[WCCFTECH]AMD Rolling Out New Polaris GPU Revisions With 50% Better Perf/Watt

[ManofGod]

It's not luck. It's the same card as the HIS which shows the exact same lower power traits, better clocking, no fan spin coding till 50-60c, etc etc, same pcb. But hey, its AMD they suck and everyone is lying.

Now you got it, AMD sucks and everyone is lying. Good to see AMD improving things and the process they are on making improvements as well. This is not the same process that Zen will be on though, correct?

 

[Shintai]

This is not the same process that Zen will be on though, correct?

Its the exact same process.

 

[lolfail9001]

It's not luck. It's the same card as the HIS which shows the exact same lower power traits, better clocking, no fan spin coding till 50-60c, etc etc, same pcb. But hey, its AMD they suck and everyone is lying.

So, did anyone actually verify sensor readings?

 

[Ieldra]

It's not luck. It's the same card as the HIS which shows the exact same lower power traits, better clocking, no fan spin coding till 50-60c, etc etc, same pcb. But hey, its AMD they suck and everyone is lying.

What same power traits? You linked total system power consumption which will be affected by factors such as more powerful card making the CPU draw more power etc etc etc, it's horribly inaccurate.and you can tell because of how skewed the results are. TPU tests at the rails. This means the measure the actual voltage supplied, they measure the actual current supplied and they multiply them.

So, did anyone actually verify sensor readings?

Is this a joke man ? Why would you verify sensor readings? WCCF says they're a respin, so there's a respin GOD DAMMIT.

Who needs verification?

 

[Ieldra]

So, did anyone actually verify sensor readings?

this is the best I could find, note 145W for current GPU power draw

[M] HIS RX 480 IceQ X2 Roaring Turbo 8GB Video Card Review

But by all means ManofGod thesmokingman don't let this stop you from turning this healthy skepticism into another one of your rants about how everyone is always against AMD, it certainly didn't stop you when he had the RX480 1600mhz "leaks" from the very same source.


Edit: scrap that, retarded readings

Edit 2: found this

HIS Radeon RX 480 IceQ X2 Roaring Turbo 8GB Review - VideoCardz.com

 

[thesmokingman]

So, did anyone actually verify sensor readings?

That'll be possible as soon as someone gets their hands on one. The two card pcb's I linked above are so far completely different from designs that came before and they share many similarities besides the pcb so it leads me to believe its not a fluke or custom job.

What same power traits? You linked total system power consumption which will be affected by factors such as more powerful card making the CPU draw more power etc etc etc, it's horribly inaccurate.and you can tell because of how skewed the results are. TPU tests at the rails. This means the measure the actual voltage supplied, they measure the actual current supplied and they multiply them.



Is this a joke man ? Why would you verify sensor readings? WCCF says they're a respin, so there's a respin GOD DAMMIT.

Who needs verification?

lol, you compare the charts to themselves, ya can't cross compare duh. But hey, I'm not surprised your shitting a brick. How dare these peons improve their process or design or whatever!

 

[Ieldra]

That'll be possible as soon as someone gets their hands on one. The two card pcb's I linked above are so far completely different from designs that came before and they share many similarities besides the pcb so it leads me to believe its not a fluke or custom job.



lol, you compare the charts to themselves, ya can't cross compare duh. But hey, I'm not surprised your shitting a brick. How dare these peons improve their process or design or whatever!

Yes, you can't compare power consumption of just the card to the entire system, must be a shocker for you.

I'll give you an example so simple even you should have no trouble understanding it, see on the TPU charts the reading for the 390 and 980Ti respectively ? Compare to the chart you just posted. There's your margin of error right there, total system power consumption IS NOT good for measuring the power consumption of single cards.



I make a habit of not shitting bricks because that would entail a significant widening of my sphincter I'm not sure I'm capable of, I'll leave that to you.

Now let’s get to the point. I’ve successfully overclocked this card to 1430 MHz. It was stable configuration that allowed me to finish all synthetic tests. However, once we got into gaming tests, it wasn’t so stable anymore, as some tests wouldn’t even start. It got to the point when our ASUS motherboard’s circuit protection would shut down whole PC, as too much power was being drained by the GPU. According to GPU-Z the card was consuming 203W of power, that’s really close to 225W we could get through 8pin connector and PCIe slot…

203W just the GPU.

Edit: you also posted the incorrect (and irrelevant) TPU chart

You're welcome.

Edit 2:

I mean, it doesn't take a genius to figure this out. 980ti draws 25W more than 390X. Okay.
GTX 1080 draws 30W less than 980Ti . Okay.

That means GTX 1080 draws less than a 980Ti by 30W, and that it draws 5W less than a 390X

I don't say this often, I like to save this for the really golden ones but ...

TOP LEL

Edit4:
Someone needs to send Bjorn3D a message and ask what they're smoking, how the FUCK do they post 1080 power at + 110W vs a 1060 ?

Who the fuck edits their shit? A trained baboon ?


GTX 1070 drawing more power than a 390X.

MY SIDES.

Oops, just noticed 390>390X power consumption as well..

Oh man, thesmokingman you've really dug yourself into a big hole now, and your name is so appropriate ....

 

[razor1]

What same power traits? You linked total system power consumption which will be affected by factors such as more powerful card making the CPU draw more power etc etc etc, it's horribly inaccurate.and you can tell because of how skewed the results are. TPU tests at the rails. This means the measure the actual voltage supplied, they measure the actual current supplied and they multiply them.



Is this a joke man ? Why would you verify sensor readings? WCCF says they're a respin, so there's a respin GOD DAMMIT.

Who needs verification?

Crazy theory of respins when there are none, just binned chips with low voltage.

All of you guys that want to believe this article, just ask youself in the past, how much has a respin metal or mask, ask you self how long they usually take and how much have they ever changed power usage that dramatically (outside of faults in the design)?

Lets see, only power usage not the performance of the chip.

FX respins didn't help it

r600 respins didn't help it

Fermi respins didn't help it until Fermi V2, where power was the same or a bit higher but performance was much higher.

Come on guys it takes a minimum of 1q to 2 q's after a respin (depending on which type mask or metal) to start up manufacturing of the modified chips and validating and getting them to the market. AMD didn't have that much time lol, simple logic dictates binned low voltage chip,,,,,,

 

[Ieldra]

Crazy theory of respins when there are none, just binned chips with low voltage.

All of you guys that want to believe this article, just ask youself in the past, how much as a respin metal or mask, ask you self how long they usually take and how much have they ever changed power usage that dramatically (outside of faults in the design)?

Lets see, only power usage not the performance of the chip.

FX respins didn't help it

r600 respins didn't help it

Fermi respins didn't help it until Fermi V2, where power was the same or a bit higher but performance was much higher.

Come on guys it takes a minimum of 1q to 2 q's after a respin (depending on which type mask or metal) to start up manufacturing of the modified chips and validating and getting them to the market. AMD didn't have that much time lol, simple logic dictates binned low voltage chip,,,,,,

What if P10v1 and P10v2 were developed asynchronously

 

[razor1]

What if P10v1 and P10v2 were developed asynchronously

LOL That would be a magically feat!

If its a revision, which is highly unlikely, it would be some crazy capability AMD pulled out of its ass that no other company has been able to do, ever.

 

[thesmokingman]

 

[Ieldra]

LOL That would be a magically feat!

If its a revision, which is highly unlikely, it would be some crazy capability AMD pulled out of its ass that no other company has been able to do, ever.

Fermi v2 was just metal?

 

[razor1]

fermi v2 was mask I think, I will check it up later tonight, have to find pics of the die and serials.

 

[Ieldra]

fermi v2 was mask I think, I will check it up later tonight, have to find pics of the die and serials.

GTX 480 was GF100

570/580 use GF110-A1 so mask I assume, internally referred to as GF100B according to Wikipedia

 

[thesmokingman]

Nice, 145 comma 657 thousand watts. And the maximum is what one could only surmise would be 12 comma 526 comma 526 million watts? Hmm, either... er lol smh.

 

[lolfail9001]

Nice, 145 comma 657 thousand watts. And the maximum is what one could only surmise would be 12 comma 526 comma 526 million watts? Hmm, either... er lol smh.

Hint: the same sensor that reports these values also reports 95W power draw. So, lol smh.

 

[thesmokingman]

Hint: the same sensor that reports these values also reports 95W power draw. So, lol smh.

But wait it gets better at the voltage line. 94 comma 162 thousand volts! This thing is gonna melt a hole into the earths core.

 

[Ieldra]

But wait it gets better at the voltage line. 94 comma 162 thousand volts! This thing is gonna melt a hole into the earths core.

this is the best I could find, note 145W for current GPU power draw

[M] HIS RX 480 IceQ X2 Roaring Turbo 8GB Video Card Review

But by all means ManofGod thesmokingman don't let this stop you from turning this healthy skepticism into another one of your rants about how everyone is always against AMD, it certainly didn't stop you when he had the RX480 1600mhz "leaks" from the very same source.


Edit: scrap that, retarded readings

Edit 2: found this

HIS Radeon RX 480 IceQ X2 Roaring Turbo 8GB Review - VideoCardz.com

Reading is essential, as is not ignoring the perfectly valid data below it.

In my defense when I first read the power reading I thought it was in mW, even then it seems like all the readings are off by a scale factor of 1000. Even so, the GPU Core Voltage reported is 1.146v with current peaking at 133A. So that's 1340mhz 153W GPU only.

The other sensor data I posted corroborate this with ~180W average at 1410mhz. Again, GPU only.

 

[noko]

Good grief why are there two threads dealing with this??? Anyways looks like these lower power Polaris chips are hitting the market. Look at this review, it is showing this 50% power savings on the FX 480.

The OC of 1475mhz and the low temperature confirms the power being shown.

Hmmm I may shift from a 1060 to a Rx480 for my Daughters build if the 290 does not work in the Node 202 SFF case. Just need to make sure it is the low power version and not the initial batch of higher power lower capable clocked chips.

So can Polaris with the improved power characteristics make it to a 490 version with DDR5(x)? $349 MSRP?

 

[thesmokingman]

Reading is essential, as is not ignoring the perfectly valid data below it.

In my defense when I first read the power reading I thought it was in mW, even then it seems like all the readings are off by a scale factor of 1000. Even so, the GPU Core Voltage reported is 1.146v with current peaking at 133A. So that's 1340mhz 153W GPU only.

The other sensor data I posted corroborate this with ~180W average at 1410mhz. Again, GPU only.

Dude, don't take things so seriously omg. But man, yer still trying to fight some fight...? Its clear hwinfo needs updating for this card as its actually different from known cards. I'm not taking any of these readings or whatever seriously until a reputable source confirms it but what info that is out there does point to a very different pcb w/o a doubt and it definitely performs differently as well. Does that emphatically mean a new revision who knows? But imo it looks like there is something going on with the card that twocents got and that same pcb underpins the roaring. You can even compare twocents video findings against some of his other cards like the terribad G1 which hit 190w with a not great overclock of 1390 where the GTR at 1475 mhz only drew 145w something.

 

[Ieldra]

Dude, don't take things so seriously omg. But man, yer still trying to fight some fight...? Its clear hwinfo needs updating for this card as its actually different from known cards. I'm not taking any of these readings or whatever seriously until a reputable source confirms it but what info that is out there does point to a very different pcb w/o a doubt and it definitely performs differently as well. Does that emphatically mean a new revision who knows? But imo it looks like there is something going on with the card that twocents got and that same pcb underpins the roaring. You can even compare twocents video findings against some of his other cards like the terribad G1 which hit 190w with a not great overclock of 1390 where the GTR at 1475 mhz only drew 145w something.

You can also go look at customer experience with the XFX GTR card Jay2Cent reviewed and find that it's just silicon lottery, the PCB has nothing to do with it, and this PCB is still using the same VRM IC as the previous ones. Unless you think board losses cover all of this power consumption that is missing the PCB is totally unrelated.

Good grief why are there two threads dealing with this??? Anyways looks like these lower power Polaris chips are hitting the market. Look at this review, it is showing this 50% power savings on the FX 480.

The OC of 1475mhz and the low temperature confirms the power being shown.

So there's more than one of them? Because I've found nothing whatsoever

 

[lolfail9001]

Good grief why are there two threads dealing with this??? Anyways looks like these lower power Polaris chips are hitting the market. Look at this review, it is showing this 50% power savings on the FX 480.

The OC of 1475mhz and the low temperature confirms the power being shown.

Hmmm I may shift from a 1060 to a Rx480 for my Daughters build if the 290 does not work in the Node 202 SFF case. Just need to make sure it is the low power version and not the initial batch of higher power lower capable clocked chips.

So can Polaris with the improved power characteristics make it to a 490 version with DDR5(x)? $349 MSRP?

Half of this thread talks about how Jayz data needs verification to begin with. Then it mentions that supposedly "identical" card is the same power hog as every other 480 on the market, so it is silicon lottery, nothing else.

 

[Ieldra]

Half of this thread talks about how Jayz data needs verification to begin with. Then it mentions that supposedly "identical" card is the same power hog as every other 480 on the market, so it is silicon lottery, nothing else.

As much as I hate doing this


Also Anandtech's Ryan Smith explicitly stated there's no difference between the embedded boards and the mobile ones; it's just binning. So either
A). The XFX card in question uses binned dies
B). Jay guy is a lucky fucker
C). They sent him a golden sample

Anyway all of this discussion about the validity of the claims is superfluous imo because at the end of the day a move like this makes no sense. It would necessarily entail the existence of another P10 revision whose development overlapped with the original, which means AMD shafted early buyers with a POS batch.

Not only, but what logic would lead to them sitting on this and not releasing these into the mobile market ? You know, the market they were thrashed by Pascal in...

For this to be true AMD would have to be moving very counter productively as this rate, as their aim is to drive Polaris into the ground for real. They hemorrhaged sales with the RX480 power, starting with the issues on the reference cards, to most OEMs preferring the NV parts, total absence in mobile market. All of that would have been fixed by this purported new revision.

 

[thesmokingman]

Half of this thread talks about how Jayz data needs verification to begin with. Then it mentions that supposedly "identical" card is the same power hog as every other 480 on the market, so it is silicon lottery, nothing else.

You could be right for all I know. But that would have to be one helluva gold chip man as that GTR he got on air equals the clocks and power draw of the ref card with an ek block on it, almost exact draws 153w vs 145w at 1475mhz.

 

[lolfail9001]

You could be right for all I know. But that would have to be one helluva gold chip man as that GTR he got on air equals the clocks and power draw of the ref card with an ek block on it, almost exact draws 153w vs 145w at 1475mhz.

Well, XFX probably are aware of effect Jay's review of MSI R9 390 had on it's sales, so they made sure to cause a good impression

 

[razor1]

GTX 480 was GF100

570/580 use GF110-A1 so mask I assume, internally referred to as GF100B according to Wikipedia

Yep then its a mask respin.

 

[Darakian]

It is possible that the 14nm process has improved since the polaris launch and improved to the point where higher binned chips are simply more common.

 

[Ieldra]

It is possible that the 14nm process has improved since the polaris launch and improved to the point where higher binned chips are simply more common.

This is totally possible, but would entail the existence of higher binned chips previously

 

[noko]

You can also go look at customer experience with the XFX GTR card Jay2Cent reviewed and find that it's just silicon lottery, the PCB has nothing to do with it, and this PCB is still using the same VRM IC as the previous ones. Unless you think board losses cover all of this power consumption that is missing the PCB is totally unrelated.



So there's more than one of them? Because I've found nothing whatsoever

duh, should have checked earlier in the thread - sorry folks for same video posting. Still is very encouraging and I doubt one GPU is incredibly better then the rest of the batch, I do believe it is something more than that.

 

[razor1]

np it happens

anycase AMD would have to do a mask layer change to drop that kinda of power usage that kind of changes takes 2 Q's..... this kind of change that would drop power usage to such a degree can't not be done in a metal layer, so this is most likely no respin...... timing doesn't fit with such a change.

 

[Ieldra]

duh, should have checked earlier in the thread - sorry folks for same video posting. Still is very encouraging and I doubt one GPU is incredibly better then the rest of the batch, I do believe it is something more than that.

It isn't incredibly better than the rest of them though, I've seen other people with RX480 getting decent clocks at relatively low voltages, this is the equivalent of that one guy who manages 2300mhz on a GTX 1080.

Mark my words this is what happened.

AMD announce embedded radeon GPUs

Devious orcs over at WCCF seize chance and claim 50% more efficiency (150w vs 100w) and start rumor mill about respin

Mad clicks

 

[{NG}Fidel]

As much as I hate doing this


Also Anandtech's Ryan Smith explicitly stated there's no difference between the embedded boards and the mobile ones; it's just binning. So either
A). The XFX card in question uses binned dies
B). Jay guy is a lucky fucker
C). They sent him a golden sample

Anyway all of this discussion about the validity of the claims is superfluous imo because at the end of the day a move like this makes no sense. It would necessarily entail the existence of another P10 revision whose development overlapped with the original, which means AMD shafted early buyers with a POS batch.

Not only, but what logic would lead to them sitting on this and not releasing these into the mobile market ? You know, the market they were thrashed by Pascal in...

For this to be true AMD would have to be moving very counter productively as this rate, as their aim is to drive Polaris into the ground for real. They hemorrhaged sales with the RX480 power, starting with the issues on the reference cards, to most OEMs preferring the NV parts, total absence in mobile market. All of that would have been fixed by this purported new revision.

I think it just kinda proves that these cards are capable of hitting that 1400mhz range that you guy said was almost impossible before.
Re-spin this soon obviously not.

 

[razor1]

I think it just kinda proves that these cards are capable of hitting that 1400mhz range that you guy said was almost impossible before.
Re-spin this soon obviously not.

It was know rx480 could go higher then 1400 around launch time but just looking at how people talked about it namely WTFtech at that time, PS same author too as this article, gave false hope to many people that it was going to be a great overclocker from their ref. design. The power usage curve looks crazy, it climbs up so fast its easy to see its out of its ideal range. Even going past its base clocks of 1266 we see that happening. It just shows the core has been pushed way beyond what it should be doing. Anytime you see power curves like what you see with the rx480 with frequency increases, its the same thing, doesn't matter if its a CPU, GPU any silicon chip for that matter. Also once you are out of that ideal range, its not good for the chip too, cause leakage increases at an alarming rate and leakage increases the chance of chip death. This is probably why we see more RMA's on AMD's side too for the past 2 or 3 gens.

We saw this with Fermi v1 too, it also had RMA increased rates, and the r600, don't remember the FX series, then again never looked that up for them.

 

[Darakian]

This is totally possible, but would entail the existence of higher binned chips previously

Well take a look at the xfx 480 video posted. Beyond that, if a process improves over time it's possible for chips of higher bins period. The whole idea of process improvement is that the distribution of quality rises.

 

[Shintai]

Good grief why are there two threads dealing with this??? Anyways looks like these lower power Polaris chips are hitting the market. Look at this review, it is showing this 50% power savings on the FX 480.

The OC of 1475mhz and the low temperature confirms the power being shown.

Hmmm I may shift from a 1060 to a Rx480 for my Daughters build if the 290 does not work in the Node 202 SFF case. Just need to make sure it is the low power version and not the initial batch of higher power lower capable clocked chips.

So can Polaris with the improved power characteristics make it to a 490 version with DDR5(x)? $349 MSRP?

It doesn't get more right the more you post it. Why dont you show other sites, specially known respected review sites to show it

Or is it because the only source is a single youtube video?

 

[thesmokingman]

I think it just kinda proves that these cards are capable of hitting that 1400mhz range that you guy said was almost impossible before.
Re-spin this soon obviously not.

Ironic huh lol. Most there hit 1400 or higher. The reality is that a lot of customs cannot hit 1400. This pcb though doesn't look custom as in one offs by an AIB since both HIS and XFX share the same pcb. I found a pic of the underside of the cooler. There's something interesting about the vrm cooling that you don't see very often. The vrm block is integrated right into the fins of the main cooler with its own heatpipe. This makes logical sense because the metal mosfets run hotter than the not metal faced. The active air cooling seems to be working pretty well here. I wish I could get my hands on one.

 

[thesmokingman]

So 76 ºC and only 45% fan speed? At full load?

Well it seems Ive buyed the wrong card then, My sapphire Nitro +OC is as hot and loud as it is beautiful. Even reaching 80-82ºC with 90% fan speed :/

Some of these reddit comments are so interesting referring to the GTR. What makes the GTR's so much better?

I backed my OC down, I did hit 1450 core/2200mem @ 1150mv and 68C. This is my first decent build and don't want to damage the card so I now run 1350 core/2100 mem @ 1070mv and rarely hit 65C with stock fan curve. oh, XFX GTR XXX 8GB here...and I also cleaned and replaced thermal compound with AS Ceramique 2

 

[Ieldra]

Some of these reddit comments are so interesting referring to the GTR. What makes the GTR's so much better?

There are also comments saying they can't run it stably above 1390 no matter the voltage.

This is likely a card using binned GPUs, there have been good overclocking cards in the past as well, there's no new revision - if there were it would be reflected in the serial on the die

 

[ManofGod]

Its the exact same process.

Oh, so you are saying that Global Foundries is producing these cards?

 

[ManofGod]

Edit: Removed off topic stuff.