AMD Radeon R9 Fury X Reviews
@fellix said:
GTX680 comes with 2GB as well. Maybe you're thinking about the older GTX580, but it wasn't meant to compete with Tahiti anyway.
@3dilettante said:
There is a provision in the spec to allow for ECC, if the DRAM in the stack decides to implement it, but I don't think that is showing up as a consumer option.
One thing I am looking for is whether thermal effects can cause HBM to enter into a high-temperature refresh mode for insufficiently cooled air-cooled cards.
GDDR5 implemented a CRC check to allow it to reach high bus speeds while maintaining an acceptable error rate. HBM has dropped that feature since it has dropped the very high bus speeds.
There is a provision in the spec to allow for ECC, if the DRAM in the stack decides to implement it, but I don't think that is showing up as a consumer option.
One thing I am looking for is whether thermal effects can cause HBM to enter into a high-temperature refresh mode for insufficiently cooled air-cooled cards.
@firstminion said:
Sorry if I'm out of the loop, but you can't grasp that from topics like Is4Genough?.
Sorry if I'm out of the loop, but you can't grasp that from topics like Is4Genough?.
@Jawed said:
http://forums.overclockers.co.uk/showpost.php?p=28248073&postcount=1944
He posted GPU-Z screenies later. He's only overclocked by 50Mhz.
He, like seemingly half the people posting on the OcUK forum, is experiencing unpleasant pump noise (as opposed to normal pump noise). I suspect this is going to erupt soon, especially as OcUK won't replace them as faulty, despite the fact that AMD said they are faulty and should not be in retail.
http://forums.overclockers.co.uk/showpost.php?p=28248073&postcount=1944
He posted GPU-Z screenies later. He's only overclocked by 50Mhz.
He, like seemingly half the people posting on the OcUK forum, is experiencing unpleasant pump noise (as opposed to normal pump noise). I suspect this is going to erupt soon, especially as OcUK won't replace them as faulty, despite the fact that AMD said they are faulty and should not be in retail.
@3dilettante said:
Possibly, there's the foam rubber insert hack that was noted earlier, but that is unfortunate since this time it seemed like AMD had done the right thing by contracting the cooler out to the specialists. Perhaps their inventory tracking is not as good as they think, this is not the same pump whine as before, this is a case of high-frequency hearing loss among the quality-control staff, or their quality control's rigor may be as suspect as before.
Does Cooler Master have other examples of this? It seems off that they'd let this out of their manufacturing process, but they obviously did for the review samples.
Possibly, there's the foam rubber insert hack that was noted earlier, but that is unfortunate since this time it seemed like AMD had done the right thing by contracting the cooler out to the specialists. Perhaps their inventory tracking is not as good as they think, this is not the same pump whine as before, this is a case of high-frequency hearing loss among the quality-control staff, or their quality control's rigor may be as suspect as before.
Does Cooler Master have other examples of this? It seems off that they'd let this out of their manufacturing process, but they obviously did for the review samples.
@silent_guy said:
The difference between stock and +10% mem overclock seems to be minimal.
http://forums.overclockers.co.uk/showpost.php?p=28248226&postcount=1957
Not unexpected, of course.
The difference between stock and +10% mem overclock seems to be minimal.
http://forums.overclockers.co.uk/showpost.php?p=28248226&postcount=1957
Not unexpected, of course.
@gamervivek said:
20Mhz on core clock and like 100Mhz on CPU are quite like to like. Though other user didn't report that big of an increase with 10% overclock.
crossfire scaling in some games by AMDmatt,
http://www.overclock.net/t/1561860/various-amd-radeon-r9-fury-x-reviews/2530
20Mhz on core clock and like 100Mhz on CPU are quite like to like. Though other user didn't report that big of an increase with 10% overclock.
crossfire scaling in some games by AMDmatt,
http://www.overclock.net/t/1561860/various-amd-radeon-r9-fury-x-reviews/2530
@pharma said:
Fury X CrossFire vs reference SLi 980ti results are incoming ...
Poor man's transulation:
http://www.hardware.fr/focus/111/crossfire-radeon-r9-fury-x-fiji-vs-gm200-round-2.html
Fury X CrossFire vs reference SLi 980ti results are incoming ...
Poor man's transulation:
http://www.hardware.fr/focus/111/crossfire-radeon-r9-fury-x-fiji-vs-gm200-round-2.html
@gamervivek said:
Far Cry 4 now runs smooth as butter on 295x2 in TR's recent review of 980Ti while it ran pretty badly during Titan X's release. AMD won't get all developers to play ball like that and of course it doesn't matter much if the game is fixed months after everybody has had their fill of it.
Another HBM overclock and a decent performance increase at 1080p/win7 in kombustor test. From 49 to 57fps just by HBM overclocking. 105Mhz on core only gives 1fps increase in comparison.
http://www.overclock.net/t/1547314/...nano-x-x2-fiji-owners-club/1720#post_24106947
Newest version of gpuz also shows the correct bandwidth value.
Far Cry 4 now runs smooth as butter on 295x2 in TR's recent review of 980Ti while it ran pretty badly during Titan X's release. AMD won't get all developers to play ball like that and of course it doesn't matter much if the game is fixed months after everybody has had their fill of it.
Another HBM overclock and a decent performance increase at 1080p/win7 in kombustor test. From 49 to 57fps just by HBM overclocking. 105Mhz on core only gives 1fps increase in comparison.
http://www.overclock.net/t/1547314/...nano-x-x2-fiji-owners-club/1720#post_24106947
Newest version of gpuz also shows the correct bandwidth value.
@ToTTenTranz said:
A little context from that Hardware.fr quote:
They're complaining about stuttering from 3 games, out of a total of 12 tested games:
[ Anno 2070 ]
[ Batman Arkham Origins ]
[ Battlefield 4 ]
[ Crysis 3 ]
[ Dying Light ]
[ Evolve ]
[ Far Cry 4 ]
[ GRID 2 ]
[ Hitman Absolution ]
[ Project Cars ]
[ Splinter Cell Blacklist ]
[ The Witcher 3 ]
[ Tomb Raider ]
Agreed, no one likes big jerks. They're really unpleasant.
Those jerks...
A little context from that Hardware.fr quote:
They're complaining about stuttering from 3 games, out of a total of 12 tested games:
[ Anno 2070 ]
[ Batman Arkham Origins ]
[ Battlefield 4 ]
[ Crysis 3 ]
[ Dying Light ]
[ Evolve ]
[ Far Cry 4 ]
[ GRID 2 ]
[ Hitman Absolution ]
[ Project Cars ]
[ Splinter Cell Blacklist ]
[ The Witcher 3 ]
[ Tomb Raider ]
Agreed, no one likes big jerks. They're really unpleasant.
Those jerks...
@pjbliverpool said:
I find it amazing that even with HBM + Colour compression, this GPU is STILL bandwidth limited!
I find it amazing that even with HBM + Colour compression, this GPU is STILL bandwidth limited!
@pharma said:
Fury X DriverGate testing results - no magic sauce.
http://www.gamersnexus.net/guides/1996-amd-r9-fury-x-driver-benchmark
Edit: Url fixed.
Fury X DriverGate testing results - no magic sauce.
http://www.gamersnexus.net/guides/1996-amd-r9-fury-x-driver-benchmark
Edit: Url fixed.
@silent_guy said:
I'm not (yet) willing to accept that: if the memory timings (in cycles) don't change with increased clocks, it could be due to reduced latency as well.
@3dilettante said:
Utilization figures in some of the tests like the Techreport's show a sizeable gap between theoretical and realized bandwidth, and AMD's color compression appears to be about as helpful as last-gen Nvidia compression.
Maybe in an ideal world it wouldn't be limited in certain scenarios, but at least some measurements show that utilization is far enough from peak to be a factor.
Utilization figures in some of the tests like the Techreport's show a sizeable gap between theoretical and realized bandwidth, and AMD's color compression appears to be about as helpful as last-gen Nvidia compression.
Maybe in an ideal world it wouldn't be limited in certain scenarios, but at least some measurements show that utilization is far enough from peak to be a factor.
@3dilettante said:
Anecdotal reports appear to have some samples without noise, but multiple with it.
PC Perspective's using retail cards, so it appears to be a substantiated data point.
Best-case, AMD's statement to reviewers was factually inaccurate with respect to some number of retail cards, which would join an impressive run of divergences from reality that executives, documentation, and emails have provided for this product and its rehashed launch brethren.
I can think of two worst-case scenarios, one where AMD's factual inaccuracy to reviewers and paying customers is a universal departure from reality, and one where there was a problem that was fixed, but it's not the same one and it's not AMD's statements but AMD itself that remains in another reality.
I could go on, but maybe someone can track down an mp3 of silence, then a long sigh of resignation, then the sound of someone running their hand down their face, then going Christ.
Anecdotal reports appear to have some samples without noise, but multiple with it.
PC Perspective's using retail cards, so it appears to be a substantiated data point.
Best-case, AMD's statement to reviewers was factually inaccurate with respect to some number of retail cards, which would join an impressive run of divergences from reality that executives, documentation, and emails have provided for this product and its rehashed launch brethren.
I can think of two worst-case scenarios, one where AMD's factual inaccuracy to reviewers and paying customers is a universal departure from reality, and one where there was a problem that was fixed, but it's not the same one and it's not AMD's statements but AMD itself that remains in another reality.
I could go on, but maybe someone can track down an mp3 of silence, then a long sigh of resignation, then the sound of someone running their hand down their face, then going Christ.
