AMD: Speculation, Rumors, and Discussion (Archive)
- Thread starter iMacmatician
- Start date
- Status
I undervolted my ancient AMD K6-III several tenths of a volt back around the turn of the millennium. Initially I did it simply because I set the mobo jumpers wrong without realizing it I believe (this was before the rise of software programmed voltage regulators on mobos, everything was hardwired with jumper blocks back then). Then when I realized the CPU ran just as well anyway I just kept it that way.
Had a 50W peltier slapped on there, and a Japanese Alpha heatsink (the one with those hexagonal pins set into hybrid alu/copper base) with twin stacked YS-Tech 60mm fans to cool off the heat.
It let the K6-III run at 83MHz bus speed at default multiplier for over 400MHz CPU speed IIRC, and core temp of below 30C. Oh, those were the days! (No, I DON'T want to go back there! )FWIW, some social media types of AMD says that this fix also targets the PCIE issue:
http://m.imgur.com/CFJYSWK?r
I agree wholeheartedly; AMD needs our support
There's a big difference between going out of spec on the PCIe slot and doing the same on the aux PCIe molex connectors. For starters, you can't blow your mobo by drawing too much from the molex connectors. Second, molex connectors have a huge amount of headroom built into them - each specced for close to 300W, and they can undoubtedly withstand even more than that before Bad Things happen, simply because they're massive compared to the piddly power pins in the PCIe slot.
Your argument that we shouldn't care now about this spec violation because nobody allegedly cared in the past about other alleged spec violations is as irrelevant as it is dumb. Sorry, but it's true. By the way, what evidence do you have that every hawaii card violates PCIe specs? That's quite the accusation, you know. I don't think you've got shit, by the way.
You have to make up your mind now, is violating specs "BAD, very bad" or not, because in the next paragraph you say it's really not that bad if you violate some specs, I guess you're the expert which ones. And at the same time call me dumb for not caring about specs.
So you can't blow up your PSU if you pull too much from 12V rail? It's perfectly fine to go 100% over specs on 295? Specs are not there for a reason now?
Hawaii can pull up to 300W. And I guess most of that power comes over 6 and 8 pins NOT the slot, right? How much is 150+75?
Would love to see figures for GTX 480 too.
ImSpartacus
Regular
Yeah, I was hoping that the 480 would be a convenient upgrade for those with older machines whose PSUs literally only have one 6-pin PEG (sorta how the <75W 460 will hopefully be cool for those without any PEG connectors, like the <75W 750 Ti before it). But if it's breaking the PCI-e spec, then that's not going to be a good drop-in upgrade for older machines. Pretty disappointing.
I always thought the spec on the 6-pin & 8-pin PEG connectors were a little "fuzzy".
Like when the 295X2 drew 500W on only a 2x 8-pin setup (theoretically 375W spec). I guess that was "ok".
Obviously it's not a "good" thing, but it might be a lesser of two evils.
EDIT I just noticed that you mentioned the 295X2, lol...
There's nothing fuzzy about the specs, its 75W for the slot, 75W for 6pin, and 150W for 8pin. AMD had to certify individual PSUs for the 295X2 since it broke the spec for 8pin connectors.
In reality any half-decent PSU can supply well over 75/150W on the 6/8pin connectors respectively, whereas the PCIe slot (and other electrical components) on any mobo is not nearly as robust. So if you're gonna break the spec it is better to break it through the connectors than through the motherboard.
Havent read the whole article but 30W less with undervolt(0.98) 80C 44% fan.
http://wccftech.com/article/radeon-rx-480-reducing-voltage-increasing-efficiency/
http://wccftech.com/article/radeon-rx-480-reducing-voltage-increasing-efficiency/
Regarding custom 480 AIB models.
The Sapphire Nitro OC is looking really nice, especially the backplate, and considering the price of the 8GB reference it is pretty good, about £20 more.
More details in the OCUk forum, seems they will be the very 1st to have this product.
https://forums.overclockers.co.uk/showthread.php?t=18738486
I can see this selling incredibly well, putting powergate to the side and even without clocks yet confirmed.
They note an official TDP of 175W.
Cheers
The Sapphire Nitro OC is looking really nice, especially the backplate, and considering the price of the 8GB reference it is pretty good, about £20 more.
More details in the OCUk forum, seems they will be the very 1st to have this product.
https://forums.overclockers.co.uk/showthread.php?t=18738486
I can see this selling incredibly well, putting powergate to the side and even without clocks yet confirmed.
They note an official TDP of 175W.
Cheers
Perhaps if you stopped trying to push your apologist agenda for a moment you'd realize there isn't a contradiction here, just you being a bit of a dick really. Breaking specs is bad, because ONE, while it may not necessarily cause trouble for people with quality and/or enthusiast-class PCs, it could cause problems for others - we've already heard of several people whose PCs either shut down while gaming/benchmarking or even blew out their PCIe slot, and this card only launched the other friggin day. This is the tip of the iceberg we've heard of this far unless AMD gets on the ball here real quick and cleans up its mess.
SECOND, breaking specs is bad, as mentioned by another poster here in this very thread, because OEMs and system builders won't touch a card that violates specs with a ten-foot pole, and it causes bad word of mouth out amongst Joe Public.
Third, I already went over why breaking specs on one connector is worse than doing it on the other, I feel no need to repeat myself.
Your PSU has to be as badly designed as the RX 480 if it blows from a card drawing 200W from a 6-pin PCIe molex plug...
Besides, any properly built PSU features overcurrent protection.Where did I say that? Stop with the strawman bullshit already, alright? Go test some hawaii-based boards to check their PCIe compliancy, then go make a thread about your findings. It'll be interesting hearing your conclusions, in that thread.
Regarding custom 480 AIB models.
The Sapphire Nitro OC is looking really nice, especially the backplate, and considering the price of the 8GB reference it is pretty good, about £20 more.
More details in the OCUk forum, seems they will be the very 1st to have this product.
https://forums.overclockers.co.uk/showthread.php?t=18738486
I can see this selling incredibly well, putting powergate to the side and even without clocks yet confirmed.
They note an official TDP of 175W.
Cheers
An 8-pin power connector and a good cooler: it's what the reference design should have been.
It could very well be that this power balancing behavior was never an issue within a particular (original?) design envelope of <= 150W. Only once process was shown to be underperforming, pushing the power consumption at chosen performance level beyond original projection did this become a serious issue.
Well, i dont use cheap Mobo ( more in the 500+ $ range ), but i have some real doubt that you can kill a motherboard this way. What happend when i OC my CPu's, 2 GPU's, by the past we was even overlock the PCI express bus pushing voltage and watts on it. I have never seen one motherboard dying due to that.
Motherboard are so much fulll of protection now who will tick at the moment a abnormal power condition is detected.
Just for tell you a little story, sometimes ago, my pump fall down in connection ( dont know what have happend for be honest as the pump is working now), the water start to become really hot, the tubes attached on the radiator who is on the top of my case, fall down. with nearly 2liters of hot water who fall on my motherboard and GPU's:
Power have been cut off immediately by the motherboard.. I dont even have a DDR sticks who is damaged. Everything is working good. sensors have detect a peak on the rails on the motherboard, and cut everything immediately. the thing is i was more suprised to see the gpu's alive ( no backplates on them, they are naked )
Well that is true with ohms law but its highly dependent on the electronic components, the reason why amperage went up like that is because voltage was being reduced, you still need a certain amount of power for the component, let be more amps or volts, something has to be there. So its a little tricky for AMD to balance things out without changing frequency I think. But we will see
el etro
Newcomer
Had to sign up just to like it.
Well my concern are for those who have the $45 - $60 and maybe a little higher motherboards/budget PSU and looking to use the 480 on them.
The single 480 left in standard form will be OK for most IMO , but the x16 slot spec (and to a very much lesser degree the auxiliary connectors) is important for those who are at the lower end of the mainstream spectrum, and I was surprised to read quite a few with the lower end of the motherboard market asking if the 480 would work fine with them on some other forums/Youtube.
This is an ususual situation that this card is priced and appealing to consumers that would not had used a card with these power consumption traits before, agree it is a minority but still.
The other unusual situation may be 2x480 now being used also at a lower budget than before, and the power distribution will possibly be a headache for the entry mGPU board/PSU, again more applicable to the minority I think.
That is ignoring the OCing.
But fingers crossed AMD delivers on that fix soonish and without noteably impacting performance.
Cheers
Last edited:
- Status
