The 480: power consumption, PCI-E powerdraw

[MrTeal]

it would be interesting to know when this happened, when the 5th pin started to be used,


but I'm seeing from other posts some good news, if they deliver a driver fix in 2 days, this is the best case scenario I could think of :thumbsup:

Well, the oldest PCIe cards I have kicking around are a 5450 and a 5770, and they all have 5 fingers used for 12V on the card, so it's been awhile. Even older, I have a BIOSTAR N68S from 2007 that has all 5 commoned. I don't have anything left that's P4 vintage, but I'm guessing you'd need to go that far back to find one.

How CPU bottlenecked do you think a Pentium D 3.2GHz would be GTA V or Fallout 4?

 

[Bacon1]

AMD will never learn. If they can churn out a new driver with these fixes why didn't they do it before the release? Now the RX 480 already has a bad rep just like 290 did and reviews are done. Any performance increase will not be seen by people looking at reviews. Plus a press release with a typo: 3%1. hat is that supposed to mean? 31%?

Because they obviously wanted it to get bad reviews! Bugs and mistakes never happen...

The pros at Nvidia still haven't fixed the failure to boot for DVI 144hz monitors though, so guess they are just as bad.

 

[SolMiester]

Because they obviously wanted it to get bad reviews! Bugs and mistakes never happen...

The pros at Nvidia still haven't fixed the failure to boot for DVI 144hz monitors though, so guess they are just as bad.

That is actually a false statement and you know it. The NV cards are quite capable right now of running the DVI port within spec.

The upcoming driver update is for those that run out of spec on legacy technology.

 

[Seba]

Plus a press release with a typo: 3%1. hat is that supposed to mean? 31%?

That "1" should be a footnote.

See the Anandtech article.

 

[RussianSensation]

AMD will never learn. If they can churn out a new driver with these fixes why didn't they do it before the release? Now the RX 480 already has a bad rep just like 290 did and reviews are done. Any performance increase will not be seen by people looking at reviews. Plus a press release with a typo: 3%1. hat is that supposed to mean? 31%?

The bigger mistake by far was launching reference cards in the first place. Just look at Sapphire Nitro RX 480 and the reference card. There are many more benefits there than just the 8-pin power fix. So few RX 480 4GB cards on sale and no $149 RX 470 4GB on sale are also eye brow raising - "what were they thinking"? They could have easily launched reference blower 480 for the OEM market and released AIB 480 into the retail space. The card would have come out performing better, running cooler, quieter, overclocking better.

The more concerning aspect is 480's terrible performance/watt on 14nm. After AIBs fix all the issues with the reference card, we still have a card that comes nowhere close to AMD's perf/watt marketing BS claims and it sets a very pessimistic tone for Vega in the future. In the eyes of $300+ buyers, would you bother waiting for Vega at all when the excellent 1070/1080 are here now? The perf/watt damage from RX 480 will also ensure AMD gets mopped in the mobile dGPU space. If you want to talk about the greatest fail here, it's actually the entire Polaris 10/11 ASIC. Its only saving grace is the low price and that's partially to do with NV releasing the most overpriced garbage in the $150-250 space since GTX760. All it takes is for NV to price 1060 6GB at $239-249 and the entire RX 480 line is finished in the eyes of the vast majority of PC gamers who view AMD as a 2nd tier graphics brand. The gap in perf/watt continues to grow which paints a very bleak future for AMD's graphics division. One may even start to have doubts about Zen I'd 14nm FF from GloFo is this bad.

 

[antihelten]

That "1" should be a footnote.

See the Anandtech article.

It's kinda weird that they state "up to 3%", seeing as the highest improvement is actually 5.5% (total war) and the average improvement is 3.5%.

 

[3DVagabond]

That is actually a false statement and you know it. The NV cards are quite capable right now of running the DVI port within spec.

The upcoming driver update is for those that run out of spec on legacy technology.

Why don't they fix it for the people who own O/C'd monitors who bought their cards? Would it kill them? Would likely cost them a minimal amount to sort that.

 

[3DVagabond]

The bigger mistake by far was launching reference cards in the first place. Just look at Sapphire Nitro RX 480 and the reference card. There are many more benefits there than just the 8-pin power fix. So few RX 480 4GB cards on sale and no $149 RX 470 4GB on sale are also eye brow raising - "what were they thinking"? They could have easily launched reference blower 480 for the OEM market and released AIB 480 into the retail space. The card would have come out performing better, running cooler, quieter, overclocking better.

The more concerning aspect is 480's terrible performance/watt on 14nm. After AIBs fix all the issues with the reference card, we still have a card that comes nowhere close to AMD's perf/watt marketing BS claims and it sets a very pessimistic tone for Vega in the future. In the eyes of $300+ buyers, would you bother waiting for Vega at all when the excellent 1070/1080 are here now? The perf/watt damage from RX 480 will also ensure AMD gets mopped in the mobile dGPU space. If you want to talk about the greatest fail here, it's actually the entire Polaris 10/11 ASIC. Its only saving grace is the low price and that's partially to do with NV releasing the most overpriced garbage in the $150-250 space since GTX760. All it takes is for NV to price 1060 6GB at $239-249 and the entire RX 480 line is finished in the eyes of the vast majority of PC gamers who view AMD as a 2nd tier graphics brand. The gap in perf/watt continues to grow which paints a very bleak future for AMD's graphics division. One may even start to have doubts about Zen I'd 14nm FF from GloFo is this bad.

They want a $200 card. That's the reason. Better than setting a lower price for their AIB partners to hit like nVidia did. At least they are footing the expense. Also, all the 4GB reference cards are actually 8GB cards. Free RAM to go along with Free mining money.

 

[guskline]

I just read the AMD release mentioned by Silverforce above and my observations are that RTG is much better run now.

Despite RS's concerns which are legit to some discerning forum posters, I find the attitude of AMD is changing from oops ignore it to get it right.

What will be telling is if Tom's Hardware AND PC Perspective test the cards with the new software and verify the improvement.

This will be telling as to the perception of Vega.

My own decision to opt for a GTX 1080 instead of waiting for Vega was mostly fueled by the delay of Vega coupled with my experience with the GTX 980 TI.

The RX480 was and is focused on AMD bumping up its market share. From initial figures, it appears that AMD is off to a flying start.

 

[krumme]

The bigger mistake by far was launching reference cards in the first place.

Yep. And so incredible easy to foresee.

Regarding vega. Now ms might be paying most of the the dev cost but as i have said for months its a idiotic product from a business perspective. Stuck between cheaper non hbm 1080 and much faster big die 1080ti and should never have been made.
From a consumer perspective i wouldnt worry. If amd want to give you hbm2 and binned big die for cheap its their problem. They will not sell the card if its not compettitive on value.
Gf still needs to move those wafers to look good. Imo it will transfer into a value card that will be pretty futureproof because of consoles.

 

[StrangerGuy]

The more concerning aspect is 480's terrible performance/watt on 14nm. After AIBs fix all the issues with the reference card, we still have a card that comes nowhere close to AMD's perf/watt marketing BS claims and it sets a very pessimistic tone for Vega in the future. In the eyes of $300+ buyers, would you bother waiting for Vega at all when the excellent 1070/1080 are here now? The perf/watt damage from RX 480 will also ensure AMD gets mopped in the mobile dGPU space. If you want to talk about the greatest fail here, it's actually the entire Polaris 10/11 ASIC. Its only saving grace is the low price and that's partially to do with NV releasing the most overpriced garbage in the $150-250 space since GTX760. All it takes is for NV to price 1060 6GB at $239-249 and the entire RX 480 line is finished in the eyes of the vast majority of PC gamers who view AMD as a 2nd tier graphics brand. The gap in perf/watt continues to grow which paints a very bleak future for AMD's graphics division. One may even start to have doubts about Zen I'd 14nm FF from GloFo is this bad.

Nope, I'm not going to play the Vega waiting game when I getting serious Phenom 1 and 2900XT vibes all over again with Polaris. I don't ever remember a perf/watt uarch margin ever this large between NV and AMD/ATI.

And yes, the lucrative mobile dGPU business is definitely over for AMD.

 

[Wall Street]

it would be interesting to know when this happened, when the 5th pin started to be used,

On a 6-pin connector that top middle pin is unassigned. On an 8-pin connector it is assigned as +12V. Once PSU manufacturers started adding 6+2 pin and 8-pin connectors to the PSUs, it just made sense to connect the middle pin to +12V on all of the 6-pin connectors too. Card manufacturers also realized that there was no downside to connecting all three of the top pins together. In the best case you get three +12V pins which spreads the current loading, on the worst case the middle pin isn't there or is an open connection so it doesn't negatively impact the card in any way.

 

[Wall Street]

Because they obviously wanted it to get bad reviews! Bugs and mistakes never happen...

The pros at Nvidia still haven't fixed the failure to boot for DVI 144hz monitors though, so guess they are just as bad.

The funny thing is that this is the opposite problem. The DVI spec specifies a max pixel clock maximum of 330 Mhz. Previously nobody checked that monitors are within spec. The problem now is the NVidia cards check that the pixel clock is within spec upon bootup. For overclocked high-res high-refresh monitors, the highest settings violate the DVI standard. Now everyone wants NVidia to work on a driver that will ignor the DVI spec for max pixel clock.

 

[Despoiler]

AMD will never learn. If they can churn out a new driver with these fixes why didn't they do it before the release? Now the RX 480 already has a bad rep just like 290 did and reviews are done. Any performance increase will not be seen by people looking at reviews.

Probably because they didn't see any abnormalities in the samples they tested with. There is some production variance happening. That is why some cards are in spec and some are running loose to spec. Product design, development, and launch always has bugs to fix. Stating that AMD is the only company that releases fixes or they should have crystal balled the future is extraordinarily ignorant. I'm not referring solely to graphics cards either.

Maybe AMD should hire The_Stilt? He seems to know more about their products than they do...

He likely already does work for AMD or one of their AIBs. That guy knows a lot about every technical aspect of their products.

 

[parvadomus]

The bigger mistake by far was launching reference cards in the first place. Just look at Sapphire Nitro RX 480 and the reference card. There are many more benefits there than just the 8-pin power fix. So few RX 480 4GB cards on sale and no $149 RX 470 4GB on sale are also eye brow raising - "what were they thinking"? They could have easily launched reference blower 480 for the OEM market and released AIB 480 into the retail space. The card would have come out performing better, running cooler, quieter, overclocking better.

The more concerning aspect is 480's terrible performance/watt on 14nm. After AIBs fix all the issues with the reference card, we still have a card that comes nowhere close to AMD's perf/watt marketing BS claims and it sets a very pessimistic tone for Vega in the future. In the eyes of $300+ buyers, would you bother waiting for Vega at all when the excellent 1070/1080 are here now? The perf/watt damage from RX 480 will also ensure AMD gets mopped in the mobile dGPU space. If you want to talk about the greatest fail here, it's actually the entire Polaris 10/11 ASIC. Its only saving grace is the low price and that's partially to do with NV releasing the most overpriced garbage in the $150-250 space since GTX760. All it takes is for NV to price 1060 6GB at $239-249 and the entire RX 480 line is finished in the eyes of the vast majority of PC gamers who view AMD as a 2nd tier graphics brand. The gap in perf/watt continues to grow which paints a very bleak future for AMD's graphics division. One may even start to have doubts about Zen I'd 14nm FF from GloFo is this bad.

Well, Pascal architecture is not as great as Nvidia made a lot of people believe. If you take a look at aftermarket 1080s reviews its obvious NV sent higher binned chips for the FE edition review.
With Maxwell a lot of aftermarket cards were even more efficient than the reference version.
Had AMD sent binned chips for the 480, it would appear like 20/25% more efficient in reviews (in case you havent noticed, a lot of people have undervolt the 480s and reduced power consumption a good chunk). I bet it can be reduced further with better aftermarket coolers.

 

[littleg]

Well, Pascal architecture is not as great as Nvidia made a lot of people believe. If you take a look at aftermarket 1080s reviews its obvious NV sent higher binned chips for the FE edition review.

Yeah, that's just common sense. Which scores are everyone going to be using when comparing cards and making buying decisions? The binned review scores.

Sites aren't going to go back and retest the cards so those high binned scores are pretty much set in stone.

 

[railven]

Yeah, that's just common sense. Which scores are everyone going to be using when comparing cards and making buying decisions? The binned review scores.

Sites aren't going to go back and retest the cards so those high binned scores are pretty much set in stone.

That's the difference in a good marketing arm.

NV: play ball, send the best of the best to represent you in the list of reviews people will most likely see due to search engines and forum bickering.

AMD : whatever we got, just send to review sites, fan noise, power issues, 'ain't no one got time for that' - what do you mean the reviews are terrible?

There was a time when AMD...correction...ATI had glowing reviews. But AMD made sure to put an end to that, can't have their GPU sub-division outshine their CPU main-division.

 

[exar333]

The bigger mistake by far was launching reference cards in the first place. Just look at Sapphire Nitro RX 480 and the reference card. There are many more benefits there than just the 8-pin power fix. So few RX 480 4GB cards on sale and no $149 RX 470 4GB on sale are also eye brow raising - "what were they thinking"? They could have easily launched reference blower 480 for the OEM market and released AIB 480 into the retail space. The card would have come out performing better, running cooler, quieter, overclocking better.

The more concerning aspect is 480's terrible performance/watt on 14nm. After AIBs fix all the issues with the reference card, we still have a card that comes nowhere close to AMD's perf/watt marketing BS claims and it sets a very pessimistic tone for Vega in the future. In the eyes of $300+ buyers, would you bother waiting for Vega at all when the excellent 1070/1080 are here now? The perf/watt damage from RX 480 will also ensure AMD gets mopped in the mobile dGPU space. If you want to talk about the greatest fail here, it's actually the entire Polaris 10/11 ASIC. Its only saving grace is the low price and that's partially to do with NV releasing the most overpriced garbage in the $150-250 space since GTX760. All it takes is for NV to price 1060 6GB at $239-249 and the entire RX 480 line is finished in the eyes of the vast majority of PC gamers who view AMD as a 2nd tier graphics brand. The gap in perf/watt continues to grow which paints a very bleak future for AMD's graphics division. One may even start to have doubts about Zen I'd 14nm FF from GloFo is this bad.

I completely agree...

This very likely will result in many people waiting for Vega to just grab a 1070/1080 in 2016.

The 480 still is compelling <$200, but that is about it. This is very likely why the mGPU P10/P11 is totally MIA at this point.

Disappointing is the best word I can use to describe this release and this is worrisome for Zen as well. Still trying to be optimistic though...

 

[sirmo]

The bigger mistake by far was launching reference cards in the first place. Just look at Sapphire Nitro RX 480 and the reference card. There are many more benefits there than just the 8-pin power fix. So few RX 480 4GB cards on sale and no $149 RX 470 4GB on sale are also eye brow raising - "what were they thinking"? They could have easily launched reference blower 480 for the OEM market and released AIB 480 into the retail space. The card would have come out performing better, running cooler, quieter, overclocking better.

The more concerning aspect is 480's terrible performance/watt on 14nm. After AIBs fix all the issues with the reference card, we still have a card that comes nowhere close to AMD's perf/watt marketing BS claims and it sets a very pessimistic tone for Vega in the future. In the eyes of $300+ buyers, would you bother waiting for Vega at all when the excellent 1070/1080 are here now? The perf/watt damage from RX 480 will also ensure AMD gets mopped in the mobile dGPU space. If you want to talk about the greatest fail here, it's actually the entire Polaris 10/11 ASIC. Its only saving grace is the low price and that's partially to do with NV releasing the most overpriced garbage in the $150-250 space since GTX760. All it takes is for NV to price 1060 6GB at $239-249 and the entire RX 480 line is finished in the eyes of the vast majority of PC gamers who view AMD as a 2nd tier graphics brand. The gap in perf/watt continues to grow which paints a very bleak future for AMD's graphics division. One may even start to have doubts about Zen I'd 14nm FF from GloFo is this bad.

How closely have you looked at those power numbers? It is greater than 2x improvement over Tonga based on my estimation.

And that 285 is a 2gb version.. so rx480's 8Gb of VRAM adds quite a bit more power consumption. Also when I play overwatch my GPU uses something like 75 watts on avg. according to GPUZ..

 

[MrTeal]

The 285 is a poor comparison. For whatever reason it launched with much poorer efficiency than later full chips.

 

[sirmo]

The 285 is a poor comparison. For whatever reason it launched with much poorer efficiency than later full chips.

There is also a 280X.. it's the best comparison we have as comparing bigger chips which have a more favorable performance contributing transistor / uncore transistor ratio is not fair either (see 960 vs 980 or even 970 vs 980 (since they are the same chip.. it's about core/uncore transistor ratios).

 

[R0H1T]

The more concerning aspect is 480's terrible performance/watt on 14nm. After AIBs fix all the issues with the reference card, we still have a card that comes nowhere close to AMD's perf/watt marketing BS claims and it sets a very pessimistic tone for Vega in the future. In the eyes of $300+ buyers, would you bother waiting for Vega at all when the excellent 1070/1080 are here now? The perf/watt damage from RX 480 will also ensure AMD gets mopped in the mobile dGPU space. If you want to talk about the greatest fail here, it's actually the entire Polaris 10/11 ASIC. Its only saving grace is the low price and that's partially to do with NV releasing the most overpriced garbage in the $150-250 space since GTX760. All it takes is for NV to price 1060 6GB at $239-249 and the entire RX 480 line is finished in the eyes of the vast majority of PC gamers who view AMD as a 2nd tier graphics brand. The gap in perf/watt continues to grow which paints a very bleak future for AMD's graphics division. One may even start to have doubts about Zen I'd 14nm FF from GloFo is this bad.

I'll just put this here ~

Aging (maturity) of the process can change things drastically. The 32nm SHP SOI is one of the best examples. It started as a complete lemon with some huge issues with the leakage and the general inconsistency. In it's current form the 32nm SHP SOI is a completely different animal. Most likely one of the best processes ever made. The silicon consistency is immaculate and the average static leakage on full Vishera dies is up to 45% lower than it was at the time of launch.

The only trouble is that it took around three years for it to reach it's peak D:

So the 14nm LPP will most likely still improve heavily as the machinery and the process itself grow older, however the only question is will the improvement be enough. If Zeppelin would ship on as immature process as Polaris 10 did, it would be a complete disaster. Zeppelin is obviously somewhat smaller than Polaris 10, but not by a huge margin. Personally I would rather see Zeppelin to be delayed slightly than to be released on the 14nm LPP in it's current state.

There's clearly a lot that can be done with 14nm, even adjusting power target, driver fixes et al can net you 20~40W savings, from the worst case scenario. This just from software tweaks, so I have no doubt that the upcoming iterations will be so much better; also next gen consoles on GCN4 makes the current Polaris 480 (4<=>8 GB) arguably the best perf/$ card in quite a while.

The 470 could change all of this & be the card to get, perf/$ & perf/mm^2 as well as perf/W wise, but there's no doubt that this was a lemon launch however you have to consider that the best quality stuff went to Apple, consoles, OEM/AIB's just to name a few.

 

[MrTeal]

There is also a 280X.. it's the best comparison we have as comparing bigger chips which have a more favorable performance contributing transistor / uncore transistor ratio is not fair either (see 960 vs 980 or even 970 vs 980 (since they are the same chip.. it's about core/uncore transistor ratios).

Yeah, but the 380X and 285 are the same chip, the 285 is just a cut down version of it. For whatever reason (sending the best binned ones to Apple, perhaps) the 285 just used a lot of power. As the 480 is also a full chip, it's more natural to compare it to full Tonga.

 

[krumme]

Nope, I'm not going to play the Vega waiting game when I getting serious Phenom 1 and 2900XT vibes all over again with Polaris. I don't ever remember a perf/watt uarch margin ever this large between NV and AMD/ATI.

And yes, the lucrative mobile dGPU business is definitely over for AMD.

Lets wait and see. 480 is selling very well. Cost is important also for mobile.

 

[sirmo]

Yeah, but the 380X and 285 are the same chip, the 285 is just a cut down version of it. For whatever reason (sending the best binned ones to Apple, perhaps) the 285 just used a lot of power. As the 480 is also a full chip, it's more natural to compare it to full Tonga.

I agree, but we don't have that direct comparison.

Maybe it could be extrapolated from this:

So it's still a substantial gain in performance per watt. Considering the more VRAM.. I still think it gets close to 2x.