• Please review our updated Terms and Rules here

Nvidia gtx 1070/1080

this is my motherboard: http://www.supermicro.com/products/motherboard/QPI/5500/X8DAH_-F.cfm

Supermicro is NOT a company that is going to save a few pennies by making thinner traces. Quite the opposite. Everything SM makes is built like a brick shithouse. I have the utmost confidence in their engineering team. As for the power supply, I was trying to indicate I had one with more than sufficient wattage and quality.
 
luckybob said:
They fixed the power thing with a driver update. => https://www.youtube.com/watch?v=YHEJ5cB9gmA
Click to expand...

They didn't quite fix it. They redistributed the power draw, so now it draws slightly less from the PCI-e slot (*just* inside spec), and draws the extra power from the 6-pin connector (going even further out-of-spec than it already did).
Most PSUs are more resilient in coping with excess power draw than PCI-e slots are though, so AMD can get away with it.
But the power problem is not fixed. It still draws more than 150W from a card designed to only handle 150W max (75W from PCI-e slot and 75W from a 6-pin connector).
 
Agent Orange said:
I never gave Vulkan much thought. It's been out there a while and the game availability seems fairly sparse - maybe DOOM is the only one? My HD 7970's would qualify for Vulkan, but I don't think I'm going to buy another game at this time just for a look-see. Anyway, I went ahead and put my name in the hat for a ZOTAC 1080 from Newegg. I know I going to take it in the shorts price-wise, but like any kid, just gotta have it.
Click to expand...

On my GTX970 Vulkan seemed to perform slightly less in DOOM than OpenGL did.
It seems that it is not all that mature yet. That could be partly NVidia's drivers, but I also suspect that the engine itself isn't as good at things like memory management as NVidia's OpenGL drivers are.
For example, the game has a minimum requirement of 2 GB VRAM in OpenGL (I've actually tried to run it on a 1 GB GTX460 for teh lulz, and at the lowest detail levels it was actually playable).
With Vulkan, the 2 GB GeForce cards are not supported. That sounds to me like they can't manage memory efficiently enough to stay within the 2 GB.
I think that is also what I'm seeing with my 970: It seems to slow down in the more open spaces (where more objects/textures are active, hence requiring more resources from VRAM to be active), where it was silky smooth with OpenGL.

I will just wait and see what future updates to the game and drivers will bring.
 
Last edited:
The rated maximum of standard molex connectors is 10A per circuit. so the 6 pin power can deliver 30A. or 360W Granted this is for the physical connector. other factors will change the maximum. Such as the gauge of conductors from the PSU. There is still only 3 lines, even for the 8-pin connectors, and those are rated 150W. There is still only 3 12v wires.

So yea, they can really go nuts on the external power connector. imho, they should have used used a 8 pin connector and drew 100% from it and nothing from the mobo. But what do I know?
 
luckybob said:
The rated maximum of standard molex connectors is 10A per circuit. so the 6 pin power can deliver 30A. or 360W Granted this is for the physical connector. other factors will change the maximum. Such as the gauge of conductors from the PSU. There is still only 3 lines, even for the 8-pin connectors, and those are rated 150W. There is still only 3 12v wires.

So yea, they can really go nuts on the external power connector. imho, they should have used used a 8 pin connector and drew 100% from it and nothing from the mobo. But what do I know?
Click to expand...

Erm, no.
The standard specifically states that 6-pin power connectors can only deliver 75W max. 8-pin can deliver 150W max (the function of the extra pins is only to detect that a 150W-capable connector is present. Many cards will work with a 6-pin connector, but will limit power draw to 75W in that case).
These are the baseline design specs for PSU and board makers.
See also: http://www.playtool.com/pages/psuconnectors/connectors.html
 
Last edited:
luckybob said:
no. that is what the specification calls for. The physical connector is capable of up to 23A per circuit. http://www.molex.com/molex/products...Title=Introduction&parentKey=power_connectors
Click to expand...

Yes, but that is *only* the connector.
That means that, given the right wiring and a powerful enough PSU attached to it, you can run up to 23A through the connector, without it melting or otherwise malfunctioning.
This is ofcourse completely irrelevant in the greater scheme of things.

What REALLY matters, is what the specification says you can reasonably expect the PSU to deliver to that connector, within the context of an ATX and PCI-e compliant PC system. Which is specified as being 75W for a 6-pin and 150W for an 8-pin connection.
The Radeon RX480 draws significantly more than 75W from a 6-pin connector.
 
Last edited:
luckybob said:
this is my motherboard: http://www.supermicro.com/products/motherboard/QPI/5500/X8DAH_-F.cfm

Supermicro is NOT a company that is going to save a few pennies by making thinner traces. Quite the opposite. Everything SM makes is built like a brick shithouse. I have the utmost confidence in their engineering team. As for the power supply, I was trying to indicate I had one with more than sufficient wattage and quality.
Click to expand...

SuperMirco may have been good back in the late 90s, early 2000s, but that isn't the case anymore. Their quality went down the toilet when they stopped manufacturing boards in the US and moved to Taiwan. Case and point is their inexplicable use of a Matrox G200eW, which is a video chip from 1997 on a wide range of their motherboards. No operating system has supported this thing in nearly 15 years, resulting in terrible video performance.

And even if they had the entire motherboard as a power trace, this still doesn't mean the slot won't burn out. SuperMicro and pretty much every other board manufacturer uses run of the mill slots made by Foxconn or a number of other Chinese manufacturers.

luckybob said:
no. that is what the specification calls for. The physical connector is capable of up to 23A per circuit. http://www.molex.com/molex/products...Title=Introduction&parentKey=power_connectors
Click to expand...

Doesn't matter what the connector is capable of, the card is still out of spec. AMD doesn't get a free pass to violate standards just because the connector can handle the extra current draw.
 
Concerning the 480: A lot has been said here about what the 480 can or cannot do. I've be buying AMD/ATI/Radeon video cards for a long time and have never experienced any type of failure. It seems implausible that a company of AMD's stature would risk their reputation and attempt to perpetrate a hoax on the consumer by permitting a new product to enter the marketplace without being fully tested for performance and safety. I understand that there is normally software issues associated with any new product release, but it it completely absurd to believe that your PC is in danger of melting down because of an assumed engineering glitch. There may be a problem or two with the release, but I'm willing to bet that AMD will standby their product and replace whatever goes up in smoke do to some unfortunate and arcane event such as that.
 
Agent Orange said:
It seems implausible that a company of AMD's stature would risk their reputation and attempt to perpetrate a hoax on the consumer by permitting a new product to enter the marketplace without being fully tested for performance and safety.
Click to expand...

It's not so much about performance and safety.
It's about not meeting PCI-e specifications.

You're right, I can't imagine that this would slip by an outfit the size of AMD.
Therefore, what I think happened is this:
1) They designed the PCBs for sub-150W performance, thinking this would be the ballpark, and performance-per-watt was promoted as a big selling-point for the Polaris architecture (having only a 6-pin connector gives off the impression of a card that is very power-efficient... note that nVidia's 145W GTX970 still uses 8-pin, just to be safe).
2) When the first 14nm FinFET GPUs arrived, the process was rather immature, and power consumption was considerably higher than originally expected (a similar situation to the GeForce GTX480 for example)
3) They had 2 choices: Either they'd clock the GPUs lower than intented, and they wouldn't hit their intented performance-envelope. Or, they could just jack up the GPUs so they'd have competitive performance, and pray nobody would notice/care.

It's obvious which option they took (I guess a lot was riding on them at least reaching 970/980-like performance levels, because anything lower would be viewed as a failure).
And various sites noticed.

No, your PC is not likely to melt down (although there is a reason why AMD made this driver fix... the PCI-e slot is a lot more fragile than the 6-pin connector, and the slot could indeed get damaged, especially on older boards... some people even report that this has happened), but that doesn't mean that the parts are in-spec (not meeting specifications means that they would not have the right to use the PCI-e trademark, which could mean legal consequences, as PCI-SIG would have to actively defend their trademark against any infringement). There's a big gray area between 'in spec' and 'melting down', it's called a safety margin. RX480 operates somewhere in that gray area.
 
Last edited:
Here some late info on the 480: http://themerkle.com/new-amd-video-card-fails-to-comply-with-pci-e-specifications/

There is indeed a power problem with the 480 card. I have a feeling a general recall is in the wind and AMD will probably offer refunds or upgrades, just speculation on my part but not unknown in the industry. I can imagine that there are some folks at AMD filling out their resumes about this time. My gut feeling is the card will perform okay, but might be kind of dicey in Crossfire.
 
luckybob said:
They fixed the power draw with a driver update.
Click to expand...

They didn't fix 'the power draw', the card still draws more than 150W, which is still out of spec for a PCI-e card with 6-pin connector.
They fixed the power draw on the PCI-e slot, at the cost of running the 6-pin connector even more out-of-spec.
In fact, the default setting doesn't even put the PCI-e slot draw entirely in spec. There is a 'compatibility' mode that does, at the cost of some performance (but still drawing more than 75W from the 6-pin connector).

Also, Tomshardware was not the only site that measured this. Other sites did their own measurements and verified Toms findings. They also re-tested after the driver update.
 
You people are making mountains out of mole hills. Its not GROSSLY over, just slightly and in small spikes.

GET OVER IT. I'm buying two of them. You haters can kiss my fat white hairy ass.
 
No, it wasn't just the spikes, it was the *average* power draw that was out of spec.
We're not making mountains out of mole hills, we're just trying to be factual and accurate. You seem to be downplaying/in denial.
Whatever.
 
Scali said:
No, it wasn't just the spikes, it was the *average* power draw that was out of spec.
We're not making mountains out of mole hills, we're just trying to be factual and accurate. You seem to be downplaying/in denial.
Whatever.
Click to expand...

AMD fanboys have been in denial for years.

AMD has been in a slow death spiral for nearly a decade, selling off their fabs and other profitable sectors just to prop up the failing x86 division. They also grossly mismanaged the GPU division soon after they bought it and ran that into the ground too. Their entire existence now hinges on their Zen platform, which if it flops, AMD will be finished.
 
AMD has been short on money for ages, and that affects R&D quite a bit.

I don't think AMD will ever go bankrupt because it is in Nvidia's and Intel's best interest to have some competition so that they won't have to deal with being a monopoly. If MS propped up Apple with a few hundred million I am sure Intel will find a way to kick a few bucks over to AMD somehow.

And to be honest Intel's greatest advantage is not chip design itself, but the chip making process it controls with mega billion dollar fabs. If you are the first to get a smaller process going you will have faster processors by default.
 
Thread back on track:

I was waiting for FedEx to drop off my GTX 1080 this afternoon and nothing happened. I ran the tracking number on my laptop and it said DELIVERED. The A/C was off today and the house was wide open. I can usually hear the truck, and there was no traffic except for locals. I went into the panic mode and stomped around the neighborhood, and still no sign of my package. I was walking down my driveway and getting ready to jump in my 2007 Ford Ranger beater (220,200+ miles), and run over to the local FedEx center, when my neighbor came trudging across the front lawn with a box in hand. I almost did a Snoopy happy dance right on the driveway. I could see the "This box may contain Awesome" on the side of the package and so I knew my latest investment in Newegg had arrived and was in good hands.

Cut to the chase. I installed the thing and it posts as you might expect. Took extra precautions in removing the old Radeon Crimson drivers, including a complete sweep of the registry. NVidia released the driver fix yesterday and I'll be ringing the thing out for most of the evening.
 
You must log in or register to reply here.
Back
Top