sirplayalot
Senior member
- Nov 27, 2003
- 595
- 23
- 81
My 4770k was not OC'd when I ran that benchmark on the 390x. Stock 4770k and 32GB Mushkin DDR3.
In CCC, if you will bring down the TDP to lets say 50% Hawaii will run at 145W of TDP at around 85-90% of base performance.
http://forums.macrumors.com/threads/nmp-2015-where-is-it.1849662/page-8#post-20880421 The quote is most important.
Power efficiency is an oft-used negative against the large-die Hawaii chips, but I've been playing with powertune settings and Furmark recently as an experiment to fit a "hot and noisy" AMD card into an SFF with limited cooling.
Actually, I stand by an earlier post I made that says I think AMD pushed Hawaii silicon too far.
With both GPU-Z and Furmark able to report power consumptions, I can see a 100W reduction in power consumption on 290X cards for as little as 5% performance loss.
If you have a Hawaii card, I urge you to crank power limits down in the overdrive tab of CCC and see what the resulting clockspeed is under full load. Even in a worst-case scenario, I'm seeing a typical clockspeed of 850MHz with the slider all the way to the left at -50%
That means that Hawaii (the two samples I personally own, at least) can run at 850+MHz on only 145W (half the 290W TDP). As mentioned, that's a worst-case scenario using a power-virus like Furmark. Under real gaming situations (I was messing around with Alien Isolation on 1440p ultra settings) the clocks averaged about 925MHz yet my PC was inaudible; Fans that normally hum along at 55% were barely spinning at 30% during my gameplay.
As Nvidia has proved, you can make a 28nm chip run efficiently. I think the design of Hawaii holds up very well under vastly reduced power constraints - AMD just pushed it outside its comfort zone in order to get the most out of it.
In saying that, the "underpowered" 290X is around the same performance as my GTX970 and also the same cost - significantly higher than a GTX960 4GB. I don't know if die-harvested 290 cards deal with power limit caps like the cherry-picked 290X cards.
So that means your 390x has to have something else going for it since you're 3.5-3.9 ghz 4770k is going against a 4.4ghz 4770k of the same architecture and your 390x only has a 20 mhz advantage over HEXUS's 1030mhz 290x, yet you're still beating them by 200 points. pcb shots and leaks may indicate rebrands, but the numbers are irrefutable. Something else has to be going on with the 390x.My 4770k was not OC'd when I ran that benchmark on the 390x. Stock 4770k and 32GB Mushkin DDR3.
But still an issue at this stage? That would be an outright catastrophy.
Remember this slide that Maddie posted a while back? Not sure where it came from, but it discusses two GPU tapeouts on "28HPM": one ">500sq.mm" and one ">350sq.mm". The GPU with a size of >350 sq. mm. could easily be Tonga; though I don't think AMD released an official figure, some sites estimated it at 359 sq. mm., presumably using calipers. The thing is, a Google search shows that 28HPM is a TSMC mobile process. That would explain a lot... yield problems (if it's a different process than AMD was using on their previous GPUs), the higher transistor density, and why the R9 285 is so much less efficient than the R9 M295X... because the desktop version is being clocked outside its comfort zone. (Note that the 918 MHz clock speed on the R9 285 is lower than most desktop GCN parts.)
I suspect the only reason AMD made this part at all was because Apple wanted it. And since Apple was in the driver's seat, they got a part optimized for their application. This still doesn't explain why they left in junk silicon such as the extra memory controller blocks, unless that was to increase yields by allowing for redundancy. Or unless they started with Tahiti as a base and did a rush job making as few changes as possible. I still think Tonga was supposed to be on 20nm, and they had to hurriedly back-port it to meet Apple's deadline. If it wasn't for the Retina iMac contract, I bet it would have been cancelled entirely.
That still leaves open the question of what the >500 sq. mm. chip on a mobile process could be. Surely not Fiji... why would they even consider 28HPM for that?
Maybe all of the evidence isn't in yet?
So that means your 390x has to have something else going for it since you're 3.5-3.9 ghz 4770k is going against a 4.4ghz 4770k of the same architecture and your 390x only has a 20 mhz advantage over HEXUS's 1030mhz 290x, yet you're still beating them by 200 points. pcb shots and leaks may indicate rebrands, but the numbers are irrefutable. Something else has to be going on with the 390x.
Fury is looking like, at the company level (not consumer level), a major disaster. HBM1 is troublesome and yields are low and costs are high.
HBM1 yields are apparently exactly as expected. HBM2 yields, however, are possibly more problematic - which could be from whence the rumors have arisen - HBM2 concerns will delay the next generation (for AMD and nVidia).
Now you've resorted to extreme exaggerations. Grossly uncompetitive, really?until that happens, they're stuck selling grossly uncompetitive products
They did that before. Nothing changed but the name.
Apple does not care about this Tonga part. They don't even carry it in their stores (the ones they actually sell are all Pitcairn based). Very very low volume - there's no way AMD could recoup the costs of designing Tonga if they are relying on such a niche product. I can't believe the iMac had anything to do with AMD's Tonga strategy.
When have they done this before? Every time they have changed the name, there has been some sort of change to the silicon. There have been times where it has been a rehash of an existing one, but never an identical part with a different name to my knowledge.
Furmark still being used as a measure of anything useful... more clueless people to disregard completely.
The GPU shot we have suggests a very recent manufacturing date versus the newest 290/x GPU shots I could find on google images, T1R*~T2* seem to be 290/x GPUs and T3V* seems to be 390/x... but there is nowhere near enough data available to determine that adequately.
But a delay on HBM2 (or FinFET+) will hurt AMD a lot more than it will hurt Nvidia. AMD needs HBM2 and GloFo 14nm FinFET+ in order to bring the next generation to market; until that happens, they're stuck selling grossly uncompetitive products, many of them dating back to 2012. Nvidia, in contrast, doesn't need to worry nearly as much because they already invested in 28nm Maxwell; if anything, they benefit from being able to get a longer life out of those successful designs.
Basically, Nvidia had the R&D budget and the foresight to make a contingency plan for the failure of Moore's Law, and AMD didn't.
What?
From system to system, the score varies. Everything from different memory modules to background apps will have an effect on the score. Heck, take the same CPU,GPU, RAM, HD, and put them in a different mother board, boot the same OS and you will have a different result.
Yes, even the motherboard will impact the results.
This is the issue with not having a real review site compare results in a controlled environment. Taking some random guys result and comparing with an entirely different system is wrong wrong wrong.
Review sites make sure the comparison is fair and things are equal.
Furmark is useful for finding maximum power draw, though, so...