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

S

SighTurtle

[H]ard|Gawd
Joined
Jul 29, 2016
Messages
1,410
AMD Rolling Out New Polaris GPU Revisions With 50% Better Perf/Watt
AMD has reportedly started rolling out new revisions of its Polaris 10 & Polaris 11 GPUs with a 50%+ improvement in performance per watt. Polaris 10 and Polaris 11 are AMD’s latest GPUs powering the company’s midrange RX 480 & RX 470 as well as mainstream RX 460 graphics cards. The new GPU revisions will reportedly debut first in the embedded market and will deliver the same and/or slightly better performance compared to the previous revisions but at significantly lower power consumption.

This significant improvement in power efficiency has been attributed to an improvement to the 14nm metal mask layers of both Polaris GPUs by AMD in combination with a more refined binning process. The new revision of Polaris 10 that’s will go into the embedded market will bring down the typical board power from 150W as is found in the RX 480 to less than 95W. The updated revision of Polaris 11 will not only bring down the power from 75W to less than 50W but it will also improve clock speeds and bring up the compute throughput to 2.5 TFLOPS from 2.15.

The new revisions will be available immediately to the embedded market under the E9550 and E9260 product names for Polaris 10 and 11 respectively. The updated revisions are believed to be coming to the mobile market in the form of mobility Radeon RX 400 series graphics cards in the coming months. It’s not clear yet whether AMD will also introduce the new revision to the desktop market. However, in the mobile market where power efficiency rules supreme this could prove to be a pivotal change for the company’s competitiveness in notebooks.
Click to expand...

Crosspost from Overclock.net. Also rumor so bag of salt.
 
Well it is quite believable. Look at the new XFX RX 480 GTR. Start the video from the 5m 26s mark and watch the power consumption in comparison to the other RX 480. I have to dump voltage into my RX 480 to OC it and it's in a full water loop. This XFX card with a mild factory OC was pulling ~100 watts. Overclocked to 1475 and it only pulled a peak or 149w @1.1v and hit 60c. Factory voltage is 1.05v. The reference RX 480 cards started off at 150w. For mine to get to 1475 I need almost 1.3v and it is hit or miss stable.

 
lolfail9001 said:
Oh heck, they just noticed the embedded Polaris card announcements have seen TDP on them and concluded that they will have exact same performance.

WCCFTech in nutshell, boys.
Click to expand...

omg I completely forgot that the embedded parts were announced.
 
Looks like clickbait ;)

But I bet they make a fortune on people clicking their crap.
 
Aside from the fact that this rumor is even more ridiculous than the "GTX 2000 refresh" and WCCF is making a lot of assumptions about embedded chips... I hope it's true because it means AMD will finally fix their power efficiency problems and I'd love to see a return to 6 month product cycle.

50% efficiency improvement in less than a year represents a fuck up earth-shattering magnitude on AMD's part, probably rushing Polaris out the door to compete with Pascal. I don't think Fermi even pulled those numbers.
 
TaintedSquirrel said:
Aside from the fact that this rumor is even more ridiculous than the "GTX 2000 refresh" and WCCF is making a lot of assumptions about embedded chips... I hope it's true because it means AMD will finally fix their power efficiency problems and I'd love to see a return to 6 month product cycle.

50% efficiency improvement in less than a year represents a fuck up earth-shattering magnitude on AMD's part, probably rushing Polaris out the door to compete with Pascal. I don't think Fermi even pulled those numbers.
Click to expand...

No, just means refinement of process. Remember that 14nm is new and nobody's been doing it before. These are the types of gains that you should expect from all companies over time. If you want to make a car analogy think about gas usage from the first cars to the 1960's, 1990's, and then today. The first cars won't be remembered 10,000 years from now as complete fails; they will be remembered as ushering in a new age. Just like 14nm in CPU / GPU design is a new milestone.

Yes, I been pissed about my card after watching that video I linked 5 days ago. Thought about it more after reading the article on what happened, and I'm happy now that there was a possible breakthrough. To me it just means that Vega and Zen might have even more refinements in their process. When I hand this PC down to my grand niece, it may have a reference RX 480 in it instead HD7950 Crossfire. :)
 
14nm isn't new. And there is no 50% improvement laying waiting. I know people are desperate for a quick fix, but lets stay in the real world. Instead of being fooled by click baits and misinformation only to be let down yet again.
 
A lot of weird posts on this forum the last few days. Is 14nm the first "new node nobody has ever done before"? Has the GPU industry ever seen a 50% efficiency improvement from a refresh in <1 year? This is uncharted territory, isn't it?

The Pascal refresh leaks cited a 20% perf improvement across the board (1070 -> 2070, etc). It's odd that AMD would focus on efficiency when we already saw how it backfired on them with the "2.8x efficiency" before the RX 480's launch.

Shintai said:
14nm isn't new. And there is no 50% improvement laying waiting. I know people are desperate for a quick fix, but lets stay in the real world.
Click to expand...
Of course not, the whole thing is bunk. Same goes for the Nvidia side.
 
Just rewind the same hype train. RX 480 was to be 2560-3072SP, 1500-1600Mhz and using 100W. And it all comes from the same people that always works backwards. Here is what we need to beat, how do we do it? And this is how fairies and unicorns gets created.
 
The source is this image:

Embedded-Radeon-Polaris-04.png
 
TaintedSquirrel said:
A lot of weird posts on this forum the last few days. Is 14nm the first "new node nobody has ever done before"? Has the GPU industry ever seen a 50% efficiency improvement from a refresh in <1 year? This is uncharted territory, isn't it?

The Pascal refresh leaks cited a 20% perf improvement across the board (1070 -> 2070, etc). It's odd that AMD would focus on efficiency when we already saw how it backfired on them with the "2.8x efficiency" before the RX 480's launch.


Of course not, the whole thing is bunk. Same goes for the Nvidia side.
Click to expand...

I can say that there is a helluva efficiency boost going from a R9 290 to a RX 480. I think it's hilarious how much smaller it is than the R9 290. I doubt if it's much bigger than a Nano when you take the reference cooling skirt off it. It isn't long enough to line up with the memory on my motherboard.

So I was thinking of keeping the loop and getting her one of those Alphacool units that [H]ardocp tested. Then stuffing everything into a Mini ITX case, but then I would feel pressured to get her new processor. :)
 
Last edited:
TaintedSquirrel I also think that efficiency for mobile is the number one reason that AMD is working on this. There will be a war raged for the right to be in the new Apple ultra thin notebooks. This 14nm process being used is completely new as far as I know. There are supposed to be refinements and discoveries. Remember when every couple of years from the beginning of CPU development until 5 years ago there was a leap of power due to refinements and discovery? Moore's Law meant something. Maybe we're seeing that at the 14nm process node. Whatever it is I could care less as long as it makes stuff OC better.

Now if AMD and Intel could get a 50% boost to something in the CPU space..... 5 damn years of CPU development and no increase in performance sucks!
 
cageymaru said:
Now if AMD and Intel could get a 50% boost to something in the CPU space..... 5 damn years of CPU development and no increase in performance sucks!
Click to expand...

There has been something like 600% perf/watt increase for CPUs the last 5 years. But its targeted for a different segment. Mobile and servers.
 
cageymaru said:
Well it is quite believable. Look at the new XFX RX 480 GTR. Start the video from the 5m 26s mark and watch the power consumption in comparison to the other RX 480. I have to dump voltage into my RX 480 to OC it and it's in a full water loop. This XFX card with a mild factory OC was pulling ~100 watts. Overclocked to 1475 and it only pulled a peak or 149w @1.1v and hit 60c. Factory voltage is 1.05v. The reference RX 480 cards started off at 150w. For mine to get to 1475 I need almost 1.3v and it is hit or miss stable.

Click to expand...


I totally enjoy watching his videos because, for some reason, he always manages to make me laugh. :D To those who are clearly Nvidia fboys, nothing AMD could ever do would be good. Glad I choose what works well for me and not what they think I should buy. ;)

It will be no surprise that power efficiency will improve dramatically. The only question is how dramatic is anyone guess.
 
Shintai said:
There has been something like 600% perf/watt increase for CPUs the last 5 years. But its targeted for a different segment. Mobile and servers.
Click to expand...

That is nice but not what he was referring too. Who cares if there was a perf/watt increase when what you want is greater performance from your desktop computer, period.
 
cageymaru said:
Saw this on Reddit. So the current RX 480 cards average 164w. The card in the video above was pulling less than 95w for a great deal of the time with a mild factory OC. At one point it was ~85w.
Die (realen) Stromverbrauchswerte von AMDs Polaris und nVidias Pascal | 3DCenter.org


Click to expand...

Sensor reading of the GPU isn't the same. I thought AMD people knew it wasn't board power. Some people thought RX 480 would use 100-110W before release for the same reason.

Dont tell me all that hype is based on that and the reviewer got no clue how much the card actually uses.
 
Last edited:
Like I said, weird posts on this forum recently.
The RX 480 magically went from a 160W+ card to 95W according to one review from... Jay? The Nvidia shill?

"JayzTwoCents Are Questionable"
"JayzTwoCents spreading misinformation about the 4GB RX 480 when comparing prices to the 6GB GTX 1060 even-though benchmarks prove otherwise (benchmarks in the comments)"
"JayzTwoCents's response to AMD being able to the pcie issue with the 480 (proving him wrong)"
"so I guess JayzTwoCents was wrong"
"Jayztwocents was called out on lying about the Nvidia's 1080 throttling. This is what ended up happening."

This after we have literally dozens of reviews and outrage from the community about AMD misleading consumers with their marketing (GPU vs Board power).
Isn't it more likely Jay botched his review? Or is trying to regain views and trust from the AMD community?

I feel like I've crossed into another dimension this week. What is going on around here? This place used to be the most level-headed and informed message board for hardware discussion. Wow... I'm a little disappointed.
 
This is what a reference RX 480 at 85C uses from the sensor reading of the GPU only. And that pretty much translates into 150-165W in real usage.

GPU-Z-thermals.gif
 
Guys on Reddit have been bragging about how low they can clock their newer RX 480 cards and still OC to 1400+. I never tried because... I like more speed 100% of the time.
 
cageymaru said:
Guys on Reddit have been bragging about how low they can clock their newer RX 480 cards and still OC to 1400+. I never tried because... I like more speed 100% of the time.
Click to expand...
AMD sub brags about a lot of things that aren't necessarily true, we know this already.
Apparently they think they know more about binning GPUs than AMD does.
 
So if true, we're saying this new 480 is the same speed (assuming no magical overclocking headroom), but consumes 50% less juice.

So what? Were hordes of people complaining about a 150W consumption and demanding 100W? I mean, I know this would keep temps a bit lower, and probably save you literally pennies a year in power bills. I think people would be more than happy with a 150W card that had 10-20% more performance rather than 10-20C lower temps.

Unless you're talking a mobile GPU anyway.
 
It's a refresh so it will come with a clock bump.

Bandalo said:
Were hordes of people complaining about a 150W consumption and demanding 100W?
Click to expand...
Pretty much, yeah. If they can re-release the 480 and drop it down to 100W then it's a big deal.
 
Bandalo said:
So if true, we're saying this new 480 is the same speed (assuming no magical overclocking headroom), but consumes 50% less juice.

So what? Were hordes of people complaining about a 150W consumption and demanding 100W? I mean, I know this would keep temps a bit lower, and probably save you literally pennies a year in power bills. I think people would be more than happy with a 150W card that had 10-20% more performance rather than 10-20C lower temps.

Unless you're talking a mobile GPU anyway.
Click to expand...

It also translates into noise, climate (room), PSU size/draw, component costs and so on. If you trade power for performance it can be a different story. But assume same performance, you really want the lower power usage.
 
TaintedSquirrel said:
It's a refresh so it will come with a clock bump.


Pretty much, yeah. If they can re-release the 480 and drop it down to 100W then it's a big deal.
Click to expand...

How so? I understand it'll be an improved product, but can they release a "new" 480 at the same price point as their just-released 480 with the only improvement being lower power use? Who's going to buy it? Existing 480 users won't upgrade just to save 50W and gain no performance.

Shintai said:
It also translates into noise, climate (room), PSU size/draw, component costs and so on. If you trade power for performance it can be a different story. But assume same performance, you really want the lower power usage.
Click to expand...

Honestly, 50W is pretty insignificant in this context. No one was looking at a 480 and saying, "yeah, I want it, but my current PSU can't handle a 150W card, so I'll wait until next year"..And while this MIGHT translate into lower fan speeds and less noise, most aftermarket cards already had that covered, so I doubt there'll be an appreciable difference there.

Now, if I was buying a new card, and had a 480 and NEW 480, and the only difference was the new one was 100W vice 150W, sure, I'd pick the "new" version. But I don't think that 50W would even start to factor into a decision otherwise.

I'd imagine it'd piss off the aftermarket companies too, because now they have an over-designed cooler.
 
Bandalo said:
Honestly, 50W is pretty insignificant in this context. No one was looking at a 480 and saying, "yeah, I want it, but my current PSU can't handle a 150W card, so I'll wait until next year"..And while this MIGHT translate into lower fan speeds and less noise, most aftermarket cards already had that covered, so I doubt there'll be an appreciable difference there.

Now, if I was buying a new card, and had a 480 and NEW 480, and the only difference was the new one was 100W vice 150W, sure, I'd pick the "new" version. But I don't think that 50W would even start to factor into a decision otherwise.

I'd imagine it'd piss off the aftermarket companies too, because now they have an over-designed cooler.
Click to expand...

Well the 50W doesn't exist to begin with. But it does matter a lot more than people think. Its also easier to overclock if you use 50W less to begin with. And for me, 50W is a lot as such. My GTX 1080 is about the limit of what I would get. I wouldn't get a Titan X/1080TI. Just as I wouldn't get a HEDT CPU over the regular LGA11xx. Because it does matter. Specially if you dont have any AC.

In terms of power cost, it depends. If you are a very dedicated gamer and game 6 hours per day as a hypothetical example. 50W in my country turns into 40-45$ a year.

AIBs pissed? Have you seen some of the sub 75W cards coolers? ;)
 
Bandalo said:
How so? I understand it'll be an improved product, but can they release a "new" 480 at the same price point as their just-released 480 with the only improvement being lower power use? Who's going to buy it? Existing 480 users won't upgrade just to save 50W and gain no performance.



Honestly, 50W is pretty insignificant in this context. No one was looking at a 480 and saying, "yeah, I want it, but my current PSU can't handle a 150W card, so I'll wait until next year"..And while this MIGHT translate into lower fan speeds and less noise, most aftermarket cards already had that covered, so I doubt there'll be an appreciable difference there.

Now, if I was buying a new card, and had a 480 and NEW 480, and the only difference was the new one was 100W vice 150W, sure, I'd pick the "new" version. But I don't think that 50W would even start to factor into a decision otherwise.

I'd imagine it'd piss off the aftermarket companies too, because now they have an over-designed cooler.
Click to expand...


There is an entire 30+ page thread here detailing why the AMD RX 480 is a bad card because it pulls 150w and was going to burn down homes. :)

With that said I like the power savings because it potentially means higher overclocks.
 
Shintai said:
Well the 50W doesn't exist to begin with. But it does matter a lot more than people think. Its also easier to overclock if you use 50W less to begin with. And for me, 50W is a lot as such. My GTX 1080 is about the limit of what I would get. I wouldn't get a Titan X/1080TI. Just as I wouldn't get a HEDT CPU over the regular LGA11xx. Because it does matter. Specially if you dont have any AC.

In terms of power cost, it depends. If you are a very dedicated gamer and game 6 hours per day as a hypothetical example. 50W in my country turns into 40-45$ a year.

AIBs pissed? Have you seen some of the sub 75W cards coolers? ;)
Click to expand...

I agree with your point on overclocking. IF (and this is the problem here), that lower consumption gives you more overclocking headroom, then yes, it's a great deal. But most of the "leaked" info doesn't support that theory. Seems like in this new spin, you get lower power use, but no additional overclocking headroom. I'd honestly bet if you ran a 100W and 150W card in your machine, you wouldn't see any measurable different in temps, regardless of your AC status. It's just not that much difference when you look at the whole picture. I can't speak to power costs in your country, so I'll defer to your math there.

Well, AIBs have two choices with a re-spun 480. Redesign the whole card and cooler (and eat the design costs), or re-use the existing design (and eat the cost of the over-built cooler and power stages). All for a card that will, in theory, occupy the same price/performance range.

cageymaru said:
There is an entire 30+ page thread here detailing why the AMD RX 480 is a bad card because it pulls 150w and was going to burn down homes. :)

With that said I like the power savings because it potentially means higher overclocks.
Click to expand...

The original 480 issue wasn't with TOTAL power being drawn, it was all about WHERE the power was being drawn. You can draw all the power you want, just don't pull more than 75W from the slot. I'm sure the actual risk was way overblown since most motherboard manufacturers have a huge safety margin in their designs. But that was fixed with a software patch regardless.

Power savings USUALLY imply higher overclocks. But these new AMD designs haven't been stellar overclockers, and it's not temp or power limits.
 
cageymaru said:
Well it is quite believable. Look at the new XFX RX 480 GTR. Start the video from the 5m 26s mark and watch the power consumption in comparison to the other RX 480. I have to dump voltage into my RX 480 to OC it and it's in a full water loop. This XFX card with a mild factory OC was pulling ~100 watts. Overclocked to 1475 and it only pulled a peak or 149w @1.1v and hit 60c. Factory voltage is 1.05v. The reference RX 480 cards started off at 150w. For mine to get to 1475 I need almost 1.3v and it is hit or miss stable.

Click to expand...


I love his review using temps on a open test bench. Lol
 
cageymaru said:
Well it is quite believable. Look at the new XFX RX 480 GTR. Start the video from the 5m 26s mark and watch the power consumption in comparison to the other RX 480. I have to dump voltage into my RX 480 to OC it and it's in a full water loop. This XFX card with a mild factory OC was pulling ~100 watts. Overclocked to 1475 and it only pulled a peak or 149w @1.1v and hit 60c. Factory voltage is 1.05v. The reference RX 480 cards started off at 150w. For mine to get to 1475 I need almost 1.3v and it is hit or miss stable.

Click to expand...


It makes sense that they would or could trim the fat, especially on a new node. It seems to be a more effective revision than last time with 280x/390x. This should force a price drop on existing Polaris silicon. Glad I didn't buy Polaris just yet.


**Lmao, pos... @ getting hit with a windows update. Whose had that happen to them more than they care to? Raises hand!


jmilcher said:
I love his review using temps on a open test bench. Lol
Click to expand...

Here's a tissue.


Btw, I saw the pcb. It looks like AMD switched to ir metal mosfets seen in their higher end gpus. Apparently this is the same pcb layout as used by HIS so it's looking like an AMD design instead of a one-off custom.

XFX
1470647963_911_Approaching-accelerator-XFX-Radeon-RX-480-GTR.jpg


HIS
HIS-RX-480-ICEQX2-Roaring-Turbo-pcb-900x419.jpg


**Also, overclocking reports on the HIS Roaring are over 1400 on stock volts too so whatever was holding the OG release cards back doesn't seem to be a big issue anymore.

Power.png


^^From bjorn3d. What's interesting is that the Roaring which in HIS form is the ugliest gpu ever, is 13 watts from the Strix 1060.

Comparatively...

power_maximum.png
 
Last edited:
Ryan Smith - Anandtech

Ryan Smith - Anandtech said:
Tuesday, September 27, 2016

Keeping in mind that these are derived from the mobile parts, which are binned to begin with, it's likely that there's more aggressive throttling.

AMD Announces Embedded Radeon E9260 & E9550 - Polaris for Embedded Markets
Click to expand...

The only news outlet that mentions a new revision is WCCF, and now everyone is repeating it.

Maybe JayZ guy got super lucky with his sample, maybe the readings are incorrect either way, there's nothing to indicate Polaris is being replaced with better Polaris just yet.
 
Ieldra said:
Maybe JayZ guy got super lucky with his sample, maybe the readings are incorrect either way, there's nothing to indicate Polaris is being replaced with better Polaris just yet.
Click to expand...

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.
 
You must log in or register to reply here.
Back
Top