Nvidia/RTG Powergate

Ah that was only part of it, the number of ROP's were not what they said they were.

But again they will get their just deserts for what they did that's what the lawsuits are all about. Multiple states multiple jurisdictions.

In this case AMD can do a simple fix and the problem will go away.
 
Presbytier said:
i really don't want to defend Nvidia in ramgate, but here goes. The card had 4gb of ram period no lie there, end of story; what they did was use slower ram for the last 512mb which certainly is misleading but is not the same as what AMD may be doing here.
Click to expand...

There were other specs misrepresented... IIRC bandwidth, rops, some other crap.

On the 480 topic, it'd be interesting if one of these sites took FLIR images of a mobo at 75W and 100W. I don't think there'd be a ton of difference. Maybe they'd need quad fire 480s :)
 
razor1 said:
Ah that was only part of it, the number of ROP's were not what they said they were.

But again they will get their just deserts for what they did that's what the lawsuits are all about. Multiple states multiple jurisdictions.

In this case AMD can do a simple fix and the problem will go away.
Click to expand...
My questions is, can drivers update the BIOS?
If not, are they expecting users to update the BIOS?

If the latter, they're better off issuing a RMA.
 
Presbytier said:
i really don't want to defend Nvidia in ramgate, but here goes. The card had 4gb of ram period no lie there, end of story; what they did was use slower ram for the last 512mb which certainly is misleading but is not the same as what AMD may be doing here.
Click to expand...

Man, 1000 years from now aliens will visit Earth and the first message they get will be "The 970 only had 3.5GB of RAM!!!111"
 
Well PCPer wrote about it, and they did ask a motherboard manufacture and they did say if its a sustained amount over the pci-e specs it could affect pins and connectors.

Power Consumption Concerns on the Radeon RX 480 | Overclocking, Current Testing

I asked around our friends in the motherboard business for some feedback on this issue - is it something that users should be concerned about or are modern day motherboards built to handle this type of variance? One vendor told me directly that while spikes as high as 95 watts of power draw through the PCIE connection are tolerated without issue, sustained power draw at that kind of level would likely cause damage. The pins and connectors are the most likely failure points - he didn’t seem concerned about the traces on the board as they had enough copper in the power plane to withstand the current.
Click to expand...
 
WorldExclusive said:
My questions is, can drivers update the BIOS?
If not, are they expecting users to update the BIOS?

It the latter, they're better of issuing an RMA.
Click to expand...


Well I they can do it in drivers and bios by downclocking, if they don't want to down clock they have to put more power connectors or larger 8 pin and then do some modifications as well in the driver and bios. So its up to the way AMD wants to peruse and fix it.
 
Presbytier said:
i really don't want to defend Nvidia in ramgate, but here goes. The card had 4gb of ram period no lie there, end of story; what they did was use slower ram for the last 512mb which certainly is misleading but is not the same as what AMD may be doing here.
Click to expand...

lol lie is a lie. So what its not the first or last card to pull more power than its designed for. look in the past history bunch of cards have done this.

They sold shit load of them as 4gb cards when only 3.5 can be used and they knowingly sold it. Yea it had 4gb but that is like hiding behind a box. Its the principal you don't knowingly sell less of a product than you are promising. Here it is not less of a product it is using more power than designed. It's worse when you find out after months when they sold so many, what would have happened if people doing reviews actually noticed it and if it came out within weeks of launch. There would have been many more pissed people.

Like I said people accept losses once they have had the product for a while. It's like whatever. I would do the same because no wants to sit there and stress about it. whats done is done type of thing you know.
 
NKD said:
LOL recall? They can easily adjust the boost clocks. Boost clocks are not promised remember? I think you are just thinking way ahead of yourself. They can easily say the boost clocks were adjusting according to power draw, bamn end of story.
Click to expand...

So they marketed the card at 1266MHz typical boost clock, and now they silently reduce that to, say, 1200. You know what they'll get? Class action lawsuits. They intentionally lied about the ability of the card to buyers.
 
razor1 said:
Well I they can do it in drivers and bios by downclocking, if they don't want to down clock they have to put more power connectors or larger 8 pin and then do some modifications as well in the driver and bios. So its up to the way AMD wants to peruse and fix it.
Click to expand...

I think we will see boost clock adjusted probably in scenarios it approaches 150w. I think thats the most likely solution for cards that are already out. May be next batch put a 8 pin connector on there.
 
trandoanhung1991 said:
So they marketed the card at 1266MHz typical boost clock, and now they silently reduce that to, say, 1200. You know what they'll get? Class action lawsuits. They intentionally lied about the ability of the card to buyers.
Click to expand...

Nope. You don't understand how boost clock works? It doesn't have to stay at 1266 sustained all the time, anything higher than base clock is boost clock. 1266 is max boost clock, you know what I mean. so it could stay at 1266 depending on how much a game is taxing the card. It may boost less in more taxing games. But going to 1266 and clocking down to 1220 or so when it needs to stay under the power envelope.

Nvidia cards do this really well. They try very hard to stay under set power limits. The card will go way below the boost clock but always above the base clock to stay under power limits.
 
NKD said:
I think we will see boost clock adjusted probably in scenarios it approaches 150w. I think thats the most likely solution for cards that are already out. May be next batch put a 8 pin connector on there.
Click to expand...

yeah its a simple fix, nothing that would break AMD, since it was caught early on, if AMD lets it go and things start failing its would be a major catastrophe for them later on, cause not only they will get pulled in any one that is part of the PCI-SIG process will to, and it will come down to individual people within each corporation or group as well.

Plus AMD might not have been aware of this problem, just a simple bios update near launch could have caused this problem, which they didn't have time to test and just assumed its ok.

Added to this the difference between the 4gb to the 8gb, the wattage might be different! so there is another reason why they might have missed it.
 
  • Like
Reactions: NKD
like this
NKD said:
lol lie is a lie. So what its not the first or last card to pull more power than its designed for. look in the past history bunch of cards have done this.

They sold shit load of them as 4gb cards when only 3.5 can be used and they knowingly sold it. Yea it had 4gb but that is like hiding behind a box. Its the principal you don't knowingly sell less of a product than you are promising. Here it is not less of a product it is using more power than designed. It's worse when you find out after months when they sold so many, what would have happened if people doing reviews actually noticed it and if it came out within weeks of launch. There would have been many more pissed people.

Like I said people accept losses once they have had the product for a while. It's like whatever. I would do the same because no wants to sit there and stress about it. whats done is done type of thing you know.
Click to expand...
All 4gb was usable i know had two of them just gave them away it was a great card never had one issue with it, what could happen is you could possibly experience on an exceptionally rare circumstance stuttering when you go over 3.5 in usage, but not always and certainly not most of the time.
 
razor1 said:
yeah its a simple fix, nothing that would break AMD, since it was caught early on, if AMD lets it go and things start failing its would be a major catastrophe for them later on, cause not only they will get pulled in any one that is part of the PCI-SIG process will do, and it will come down to individual people within each corporation or group as well.

Plus AMD might not have been aware of this problem, just a simple bios update near launch could have caused this problem, which they didn't have time to test and just assumed its ok.
Click to expand...

Yep I think every one is getting overworked on this. They will patch it quick relax the boost clocks a little under more taxing scenarios. It doesn't have to try and stay at 1266 all the time.
 
Presbytier said:
All 4gb was usable i know had two of them just gave them away it was a great card never had one issue with it, what could happen is you could possibly experience on an exceptionally rare circumstance get stuttering when you go over 3.5 in usage.
Click to expand...

Its all good. I don't wanna turn this thread in to that. This will be an easy patch. good thing reviews caught it quick.
 
trandoanhung1991 said:
So they marketed the card at 1266MHz typical boost clock, and now they silently reduce that to, say, 1200. You know what they'll get? Class action lawsuits. They intentionally lied about the ability of the card to buyers.
Click to expand...

yes typical but not guaranteed. There is a reason its called boost clock to boost when it can to stay adjust when it uses too much power. Thats all it means.
 
NKD said:
Yep I think every one is getting overworked on this. They will patch it quick relax the boost clocks a little under more taxing scenarios. It doesn't have to try and stay at 1266 all the time.
Click to expand...


I'm not sure if that will help completely but yeah it might we saw that the voltage increases even with small amounts of frequency increases.....
 
razor1 said:
I'm not sure if that will help completely but yeah it might we saw that the voltage increases even with small amounts of frequency increases.....
Click to expand...

Yea may be I think its their adaptive voltage thing being too aggressive, when it shouldn't be. I think thats what going on. Probably early pains and drivers not handling it correctly.
 
NKD said:
Yea may be I think its their adaptive voltage thing being too aggressive, when it shouldn't be. I think thats what going on. Probably early pains and drivers not handling it correctly.
Click to expand...
I was going to post this and figured I'd skim the thread first. I'm fairly certain this is just an errant issue with the card itself spiking voltage higher than the spec is for. It doesn't mean that boost clocks will be affected if they place a better limiter, it might actually improve them as thermals should lower as a result.
 
Trimlock said:
I was going to post this and figured I'd skim the thread first. I'm fairly certain this is just an errant issue with the card itself spiking voltage higher than the spec is for. It doesn't mean that boost clocks will be affected if they place a better limiter, it might actually improve them as thermals should lower as a result.
Click to expand...

Well, if this turns out to be the case, then someone will deservedly get a kick in the balls over at AMD...

Ship cards out to reviewers, bugged, 170w power consumption, barely as efficient as maxwell.

170/150 score for first impression
 
Trimlock said:
I was going to post this and figured I'd skim the thread first. I'm fairly certain this is just an errant issue with the card itself spiking voltage higher than the spec is for. It doesn't mean that boost clocks will be affected if they place a better limiter, it might actually improve them as thermals should lower as a result.
Click to expand...
I imagine someone in a lab is testing this at this moment. Although the card seems power limited with its boost as far as some reviews show.
 
Ieldra said:
Well, if this turns out to be the case, then someone will deservedly get a kick in the balls over at AMD...

Ship cards out to reviewers, bugged, 170w power consumption, barely as efficient as maxwell.

170/150 score for first impression
Click to expand...
well kyle did say it was a mess over there lol

Presbytier said:
I imagine someone in a lab is testing this at this moment. Although the card seems power limited with its boost as far as some reviews show.
Click to expand...
that's a combo of poor thermal dissipation and over pulling from the 6pin and pci e slot
 
Presbytier said:
They really should have just stuck an 8-pin on it. I keep feeling so many of the issues we have here would not have mattered then. Sure it would not be the most power efficient, but at least it work fine.
Click to expand...

My assumption on this is AMD didn't want to offer a card that appeared extremely inefficient compared to the competition. I mean, I get where that view would come from and I doubt any competent engineer would do the same. I'm guessing this is a call made by management and caused an engineer to run/leak to Kyle how piss poor it is over there right now.

Look at it this way: Nvidia releases the 1070 with a 8 pin on TSMC's 16nm FinFet, you are about to release the RX 480 that isn't remotely close to the performance on a 14nm FinFet PLUS it would pull in more power than the 1070 if you gave it the power it needed. Would you (as a marketing person) let that happen? The cards legitimate saving grace right now is its price/perf and we don't know if this is a good thing for AMD or not.
 
Trimlock said:
My assumption on this is AMD didn't want to offer a card that appeared extremely inefficient compared to the competition. I mean, I get where that view would come from and I doubt any competent engineer would do the same. I'm guessing this is a call made by management and caused an engineer to run/leak to Kyle how piss poor it is over there right now.

Look at it this way: Nvidia releases the 1070 with a 8 pin on TSMC's 16nm FinFet, you are about to release the RX 480 that isn't remotely close to the performance on a 14nm FinFet PLUS it would pull in more power than the 1070 if you gave it the power it needed. Would you (as a marketing person) let that happen? The cards legitimate saving grace right now is its price/perf and we don't know if this is a good thing for AMD or not.
Click to expand...

If you gave it the power it needed? :p it already pulls 170w with just the six pin :p

What difference would the 8pin make? In terms of appearances I mean. It already looks bad compared to the 1070 from that pov, the 8pin would have allowed then to stay within spec at least
 
Ieldra said:
If you gave it the power it needed? :p it already pulls 170w with just the six pin :p

What difference would the 8pin make? In terms of appearances I mean. It already looks bad compared to the 1070 from that pov, the 8pin would have allowed then to stay within spec at least
Click to expand...
Its already wanting to pull past both the PCIe and 6pin, its clear this GPU wants more and that could of been due to them trying to ramp up clocks and being unable to.

I agree with the 8 pin, they should of stuck one on there.
 
Trimlock said:
Its already wanting to pull past both the PCIe and 6pin, its clear this GPU wants more and that could of been due to them trying to ramp up clocks and being unable to.

I agree with the 8 pin, they should of stuck one on there.
Click to expand...

This. Either the current reference board's use of power is a problem or it isn't. If you think that it is, the AIB cards SHOULD negate this. if you think this isn't a problem, then there's nothing to be alarmed about.

I personally feel it's a problem. So, my RX 480 will be an AIB variant rather than a reference variant. Problem solved and I'm happy. My 970's VRAM also never gave me any issues, so I guess I'm just easy to please.
 
Is this a tit for tat by AMD fan boys. Hate to break the reality but unfortunately spiking vs. continuously running out of spec are two different things.

Any way you slice it. AMD is sol.
 
madgun said:
Is this a tit for tat by AMD fan boys. Hate to break the reality but unfortunately spiking vs. continuously running out of spec are two different things.

Any way you slice it. AMD is sol.
Click to expand...

anyway you slice it, both companies are out of PCI-e spec. Doesnt matter if its a little or a lot. Both Companies do it.
 
Did it happen with the current NVidia series. All I see is AMD got caught. No need to take your pants off for them. They won't pay you!
 
Brackle said:
anyway you slice it, both companies are out of PCI-e spec. Doesnt matter if its a little or a lot. Both Companies do it.
Click to expand...

Nope. I know that AMD fans want this to be true. And I see that you're looking for a reason for it to be true. PCI-E spec is about sustained power draw. Fluctuations happen. It's 75W +/-8% for sustained, and the reference 8GB RX 480 goes above that spec at stock speeds. I would never put one of those cards in my motherboard. AIB is different. I will in all likelihood own an MSI Gaming RX 480. But just because I'm an AMD fan doesn't mean I'll condone them taking a shortcut like this.
 
Brackle said:
anyway you slice it, both companies are out of PCI-e spec. Doesnt matter if its a little or a lot. Both Companies do it.
Click to expand...
Comparing a gpu from years ago that spiked is not the same as a new card on a new fab that is supposed to be super energy efficient running out of spec because the company was too asinine to put a 8-pin connector on it.
 
Presbytier said:
Comparing a gpu from years ago that spiked is not the same as a new card on a new fab that is supposed to be super energy efficient running out of spec because the company was too asinine to put a 8-pin connector on it.
Click to expand...

To be 100% fair, the 750ti wouldn't have the spiking issue if it had ANY power connector on it. So we can level a similar accusation at Nvidia, even though it's not quite apples to apples here. Also, I'm willing to bet
1) The GTX 950 75W card exhibits the same issue as the 750ti above, and,
2) The upcoming RX 460 has a similar issue either to the 750ti above, or the RX 480. We'll see.
 
Presbytier said:
I will not pretend to be impartial in the least. No one is. We all carry our biases with us, some of us are just more cognizant about it than others.
Click to expand...

Correct. But we tend to have a clearer bias on things we care more about. When it comes to gaming, I do have biases regarding certain gaming companies (Blizzard, Sega, etc.). The hardware? It's a replaceable tool tome, like a hammer. Give me the one that gives me the features that I want at the price that I want, regardless of brand. That said, I have a HUGE EVGA bias.

But I'm also into monitors (AMD wins due to FreeSync) and power delivery. AMD wins on the first, but really set me off on the latter. As I've said though, AIB cards will fix that :) Having this view doesn't make me a shill.
 
Whether they fix it or not it makes no difference to me since I will undo the fix in order to OC it. I've been an electronics hobbyist for decades, I design my own circuits. The rule of thumb in electronic design is to always over provision everything by 200%. Especially if it doesn't cost you anything, like passive components additional traces and pins. A 14% isn't going to do anything.

Of course if I was a motherboard manufacturer and someone asked me if this is an issue I would say yes it is, but that's called covering your own ass.
 
muxr said:
Whether they fix it or not it makes no difference to me since I will undo the fix in order to OC it. I've been an electronics hobbyist for decades, I design my own circuits. The rule of thumb in electronic design is to always over provision everything by 200%. Especially if it doesn't cost you anything, like passive components additional traces and pins. A 14% isn't going to do anything.

Of course if I was a motherboard manufacturer and someone asked me if this is an issue I would say yes it is, but that's called covering your own ass.
Click to expand...

Someone, somewhere, is going to have issues because of the PCI-E draw. Very likely not me or you, but someone will .
 
Brackle said:
Remember; nvidia's power limit counts toward the GPU voltage and memory overclock, with AMD there is no limit to the power you want to throw at the card.
Click to expand...

All hail prophet Brackie
 
You must log in or register to reply here.
Back
Top