Anandtech's AMD Fury X review

[Stuka87]

I would be very surprised if overclocking the memory gave any game performance increase. Synthetics do great with memory OC, but games only do if there is truly a bandwidth issue. Which upper tier GCN cards typically do not have.

 

[2is]

At this point, AT should just wait to do their review when voltage tools are available for the Fury. No point in releasing a week late with no new info.

If they even become available. With the VRM's hitting >100C already I just don't see them being able to reliably handle a whole lot more.

 

[Stuka87]

If they even become available. With the VRM's hitting >100C already I just don't see them being able to reliably handle a whole lot more.

Most VRM's are rated for 125-150C depending on the chip. Majority of them are rated for 125C though.

 

[2is]

Most VRM's are rated for 125-150C depending on the chip. Majority of them are rated for 125C though.

I should have probably clarified. The 100C measurement was the back of the PCB so the VRM's themselves are undoubtedly even hotter. Hopefully they're using the 150C variant, because I'll bet they're already pretty close to the 125C mark.

 

[Makaveli]

Most VRM's are rated for 125-150C depending on the chip. Majority of them are rated for 125C though.

Even though they rated for those temps that would bother the heck out of me.

And just like a PSU is rated at 750watts you don't want to be running it at 750watts for a prolong period of time.

 

[Headfoot]

I'm still waiting to see what unlocked voltage does. Someone on OCN was able to push the memory to 600mhz and get a 3k score increase in firestrike, so I'm still keeping an open mind about the card. If it hits 1300mhz with a voltage bump, can overclock the memory consistently, and still stay under 65C while being quiet then it will be a decent alternative to the competition. Then we may get a trickle of driver improvements. Not the run away success that we were hoping for, but it may turn into a nice option.

Edit: Here's a link to the men oc:
http://www.overclock.net/t/1562593/...n-be-overclocked-after-all/0_50#post_24105109

Agreed. It all rests on voltage unlocked overclocking now. I think its safe to assume between 5-10% average increase in performance due to drivers on Fiji over time. So if overclocking can get Fiji within 5-10% of overclocked 980 Ti then its not as disappointing as many make it out to be. On the other hand, if it really does overclock poorly even with unlocked voltage, then its all up to Fury (no x). I'm fairly confident it will overclock decently with voltage. Water cooled, overengineered for overclocking. If it doesn't clock at least as high as aftermarket 290x i'll be surprised.

 

[therealnickdanger]

And just like a PSU is rated at 750watts you don't want to be running it at 750watts for a prolong period of time.

You didn't clarify, so I assume that you're referring to all PSUs on the market. Any PSU worth its weight should have no problem operating at 100% its rated load for extended periods. As long as a unit is properly cooled and has the right protective circuitry, the only downside is that it's not as efficient as a more powerful PSU at the same load.

 

[Stuka87]

I should have probably clarified. The 100C measurement was the back of the PCB so the VRM's themselves are undoubtedly even hotter. Hopefully they're using the 150C variant, because I'll bet they're already pretty close to the 125C mark.

Looks like the french site that took the temps was running furmark at the time. Which means their measurement means nothing. Gaming temps are going to be *WAY* less than furmark temps.

 

[Despoiler]

You didn't clarify, so I assume that you're referring to all PSUs on the market. Any PSU worth its weight should have no problem operating at 100% its rated load for extended periods. As long as a unit is properly cooled and has the right protective circuitry, the only downside is that it's not as efficient as a more powerful PSU at the same load.

Modern high efficiency, quality PSUs are much much better compared to older ones. Half and full load efficiency are generally +-5% of each other..

Looks like the french site that took the temps was running furmark at the time. Which means their measurement means nothing. Gaming temps are going to be *WAY* less than furmark temps.

100C running a power virus is great.

 

[AtenRa]

Looks like the french site that took the temps was running furmark at the time. Which means their measurement means nothing. Gaming temps are going to be *WAY* less than furmark temps.

From Toms,

http://www.tomshardware.com/reviews/amd-radeon-r9-fury-x,4196-8.html

During gaming, the VRMs stay reasonably cool, even though they're only covered by a small heat sink that touches a heat pipe above it. The board hits 60 °C at the slot, meaning the VRM’s heat travels across the PCB under the rubberized back plate.

 

[IEC]

If they even become available. With the VRM's hitting >100C already I just don't see them being able to reliably handle a whole lot more.

100C under Furmark/power virus isn't anything to be concerned about...

Tom's showed gaming VRM temps in the 60s IIRC.

Edit: Beat by AtenRa

 

[tonyfreak215]

I should have probably clarified. The 100C measurement was the back of the PCB so the VRM's themselves are undoubtedly even hotter. Hopefully they're using the 150C variant, because I'll bet they're already pretty close to the 125C mark.

Turns out the 100C might be a hoax.

Even the water is at 100C, if it was truly that, the water would be boiling also.

http://forums.anandtech.com/showthread.php?t=2437611

http://forums.anandtech.com/showpost.php?p=37522323&postcount=24

 

[Silverforce11]

Just touch (flir cam) the radiator. In a water loop, the radiator temp is ~ the temp of the loop once it equilibrate at a certain temp.

That hw.fr measure is idiotic they should have known the basics of water cooling.

Guru3d's FLIR results:
http://www.guru3d.com/articles_pages/amd_radeon_r9_fury_x_review,12.html

Note that the PCB and GPU/VRM area is ~the temperature of the loop.

 

[LTC8K6]

I doubt very much that there is "water" in these coolers. It has to be some sort of coolant mix, because that boils at a much higher temp and it transfers heat much better than plain water.

 

[tential]

I doubt very much that there is "water" in these coolers. It has to be some sort of coolant mix, because that boils at a much higher temp and it transfers heat much better than plain water.

I'm surprised people even think it's water....
This is an enthusiast forum right?

 

[AnandThenMan]

I doubt very much that there is "water" in these coolers. It has to be some sort of coolant mix, because that boils at a much higher temp and it transfers heat much better than plain water.

Outside of mercury water has the best heat transfer property of just about any liquid. Silicone oil may be better than water not sure but it takes a lot of energy to pump liquids like mercury or silicone oil so they are not practical.

 

[tential]

"The antifreeze/coolant contained within the liquid cooling solution for your graphics
card is a hazardous chemical that may cause bodily harm if it is allowed to escape from
the cooling solution. This section provides details about the coolant and the safety
measures that should be taken in case of coolant leakage."

http://support.amd.com/Documents/amd-radeon-r9-fury-x.pdf

Just lost a lot of respect for a lot of posters right now.... You guys can't have seriously made these posts so authoritatively without a simple google search right?

 

[Silverforce11]

People put coolant into their water loop to prevent evaporative loss over time through the tubes.

You are idiots if you think a AIO loop is running at 100C when the GPU is 50-60C. Learn some water cooling basics first.

In a water cooling loop, the temps of all the components, including the liquid is very similar to each other once it has been given some time to equilibrate. Look at the radiator & tube temps. There's no way its 100C.

Users report they can put their hand on the rad in operation and its warm. Btw my own rads on the R290s (sig) is only warm to the touch, while mining bitcoins. My thermaltake AIO has water sensors, reports ~45C.

 

[AnandThenMan]

Just lost a lot of respect for a lot of posters right now.... You guys can't have seriously made these posts so authoritatively without a simple google search right?

Who are you referring to? The glycol type coolant used in the Fury cooler is a very common chemical used in many, many products. I agree with Silverforce there is no way there is such a large temperature difference as some of these so called expert sites are showing.

Applications for MEG

Monoethylene Glycol (MEG) can be used for applications that require chemical intermediates for resins, solvent couplers, freezing point depression, solvents, humectants and chemical intermediates. These applications are vital to the manufacture of a wide range of products, including resins; deicing fluids; heat transfer fluids; automotive antifreeze and coolants; water-based adhesives, latex paints and asphalt emulsions; electrolytic capacitors; textile fibers; paper and leat

 

[Sabrewings]

A good computer liquid cooling system shouldn't reach more than 10*C delta T from ambient air temp. Any more and you have an under radiated system. If your delta T keeps rising so will your silicon temps as they lose their delta T advantage to the fluid. Of course, there will eventually be a balance point where the delta T from ambient is enough to move the heat input to the system (higher delta T means more efficient transfer of energy), but again this means higher silicon temps.

Personally, my fluid temps never rise above 34C even under max loads for long duration. This translates to a 55C CPU load temp on P95 small FFT and a 40C GPU max load temp (only fur mark, never in game). These temps (fluid, GPU, and CPU) were taken after over an hour of P95 (8 workers) and Furmark.

100C is downright ridiculous and not believable.

 

[Silverforce11]

@Sabrewings
Indeed, the coolant isn't hotter than the GPU. If hardware.fr thinks its 100C, the GPU core must be 110C at least. lol

 

[Sabrewings]

@Sabrewings
Indeed, the coolant isn't hotter than the GPU. If hardware.fr thinks its 100C, the GPU core must be 110C at least. lol

Absolutely. The air coming out the back would be screaming hot as well with a 70C delta T to work with.

 

[Abwx]

You are idiots if you think a AIO loop is running at 100C when the GPU is 50-60C. Learn some water cooling basics first.

A heatsink that get warmer than the heat source it is supposed to cool would be a first in human history..

So they got 100°C water out of a 60°C thermal source...??..

There s no other firm that is as much targeted with viral marketing practices than AMD, even absolute non sense is used as a weapon, meanwhile Hardwre.fr made a CF test..

Conclusion is that a Fury CF work at its rated frequency while a SLI throttle even in open air with a slot separating the cards, i guess that we wont hear of such issues discussed here, instead we have magical physics used as arguments...

http://www.hardware.fr/focus/111/crossfire-radeon-r9-fury-x-fiji-vs-gm200-round-2.html

 

[KingFatty]

What if they simply turned off the pump/fan when doing the test - could the Fury X temperatures get that high with the fan/pump turned off?

 

[Silverforce11]

Ouch, open bench test and SLI 980Ti loses major performance due to temperature throttling!

Would be interesting to see custom 980Ti SLI configs in a case vs Fury X CF. But the reference 980Ti SLI gets a stomping.

Edit: Just had a look at their test in detail (http://www.hardware.fr/focus/111/crossfire-radeon-r9-fury-x-fiji-vs-gm200-round-2.html), they are testing 4K with maxed settings, often with maxed AA available, 4x MSAA and even 8x MSAA. Basically the worse scenario for Fury X CF and it still beats 980Ti SLI. Albeit with more stuttering in Dying Light (and crap performance in general in that GameWorks title) and some stuttering in Evolve/Witcher 3. Witcher 3 isn't a vram issue since that game doesn't use much vram, its a driver issue again. Similar to GTA V where users report the recent "Batman beta" drivers added stutter than the prior version did not.