Penny stocks always have high volatility like that. People think they can jump in and quadruple their money in a week and get out.
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
-
Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.
AMD confirms Power Issue for RX 480, blames memory bus, will patch
- Thread starter pcgeekesq
- Start date
Penny stocks always have high volatility like that. People think they can jump in and quadruple their money in a week and get out.
Their (pro AMD RX 480 'investors') only fallback now is market share and wicked OC of AIB partner cards (but then so much for $/perf), but the GTX 1060 info tied with a probable performance reduction from the AMD driver "fix" doesn't bode well for that, either.
They had some wins but the upcoming graphics releases and Zen are the primary motivators. Su claiming they'll be back to profitability either this recent quarter or next quarter, but who knows now!
Twitch guy is on the right track when he measured the card but he really doesn't know what he is saying.
After looking the measurements what he did, its almost 100% certain that the card has a 50/50 vcore power distribution shared between 6-pin and PCI-E slot. It also seems that the phases are not linked together. 6-pin (3 phases) and PCI-E slot (3 phases), which is a GOOD THING.
So why it's a good thing? AMD is using the most sophisticated power regulator chips what you can find on the market. And that chip is programmable. So it's actually possible to change the distribution ratio between 6-pin and PCI-E slot because the phases are not linked. The ratio can be changed up to 30.6% (according to The Stilt). Heck it can most likely be done through Afterburner if you know how to program the L2C bus.
One other wrong point from that video. The 6-pin is not out of spec because the third 12v is optional and it will not blow up the mobo (PCI-E slot) if you don't have the 6-pin plug connected because the regulator chip will not start if it doesn't sense that there's 12v incoming from 6-pin. Only an ape would design the card in a way that there's no protection when you try to power it on without 6-pin plug.
Ps. If you don't know who The Stilt is, you're not [H]ard enough
Pss. That 6-pin plug can actually push up to 396w before melting because it has all the 6 HCS pins there (3x12v, 3xGND). Heck even if were using only 2x 12v old STD pins (which haven't been allowed since 2005), it could push up to 192w before melting.
6-pin plug specifications are way too conservative but what can we do about it when the specification was finalized in 2004.
After looking the measurements what he did, its almost 100% certain that the card has a 50/50 vcore power distribution shared between 6-pin and PCI-E slot. It also seems that the phases are not linked together. 6-pin (3 phases) and PCI-E slot (3 phases), which is a GOOD THING.
So why it's a good thing? AMD is using the most sophisticated power regulator chips what you can find on the market. And that chip is programmable. So it's actually possible to change the distribution ratio between 6-pin and PCI-E slot because the phases are not linked. The ratio can be changed up to 30.6% (according to The Stilt). Heck it can most likely be done through Afterburner if you know how to program the L2C bus.
One other wrong point from that video. The 6-pin is not out of spec because the third 12v is optional and it will not blow up the mobo (PCI-E slot) if you don't have the 6-pin plug connected because the regulator chip will not start if it doesn't sense that there's 12v incoming from 6-pin. Only an ape would design the card in a way that there's no protection when you try to power it on without 6-pin plug.
Ps. If you don't know who The Stilt is, you're not [H]ard enough
Pss. That 6-pin plug can actually push up to 396w before melting because it has all the 6 HCS pins there (3x12v, 3xGND). Heck even if were using only 2x 12v old STD pins (which haven't been allowed since 2005), it could push up to 192w before melting.
6-pin plug specifications are way too conservative but what can we do about it when the specification was finalized in 2004.
Last edited:
Is that doable in a "driver" fix?
westrock2000
[H]F Junkie
- Joined
- Jun 3, 2005
- Messages
- 9,489
Back in the 20th century American citizens had to go to college to learn Ohm's law. I even graduated from one of the best high schools in my city and I did not take Physics, Calculus, Algebra 2, or even a foreign language.
That was in 1999, only a mere 6 weeks after the release of the The Matrix. So I'm assuming kids now-a-days get rigorous super duper techno-educations in high school.
Funny thing is that in my country (Finland), physics is a mandatory subject in junior high school (age groups 13-15) and some schools teach physics at elementary school (age groups 7-12). We call this elementary + j hs as comprehensive school and it is mandatory for children to complete it.
But let's drop the OT here
But let's drop the OT here
D
Deleted member 126051
Guest
Yeah which is why I provided the link for poster to go to, well glad at least you looked
They also mentioned the 175W TDP.
Just to say to everyone I do not think it is that easy to conclude much beyond looks with those photos.
Link again for those that missed it 1st time: SAPPHIRE RX 480 NITRO NOW AVAILABLE FOR PRE-ORDER AT OVERCLOCKERS UK (WORLD FIRST) !!! - Overclockers UK Forums
This is probably the only place with information on the board, reason explained in the forum post as they are the 1st to sell it, and allowed to post information about this card, although by now many sites probably mention this card using said info.
Cheers
Last edited:
Lunas
[H]F Junkie
- Joined
- Jul 22, 2001
- Messages
- 10,050
In the specs it says 8 pin...
Yeah I would prefer the info from reputable company sponsored OCers such as Kingpin/Tin.
There is a few for AMD, but annoyingly cannot remember their names as would be worth seeing if they have done any guides/articles yet on extreme OC the 480.
Cheers
Lunas
[H]F Junkie
- Joined
- Jul 22, 2001
- Messages
- 10,050
It would allow booting rather than refusing to turn on the card the card would attempt to draw all of the power needed from the slot too...
![[Spectre]](/data/avatars/m/89/89728.jpg?1455675927){kind=avatar}
[Spectre]
I Will Find You
- Joined
- Aug 29, 2004
- Messages
- 24,185
That s nowhere near universally true though. I went to public school in a portion of the country people like to make fun of for being "backward", "redneck", etc before then and Physics (including ohm's law), foreign language, and calculus were standard high school curriculum.
Each state mandates core classes then the county mandates there things then the city mandates there's so its really dependent on the school district ya go to I guess.
D
Deleted member 93354
Guest
Wow I just watched that video. They covered all the same points I did including the power graphs. Great geeks think alike. :-D
- Joined
- Feb 11, 2013
- Messages
- 6,464
i've been invoking you to do advanced electric test on RX 480 together with Kyle and Daniel.. =(.. common a free job for science..
D
Deleted member 93354
Guest
I was doing Ohms law in 6th grade. (Seriously)
Last edited by a moderator:
dragonstongue
2[H]4U
- Joined
- Nov 18, 2008
- Messages
- 3,162
many many articles have said that GDDR5 especially at higher speed and even with current manufacturing you are looking at ~30w to just power the chips and can be in the 80w range to power the "board" memory subsystem as a whole, so to say that memory is not taking up a considerable part of the power budget is not proper, and given these are more or less top of the pyramid for speeds, it could very well be they are pushed to hard with a memory controller not designed for the excessive speed so making the problem "worse" EEC, clock gating etc can do all kinds of fancy things. IMO it still is being blown out of propotion likely marketing shills trying to force sales somewhere else, is it a good problem, HELL NO, is it as widespread a major nasty issue, IMO no not at all, go a head test the many millions of Nvidia products ever released I can say for certain vast majority of them overload PCIe just the same if not worse via overloading one of the rails instead of drawing from both.
Least AMD is working on and looking into asap, and they did say that they passed both internal as well as external PCI-SIG testing, I tend to believe them over the user market, as well the power controller they use can be tuned via drivers, so at least they can do something about it, vs them fancy power controllers others use that are set and forget type deal, GTX 10 series are also mass loading their PCIe rails and overheat etc, do we hear anywhere near as much, of course not, cause they can afford to pay things off
Least AMD is working on and looking into asap, and they did say that they passed both internal as well as external PCI-SIG testing, I tend to believe them over the user market, as well the power controller they use can be tuned via drivers, so at least they can do something about it, vs them fancy power controllers others use that are set and forget type deal, GTX 10 series are also mass loading their PCIe rails and overheat etc, do we hear anywhere near as much, of course not, cause they can afford to pay things off
many many articles have said that GDDR5 especially at higher speed and even with current manufacturing you are looking at ~30w to just power the chips and can be in the 80w range to power the "board" memory subsystem as a whole, so to say that memory is not taking up a considerable part of the power budget is not proper, and given these are more or less top of the pyramid for speeds, it could very well be they are pushed to hard with a memory controller not designed for the excessive speed so making the problem "worse" EEC, clock gating etc can do all kinds of fancy things. IMO it still is being blown out of propotion likely marketing shills trying to force sales somewhere else, is it a good problem, HELL NO, is it as widespread a major nasty issue, IMO no not at all, go a head test the many millions of Nvidia products ever released I can say for certain vast majority of them overload PCIe just the same if not worse via overloading one of the rails instead of drawing from both.
Least AMD is working on and looking into asap, and they did say that they passed both internal as well as external PCI-SIG testing, I tend to believe them over the user market, as well the power controller they use can be tuned via drivers, so at least they can do something about it, vs them fancy power controllers others use that are set and forget type deal, GTX 10 series are also mass loading their PCIe rails and overheat etc, do we hear anywhere near as much, of course not, cause they can afford to pay things off
Err no nV cards never did this nor did any AMD card in past do this lol, not over the motherboard. I think you need to get your facts straight. If you are confusing peaks vs sustained or lumping those together, yeah you don't know what the differences of the two are.
The GPU is what is being powered over the PCI-e bus, and this is the problem. Its drawing too much power over the PCi-e bus, the pci-e connector is secondary to the immediate problem but that too is an issue for OEM's and system builders.
- Joined
- Aug 17, 2005
- Messages
- 7,361
Where you went to school? In 87 i didn't learn it till after high school (military)
Lunas
[H]F Junkie
- Joined
- Jul 22, 2001
- Messages
- 10,050
To be honest the only thing I would like to see powered off the pci-e slot is the fan on these cards no reason why 100% of the power can't be provided directly from the PSU...
Have you never downvolted a GPU or a CPU? You don't have to reduce the frequency too to get lower consumption i.e. less amps from the power source.
D
Deleted member 93354
Guest
It was public school in a small po-dunk town in Maryland. We had volt meters and touched various materials around the science classroom to see how well they conducted. We had a sled with a metal slug on it pass through a field of wires and hooked up a galvanometer to show how field was affected. We had potato clocks, lemon clocks, and solar cells. We even had a generator attached to a bike to see how many watts a human body put out. (How many light bulbs it could light and for how long.) But yes we also learned ohms law. By 12th, I was using calculus for tc circuits with capacitors. That was ap physics though.
I bet a lot of you guys did too. You just likely don't remember it.
I graduated in 90.
Last edited by a moderator:
D
Deleted member 93354
Guest
That's purely luck of the draw. You got lucky if you could downclock and not lose frequency. Intel's T series CPU's are supposed to be the binned CPU's that work well at a high frequency with extremely low voltages.
FinFET circuits do downclock when they are undervolted very well. (Meaning you can get significant power savings for a small dropoff in voltage provided you are willing to lose some speed.)
Undervolting and Overclocking on Ivy Bridge
Pyroja
Gawd
- Joined
- Jan 4, 2005
- Messages
- 594
Just curious: Has anyone actually done the testing required to back up that statement? We know what a 970 draws in its stick form, but it there anyone out there that has tested user OC'd cards to make sure they aren't overdrawing from the PCI-E slot? I've been reading reports that this is a lot more common than people think, and it only coming to light because it's happening to a newly released card (not making excuses for it).
D
Deleted member 93354
Guest
About 42->43 Watts pver the PEG....I think you have some headroom even if all the overclock goes to the PEG (which is highly unlikely)
Ieldra
I Promise to RTFM
- Joined
- Mar 28, 2016
- Messages
- 3,539
No but what happens is this; the testing was done at stock if anything, and the cards' power draw from the PCI-E does not scale with load or with OC; all the headroom is taken from the external connectors. You can see it in the 960 comparison recently done by PCPer; the power draw from PCI is fairly constant and unchanging. I'd be interested in actually seeing results for heavily overclocked 970s though.
Ocellaris
Fully [H]
- Joined
- Jan 1, 2008
- Messages
- 19,074
AMD fans have more belief than UFO chasers, they know they will be right eventually
LOL so far its been 1 out of 7? I would say its not betting odds lol.
Ieldra
I Promise to RTFM
- Joined
- Mar 28, 2016
- Messages
- 3,539
Let's be honest, their engineer in charge of board design is a bigger troll than i can ever aspire to be out of spec pin with three 12v leads and the sense grounded ; still uses pcie for providing core power. Trolololololol
out of spec pin with three 12v leads and the sense grounded ; still uses pcie for providing core power. TrololololololWhy so much hate? So hard to understand a simple concept? let me correct you on some of it.
First it does not underperform for the price.
second it does not run hot, look at h reviews. Temp in 80s max.
third it is not any more louder than any other blower design. Read H review.
Now Yes it does take more power but it is still way more efficient than AMD last gen. Nvidia had a leg up for a while on this and its not surprising they improved a bunch on it from 28nm to 14nm.
Now this problem is real. Lets see how amd addresses it. It wouldn't have been an issue if it was drawing from 6 pin rather than pci-ex.
Now you can calm down and stop hating and wait until tuesday see how AMD responds. Nvidia has their fuck ups with 970 its not new. Give it a rest and enjoy the fourth.
Krenum
Fully [H]
- Joined
- Apr 29, 2005
- Messages
- 19,194
I'm just wondering how this got through Quality Control. Or if they even cared?
Ocellaris
Fully [H]
- Joined
- Jan 1, 2008
- Messages
- 19,074
They knew. They just didn't care or took the risk that people wouldn't figure it out. I've worked on electrical engineering projects that barely shipped hundreds of units with only two engineers and they still ALWAYS validated power consumption so they didn't burn anything up. It is unfathomable AMD would launch these new effecient cards without verifying the per pin power consumption of every board revision that led up to the retail product.
If AMD didn't know, they should burn the company down and send everyone home. That level of incompetence is unheard of for such important releases.
Last edited:
Ieldra
I Promise to RTFM
- Joined
- Mar 28, 2016
- Messages
- 3,539
It seems likely they knew, but I'm still curious what their statement will be because it makes no sense why they would do this honestly. It's like they are trying to intentionally cripple their launch lol
There is only one logical explanation, performance....... They needed to at least get it to what they stated for VR lol. Anything less than a 970 performance its not going to be VR ready, and their marketing department being marketing, put their foot in their mouth by pushing it as VR for TAM.
Last edited:
Krenum
Fully [H]
- Joined
- Apr 29, 2005
- Messages
- 19,194
Have there been any reports of peoples computers catching on fire? Is that even possible with this kind of power draw through PCI slots? AMD could be facing serious lawsuits I would assume.
- Joined
- Feb 3, 2008
- Messages
- 6,984
This is sig worthyAMD fans have more belief than UFO chasers, they know they will be right eventually