But even if much higher fan speeds, this card wasn't able to overclock well. Performance in games and benchmarks is good for a price and the card is equipped with 8 GB of GDDR5 memory with very wide memory bus, so there are no issues to expect in future, but with this kind of temperatures and power draw I don't find this card attractive. I have nothing against huge power draw and high temperatures on high end cards, but for mid end card drawing so much power and not being able to cool itself properly with maintaining good noise levels is just not reasonable. 

 

However, things are much different with Sapphire Fury Nitro. The card delivers very good performance for 1080p and 2560x1080 gaming while maintaining low temperatures at very low noise levels. The cooler on this card is just amazing. Card overclocked slightly better than R9 390, but this is quite random thing based on luck. Another amazing thing on this card is how well it takes undervolting. At stock I was able to lower the voltage so much that temperatures and power draw dropped very significantly without any performance decrease. This makes the card even more quiet, because the lowest 25% fan speed (800 RPM) is now enough to cool the card even at heavy load in demanding games. Overall card is better than I expected after experiences with 390 Nitro. Card has amazing cooler, ridiculously low temperatures, good noise levels and even power reasonable power efficiency, especially after undervolting.

 

 

 

[Krzych]

11 more games on 2560x1080 and 3440x1440 resolutions: 

 

[FIXXX]

Memory clock 500? wha?

[Mr.Meerkat]

On 10/24/2016 at 3:36 PM, FIXXX said:

Memory clock 500? wha?

That's known as HBM memory where it's on a 4096bit memory bus so it's memory clock can be extremely low while still providing more bandwidth than normal GDDR5 memory  

[FIXXX]

31 minutes ago, Mr.Meerkat said:

That's known as HBM memory where it's on a 4096bit memory bus so it's memory clock can be extremely low while still providing more bandwidth than normal GDDR5 memory  

Guess it'll become a mainstay eventually with the next AMD GPU series. 

Sure looks odd though

[Clyne]

I guess I got a golden chip...

Fury Nitro 1150/500 stable with no voltage increase or power limit, stable in Heaven/TR/GTA 5.

[Space Reptile]

question to power draw ; what is the rest of your test setup?

[Krzych]

3 hours ago, Space Reptile said:

question to power draw ; what is the rest of your test setup?

Like I said at the beginning it is in my signature. I also said that you can cut off 25-30W from power draw because this was power draw from the wall and included monitor (~25W) and speakers (~5W). You can find more specific info about power draw elsewhere, this was just quick review. 

[TheRandomness]

There's a method to overclock HBM.. I've read on OCN that the clock actually goes up in relatively large jumps (around 130~MHz iirc) and it'll be more stable the closer you are to a step.