AMD confirms Power Issue for RX 480, blames memory bus, will patch

[Tyns]

Same companies that made the PCI to PCIe 1.0 transition. Then the PCE 2.0 leap. And those crazy fools even made a PCIe 3.0 standard! What did consumers do? Upgraded their PCs and made the hardware companies lots of revenue.

If you progressed to PCI-E directly from PCI, you waited too long to upgrade!

Could bridge chips make a comeback?? Introduction and Pictures - GeForce 6600GT AGP Preview

 

[Shintai]

Since PCIe 3.0 cards work in 4.0. I think we just see PCIe 3.0 cards for a long time. Its not like the PCIe bandwidth is a big crippling problem the next 4-5 years on that front. Even with PCIe 2.0 x16 you barely lose anything.

So I think they just delay it from the card perspective. Else you need to make 2 cards and the "cheaper one" with a bridge style chip. That's not something that would look good cost structure wise. Then slowly introduce it as flagship feature only and going down as the time passes.

 

[Shintai]

Same companies that made the PCI to PCIe 1.0 transition. Then the PCE 2.0 leap. And those crazy fools even made a PCIe 3.0 standard! What did consumers do? Upgraded their PCs and made the hardware companies lots of revenue.

I assume its a joke, since you are using a PCIe 3.0 card in a PCIe 2.0 system

 

[walwalka]

You also posted a video of an idiot that didn't disconnect the power before changing the fuse, great!

It is not at all necessary to do so btw.. That live lead is completely useless without it's ground circuit. Moving it to the side to change a fuse is standard practice for the 12v industry, just don't touch it to anything metal that is also touching ground All the spark you saw was the caps recharging and it was going to happen whether or not he disconnected it at the battery. Amps with a large cap bank will do this.

 

[David-Duc]

You guys are so full of bullshit. A PCB that can't handle 300w. This guy pops a damn 300 amp fuse while making 4721 watts with components mounted to a PCB. So what you'll are really saying is that you're clueless about electronic design.

(Which we knew already.)

You seem to have short memory (or have no idea of how power draw could kill a motherboard really fast). Ask the people with burnt ATX 24-pin connector because the connector couldn't handle the power draw of multiple VGA. That's why motherboards for multi VGA now always have at least an extra power plug (6/8 pin PEG, 4 pin molex) for the graphics card. The ATX 24 pin only has 2 x 12V wire. The EPS 12V is already dedicated to the processor. So if they change the PCI-E 4.0 power specs without changing/adding 12v power socket on the mobo...

24 pin power connector burned | Overclockers Forums
24pin ATX melteted with after installing SLI

Pretty sure there is more, but 2 is probably enough for now.

 

[cageymaru]

I assume its a joke, since you are using a PCIe 3.0 card in a PCIe 2.0 system

Adios!

 

[cageymaru]

You seem to have short memory (or have no idea of how power draw could kill a motherboard really fast). Ask the people with burnt ATX 24-pin connector because the connector couldn't handle the power draw of multiple VGA. That's why motherboards for multi VGA now always have at least an extra power plug (6/8 pin PEG, 4 pin molex) for the graphics card. The ATX 24 pin only has 2 x 12V wire. The EPS 12V is already dedicated to the processor. So if they change the PCI-E 4.0 power specs without changing/adding 12v power socket on the mobo...

24 pin power connector burned | Overclockers Forums
24pin ATX melteted with after installing SLI

Pretty sure there is more, but 2 is probably enough for now.

So you're saying that when PCIE 4.0 happens they will only upgrade the slot and nothing else on the motherboard? I don't think they are going to install new slots on older motherboards. I would think they would design for it in the planning stages. Like they have been doing since the first PC came out.

 

[razor1]

It is not at all necessary to do so btw.. That live lead is completely useless without it's ground circuit. Moving it to the side to change a fuse is standard practice for the 12v industry, just don't touch it to anything metal that is also touching ground All the spark you saw was the caps recharging and it was going to happen whether or not he disconnected it at the battery. Amps with a large cap bank will do this.

Well when ya 300 watt fuses to change, I'm sure the amperage and voltage are fairly high, I would disconnect the power just for personal safety reasons.

 

[cageymaru]

It is not at all necessary to do so btw.. That live lead is completely useless without it's ground circuit. Moving it to the side to change a fuse is standard practice for the 12v industry, just don't touch it to anything metal that is also touching ground All the spark you saw was the caps recharging and it was going to happen whether or not he disconnected it at the battery. Amps with a large cap bank will do this.

I was going to say this also. Heck even if you add a circuit breaker to the system it will spark when you flip the switch. It's just less visible because they are enclosed in plastic.

 

[CSI_PC]

So you're saying that when PCIE 4.0 happens they will only upgrade the slot and nothing else on the motherboard? I don't think they are going to install new slots on older motherboards. I would think they would design for it in the planning stages. Like they have been doing since the first PC came out.

Upgrading other aspects falls under ATX guidelines and specifications, which will need technically revising and agreement to changes (this is outside of SIG) if it is to handle the additional power as some think (until we see a PCIE 4.0 motherboard many will have different views on this as we can only speculate).
The current spec supporting PCI with the 24-pin ATX connector only handles 2x12v pins with 6A at standard spec (144W), and 9A if HCS (216W) and 11A if HCS+, this figures are just the 12v.
Most IHV implement the standard or on the upper models the HCS.
The other challenge is that multiple cables will still be needed because that would be a massive amount of current going to a single connection to the motherboard and critically a single connection and rail in the PSU that also would need a re-design of all PSUs (another change that needs to be revised and agreed to the ATX specifications/guidelines).
Remember that currently the PCIe slots share the ATX 24-pin power connector so it is 66W multipled by number of slots.
If looking at the AMD server solution in the Tom's Hardware PCIe 4 article puts the PCIe 'auxiliary' connectors on the motherboard and quite a few of them, interestingly they then go back to a traditional 20-pin connection instead of the 24-pin.

From a more practical perspective, it is worth considering just how many auxiliary connectors power even a single upper mainstream GPU model, with each auxiliary Molex connector having either two or three 12V wires.
Cheers

 

[cageymaru]

Upgrading other aspects falls under ATX guidelines and specifications, which will need technically revising and agreement to changes (this is outside of SIG) if it is to handle the additional power as some think (until we see a PCIE 4.0 motherboard many will have different views on this as we can only speculate).
The current spec supporting PCI with the 24-pin ATX connector only handles 2x12v pins with 6A at standard spec (144W), and 9A if HCS (216W) and 11A if HCS+, this figures are just the 12v.
Most IHV implement the standard or on the upper models the HCS.
The other challenge is that multiple cables will still be needed because that would be a massive amount of current going to a single connection to the motherboard and critically a single connection and rail in the PSU that also would need a re-design of all PSUs (another change that needs to be revised and agreed to the ATX specifications/guidelines).
Remember that currently the PCIe slots share the ATX 24-pin power connector so it is 66W multipled by number of slots.
If looking at the AMD server solution in the Tom's Hardware PCIe 4 article puts the PCIe 'auxiliary' connectors on the motherboard and quite a few of them, interestingly they then go back to a traditional 20-pin connection instead of the 24-pin.

From a more practical perspective, it is worth considering just how many auxiliary connectors power even a single upper mainstream GPU model, with each auxiliary Molex connector having either two or three 12V wires.
Cheers

Again. Why is this relevant when we all know that they are going to design the motherboards correctly. Do you'll think that the hardware manufacturers fired all of the engineers so that the new PCIE 4.0 boards will catch fire?

If you have to upgrade your PC for new gear who cares? When DDR4 came out people upgraded. What next? Reminisce about the RS232 ports? Serial ports? Dot Matrix printers are still made it seems. I used to own full size thermal printers. I don't complain that I want to kill off the laser printers to bring these back into the mainstream. I think it's asinine that we have to auxiliary power connectors for components in a PC. Why can't it just run off the motherboard in the first place?

Technology moves forward. Deal with it or let Dell build you a system if it's too hard.

With that said I'm leaving this crap thread. It was started as a much to do about nothing in the first place issue and has devolved into a pure whine fest.

 

[NKD]

are we all still fighting about this power issue thing. Common guys ROFL. Let it go! Let there be peace!

 

[Daniel_Chang]

With that said I'm leaving this crap thread. It was started as a much to do about nothing in the first place issue and has devolved into a pure whine fest.

But only one person is whining...and thankfully he claims to be leaving the thread.

Anyway folks, false alarm. Some clarification provided by PCI-SIG.

PCI Express 4.0 Will Deliver 75 Watt over PCIe not 300W

Basically, the existing specs allows for up to 300W. You get 75W from the motherboard, 75W per 6-pin auxiliary connector, and 150W per 8-pin auxiliary connector. These configurations are certified up to 300W. Anything over 300W (2x 8-pin auxiliary connectors) is not officially certified by PCI-SIG. With PCI-E 4.0 this will change. The motherboard is still limited to 75W per PCI-E slot, but PCI-SIG will not ceritfy configurations exceeding 300W.

 

[CSI_PC]

Again. Why is this relevant when we all know that they are going to design the motherboards correctly. Do you'll think that the hardware manufacturers fired all of the engineers so that the new PCIE 4.0 boards will catch fire?

If you have to upgrade your PC for new gear who cares? When DDR4 came out people upgraded. What next? Reminisce about the RS232 ports? Serial ports? Dot Matrix printers are still made it seems. I used to own full size thermal printers. I don't complain that I want to kill off the laser printers to bring these back into the mainstream. I think it's asinine that we have to auxiliary power connectors for components in a PC. Why can't it just run off the motherboard in the first place?

Technology moves forward. Deal with it or let Dell build you a system if it's too hard.

With that said I'm leaving this crap thread. It was started as a much to do about nothing in the first place issue and has devolved into a pure whine fest.

Who cares you say...
Well those who write the specifications do and especially when it comes to ATX that are responsible for the freaking specification and guidelines for power in PC.....
Even the initial response in Tom's Hardware, SIG say implementors should be following-reference the ATX-150W power specification/guidelines, which I think also tie into the ATX12V 2.x standard.
As I said this is beyond SIG who are responsible for PCIe buy NOT ATX.
You have any ideas how long it takes to revise and amend these global standards and then followed by the hardware changes the manufacturers then do?

I have worked on some standards councils/committees myself, and it is sloooooow.
Good luck getting a non-ATX spec/guidance consumer motherboard/PSU/PC.

Anyway as Daniel Chang mentions with the link:

SIG reached out to tell us that the power increase for PCI Express 4.0 will come from secondary connectors and not from the slot directly.

Not surprised myself, notice it mentions connectors meaning requiring more than one probably like the current spec-defined auxiliary situation.
Cheers