• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

More perf/watt incomming

Good news if true for the mobile market because it was looking like Nvidia was going to maintain the large edge there with Pascal. This could actually turn the tide in AMD's favor to at least start to gain market share in the mobile space. For the desktop a higher speed 1600mhz+ with DDR5(x) would help but still fall short of the 1070 from Nvidia but the low $300 range is open though for a gpu.
 
Bandalo said:
If you ain't got the performance, push the perf/watt numbers!
Click to expand...

Compared to?

noko said:
Good news if true for the mobile market because it was looking like Nvidia was going to maintain the large edge there with Pascal. This could actually turn the tide in AMD's favor to at least start to gain market share in the mobile space. For the desktop a higher speed 1600mhz+ with DDR5(x) would help but still fall short of the 1070 from Nvidia but the low $300 range is open though for a gpu.
Click to expand...

That is not going to happen with the Polaris parts. Vega is what you're looking for and should easily wipe the floor with the 1070/1080 come launch.
 
oh boy this guy is making things up again, there is no new revision, the server market is using cherry picked low voltage chips.......

This is WTFtech's AMD's favorite author making shit up.
 
razor1 said:
oh boy this guy is making things up again, there is no new revision, the server market is using cherry picked low voltage chips.......

This is WTFtech's AMD's favorite author making shit up.
Click to expand...

This time I am taking notes, so AFTER the launch, I can point at certain people and say, backed by evidence: "Don't not listen to these guys!" ;)
 
well lets keep this in mind too, the r380 version of these types of cards, were also rated at 95 watts lol guess what that didn't really happen ;) they were heavily throttling even when they were binned low voltage parts. I expect the same here even if they are special low voltage binned parts
 
Looks like this is less of a rumor and the lower power Polaris chips are hitting the market. Look at the below review of a XFX Rx480, the temperatures, OC of 1475mhz and indicated power is showing basically a 50% lower power level then the first lauched Rx480's. This is some great news and really puts AMD tied or slightly ahead of Nvidia perf/w.

 
Factum said:
This time I am taking notes, so AFTER the launch, I can point at certain people and say, backed by evidence: "Don't not listen to these guys!" ;)
Click to expand...
After the launch of what ?

noko said:
Looks like this is less of a rumor and the lower power Polaris chips are hitting the market. Look at the below review of a XFX Rx480, the temperatures, OC of 1475mhz and indicated power is showing basically a 50% lower power level then the first lauched Rx480's. This is some great news and really puts AMD tied or slightly ahead of Nvidia perf/w.

Click to expand...


Where's the rest of them? Why is there only one SKU displaying this behavior ? Why do the other people who own this exact same SKU report wildly varying clock-voltage characteristics ? Why isn't this in mobile ?
 
Ieldra said:
After the launch of what ?



Where's the rest of them? Why is there only one SKU displaying this behavior ? Why do the other people who own this exact same SKU report wildly varying clock-voltage characteristics ? Why isn't this in mobile ?
Click to expand...
If there is a revision or process improvement it would be the later chips that would see this improvement thus later models or cards manufacture will get this benefit if the case. Now we just have to wait and see if this is a fluke or actually the case. To me it is likely a real chip level advancement.
 
noko said:
If there is a revision or process improvement it would be the later chips that would see this improvement thus later models or cards manufacture will get this benefit if the case. Now we just have to wait and see if this is a fluke or actually the case. To me it is likely a real chip level advancement.
Click to expand...

Then why do other people who own the same exact card not report the same results ? It may very well be a binned SKU. Also it's so easy to prove it's not a revision, ask someone with the card to look at their die, unless AMD are hiding it it'll be plainly obvious from the serial on the die.

If it is a process improvement which yields a greater proportion of good performing chips then why have those better clocking chips been totally absent so far? All of the high end RX480s so far have behaved fairly similarly across the board
 
noko said:
If there is a revision or process improvement it would be the later chips that would see this improvement thus later models or cards manufacture will get this benefit if the case. Now we just have to wait and see if this is a fluke or actually the case. To me it is likely a real chip level advancement.
Click to expand...


That kind of power change without a major change in the mask layer is unprecedented (which is a 2 q or 6 month change), we have seen nodes have high volatility with chips on the same wafer with voltages though (Fermi had this problem too, so its not uncommon at all to have a wide variation), and I think AMD did hint on this with Polaris if I'm not mistaken, so pretty sure that is all we are seeing on of the best of the best binned.
 
Last edited:
It's a hell of a quality distribution if this is just the best of the best. That said, if amd can get a 50% improvement then either they or GF really screwed up the first revision.
 
razor1 said:
That kind of power change without a major change in the mask layer is unprecedented (which is a 2 q or 6 month change), we have seen nodes have high volatility with chips on the same wafer with voltages though (Fermi had this problem too, so its not uncommon at all to have a wide variation), and I think AMD did hint on this with Polaris if I'm not mistaken, so pretty sure that is all we are seeing on of the best of the best binned.
Click to expand...
Vr-zone said there is a change to mask layer. It will be about 6 months or more when we actually see these cards. Looks like this power decrease when vega first.



"The gains in efficiency are attributed to a more refined binning process as well as enhancements to the 14nm metal mask layers. The revision will see the TBP go from 150W to less than 95W on the RX 480, and will see the Polaris 11-based RX 460 go from 75W to less than 50W. The revised GPUs will also be able to deliver more performance, with the RX 460 seeing a boost from 2.15 TFLOPS to 2.5 TFLOPS."

Read more: AMD rolling out revised Polaris GPUs with 50% increase in performance/watt
 
Its not getting more right the more time people post the same YouTube video and the same news chain based on the same "source.
 
So the next AMD mantra is wait for "the respin"?
Waiting seems like the key factor for AMD...
 
NKD said:
Vr-zone said there is a change to mask layer. It will be about 6 months or more when we actually see these cards. Looks like this power decrease when vega first.



"The gains in efficiency are attributed to a more refined binning process as well as enhancements to the 14nm metal mask layers. The revision will see the TBP go from 150W to less than 95W on the RX 480, and will see the Polaris 11-based RX 460 go from 75W to less than 50W. The revised GPUs will also be able to deliver more performance, with the RX 460 seeing a boost from 2.15 TFLOPS to 2.5 TFLOPS."

Read more: AMD rolling out revised Polaris GPUs with 50% increase in performance/watt
Click to expand...

VR Zone is reporting what WCCF did.

The embedded Polaris GPUs are just binned and throttle aggressively to stay within limits, and even then we don't know if they'll actually stay below 95W, see Ryan Smith's comment I posted earlier.

Changes in the metal layer won't lead to significant changes in power draw and efficiency, as someone on B3D put it ; there's no way this will help unless AMD had giant metal caps in the interconnect for absolutely no reason other than to come back and remove them to claim an improvement :p

Negligible power is lost in the interconnect wrt rest of chip, so I'm calling potatoes on this notion
 
Ieldra said:
Changes in the metal layer won't lead to significant changes in power draw and efficiency, as someone on B3D put it ; there's no way this will help unless AMD had giant metal caps in the interconnect for absolutely no reason other than to come back and remove them to claim an improvement :p
Click to expand...
Actually they can. Disconnect power form half the chip you can cut power consumption in half. Fail to properly ground something, who knows what the outcome will be? While it won't directly influence efficiency, it could do so indirectly. Ground loops and noise can be a real PITA on power systems. RF systems already turn metal layers into giant caps to counteract interference and shield circuits. If they had really high gain transistors for monitoring droop on their adaptive clocking, separation of grounds could make a difference.
 
Anarchist4000 said:
Actually they can. Disconnect power form half the chip you can cut power consumption in half. .
Click to expand...

Funny guy :p

Anarchist4000 said:
Fail to properly ground something, who knows what the outcome will be? While it won't directly influence efficiency, it could do so indirectly. Ground loops and noise can be a real PITA on power systems. RF systems already turn metal layers into giant caps to counteract interference and shield circuits. If they had really high gain transistors for monitoring droop on their adaptive clocking, separation of grounds could make a difference.
Click to expand...
Fair enough there could be a bug of sorts in the power system, but still this would be unprecedented. Anyway, all this talk of revisions originated from WCCF
 
Anarchist4000 said:
Actually they can. Disconnect power form half the chip you can cut power consumption in half. Fail to properly ground something, who knows what the outcome will be? While it won't directly influence efficiency, it could do so indirectly. Ground loops and noise can be a real PITA on power systems. RF systems already turn metal layers into giant caps to counteract interference and shield circuits. If they had really high gain transistors for monitoring droop on their adaptive clocking, separation of grounds could make a difference.
Click to expand...


And you think that wouldn't be a problem they would have fixed ASAP when they did 3 respins on Polaris to begin with lol? Yeah that is one of the first things they will look at in validation the electrical layout if their are any problems such as what you stated, not something they will fix well after release, and we know they had time to make those kinds of changes.
 
NKD said:
Vr-zone said there is a change to mask layer. It will be about 6 months or more when we actually see these cards. Looks like this power decrease when vega first.



"The gains in efficiency are attributed to a more refined binning process as well as enhancements to the 14nm metal mask layers. The revision will see the TBP go from 150W to less than 95W on the RX 480, and will see the Polaris 11-based RX 460 go from 75W to less than 50W. The revised GPUs will also be able to deliver more performance, with the RX 460 seeing a boost from 2.15 TFLOPS to 2.5 TFLOPS."

Read more: AMD rolling out revised Polaris GPUs with 50% increase in performance/watt
Click to expand...


Same news as WTFtech.

and no they did not say specific mask layer, they stated metal mask layer efficiency that means node maturation, not something on AMD's end.

Its most likely hogwash, if its sounds too good to be true, it probably is, and the complexity of these chips with the fact we know from past experiences no such thing has ever taken place not from a respin.
 
razor1 said:
And you think that wouldn't be a problem they would have fixed ASAP when they did 3 respins on Polaris to begin with lol? Yeah that is one of the first things they will look at in validation the electrical layout if their are any problems such as what you stated, not something they will fix well after release, and we know they had time to make those kinds of changes.
Click to expand...
If they knew what the problem was and had an easy fix they wouldn't need 3 respins would they? Your validation argument is nonsense as even if they found a problem they still need to devise a fix for it.
 
Anarchist4000 said:
If they knew what the problem was and had an easy fix they wouldn't need 3 respins would they? Your validation argument is nonsense as even if they found a problem they still need to devise a fix for it.
Click to expand...

If they haven't fixed it with 3 mask respins lol, man that is a year and half of work they aren't going to fix it lol.
Ok look the nonsense you post, tell me the steps of chip design once a chip is fabricated, then we can talk, you make things up based on some twisted logic that you have come up with which just isn't possible. I'll give you 15 minutes, if you don't do it, I will and I will go into depth on what each step entails.

Stop posting BS you don't know about, its better when you ask a question or understand when someone says something against what you say, most likely they know more than what they have posted.

More than 15 minutes need more time to search for it? Give ya till the end of the day how about that?
 
Last edited:
razor1 said:
Same news as WTFtech.

and no they did not say specific mask layer, they stated metal mask layer efficiency that means node maturation, not something on AMD's end.

Its most likely hogwash, if its sounds too good to be true, it probably is, and the complexity of these chips with the fact we know from past experiences no such thing has ever taken place not from a respin.
Click to expand...


So you are saying. No way in hell it is possible? No way for them to have made improvements to the design and get more per/watt. Idk. To me this may be possible. We will see. I kinda tend to believe that Polaris was being constantly tweaked as the rumors suggested. Looks like lots of improvements in the node process were yet to be made. It will be over 6 months when they release these according to rumor so it's possible.
 
I asked somebody with the card in question to test for me. He uploaded afterburner logs.

I'm not home yet

Here they are

 
NKD said:
So you are saying. No way in hell it is possible? No way for them to have made improvements to the design and get more per/watt. Idk. To me this may be possible. We will see. I kinda tend to believe that Polaris was being constantly tweaked as the rumors suggested. Looks like lots of improvements in the node process were yet to be made. It will be over 6 months when they release these according to rumor so it's possible.
Click to expand...


I'm saying its very unlikely that its what it is. They would have had to make those changes starting early on in chip design, it is not a node thing. If this comes to be truth, AMD pulled out some magic.

I'll give you an example of a 15% increase from Intel CPU's, which they have absolute control over the node libraries and architecture. But the microcode for Intel's tick tock cycle, going to tock they can only use 20% of the tick microcode as a base, that shows you the amount of changes they have to do to gain 15% on perf/watt. And we know Intel's capabilities are above and beyond AMD when it comes to the node right? And then factor in CPU's are hand layout, not automation....
 
  • Like
Reactions: NKD
like this
I'm starting to think PakistanTech should be held liable for all the bad blood arguments their rumor mill 'articles' cause.
 
Well if you have some running at 95w and then others at 153w - a rather huge deviation - but proven to be true if the same chip design -> I would say you would look at why some come out much better then the others (what is causing that) and make the changes so they all or most of them come out with the better perf/w. There is proof already Polaris can be very high perf/w raitio's while others are less. Six sigma the design for more consistency of the chips on the die. This could indeed be the node process because of results/data from the same gpu design .
 
well we will see this with the 1050ti too if it is (Samsung 14nm same process as GF's.), but for a chip like that I guess it won't matter as much......
 
You must log in or register to reply here.
Back
Top