monstercameron
Newcomer
Truly disappointed in the analyses so far. I though b3d was above the doom and gloom rhetoric...atleast outside of thst thread. Recently this forum reads like anandtech or other similar forums.
AMD: Speculation, Rumors, and Discussion (Archive)
- Thread starter iMacmatician
- Start date
- Status
Another unknown is what was being used as the representative workloads for the tests.
If the card passed the tests in that set of representative workloads considered by the certification process, then the board is certified.
There are countless devices out there that have pathological corner cases or power viruses and they have more primitive hardware than CPUs and GPUs. Even CPUs use the "representative workload" card for determining things like TDP, and they exclude power viruses or programs that inconveniently draw power like them.
If the card passed the tests in that set of representative workloads considered by the certification process, then the board is certified.
There are countless devices out there that have pathological corner cases or power viruses and they have more primitive hardware than CPUs and GPUs. Even CPUs use the "representative workload" card for determining things like TDP, and they exclude power viruses or programs that inconveniently draw power like them.
There's clearly an issue here and whilst the reaction should probably be a bit more reserved, most of us truly want information and honest expert answers on an apparent large problem.
Then it's a question of which ones, what settings, and so on. They cannot test all of them in all configurations, and I'm not sure they'd be free-playing levels manually.
If the board managed to scrape by set that was agreed upon by the specification process, then the specific PR-speak of "the board passed certification" is true, and is fundamentally true of pretty much any certification process. They can only certify based on the things they think to look at and certify. Maybe that set needs to be revisited.
D
Deleted member 13524
Guest
Regarding the suggestion of a late overclock when faced with Pascal's clocks, that doesn't seem like plausible at all.
The RX 480 has been told it would bring the minimum VR-spec to the masses, and that has been defined as R9 290 and GTX 970 for well over a year. The RX 480 delivers just that, and not more or less. So unless the GPU's single-threaded performance came up short of their expectations, these clocks are what they were aiming for, for a long time.
Perhaps no reviewer ever goes over that because no one assumes you must blame the chip maker (and not even the OEM who finalizes the BIOS? curious..) for the consequences of overclocking.
As far as anyone knows, all those home-overclocked 1500MHz Maxwells and 2GHz Pascals could be pulling 100W out of the PCIe's 12V pins and no one would be batting an eye to it.
Out of so many things that could be discussed around the new chip/card, this has turned into another hater cirlejerk about something most of the circlejerkers don't even seem to grasp a minimum comprehension about (i.e. "it's pulling more current which is even worse than we thought").
I propose the creation of a thread called "RX 480 power shenanigans" where all of it could be sent to, so that the discussion could be kept a bit more sane elsewhere.
The RX 480 has been told it would bring the minimum VR-spec to the masses, and that has been defined as R9 290 and GTX 970 for well over a year. The RX 480 delivers just that, and not more or less. So unless the GPU's single-threaded performance came up short of their expectations, these clocks are what they were aiming for, for a long time.
You should read that sentence again. 95W spikes are okay, 95W continuous isn't okay. 95W continuous in the PCIe's 12V pins is only attainable through overclock for the RX 480.
By all means, please show us the myriad of reviews for other cards explicitly showing the amperage levels on the PCIe's 12V when they are being overclocked outside factory-defined bounds.
Perhaps no reviewer ever goes over that because no one assumes you must blame the chip maker (and not even the OEM who finalizes the BIOS? curious..) for the consequences of overclocking.
As far as anyone knows, all those home-overclocked 1500MHz Maxwells and 2GHz Pascals could be pulling 100W out of the PCIe's 12V pins and no one would be batting an eye to it.
100% agreed.
Out of so many things that could be discussed around the new chip/card, this has turned into another hater cirlejerk about something most of the circlejerkers don't even seem to grasp a minimum comprehension about (i.e. "it's pulling more current which is even worse than we thought").
I propose the creation of a thread called "RX 480 power shenanigans" where all of it could be sent to, so that the discussion could be kept a bit more sane elsewhere.
Anarchist4000
Veteran
I still have no idea what this whole argument is about. 150W card pulling 200W is exceeding limits. Just back off the voltage and clocks a bit and suddenly everything is fine. Are the limits actually exceeded when WattMan is set to 150W limit?
That is all fine but at stock clocks and stock boost clocks it still goes above, that is the problem, its not just overclocking and its not just FurMark.
AMD has to address this by either lower voltages *frequency or throttle their card in a normal gaming situation with the reference card, or a 8 pin instead of 6 pin, the board has the capability of an 8 pin, then adjust where the card draws its power from....
its not such a problem that AMD can't fix and can't fix quickly, unlike the 3.5 gb 970 with the ROP counts.
But its problem that can affect the entire system as a whole, so there are more problems outside of just its own product in play.
And if you really want to into this deeper, the 8 pin and 6 pin connectors are the same, just that the 8 pin has extra grounds, so when they are plugged in it tells the board it draw more power from the connector if needed.
Last edited:
That PCPer article is great. The AMD faithful posting comments that the GTX960 had the exact same problem so they updated the article with GTX960 results proving otherwise. 
I may look into getting a 480 for my secondary rig if this issue gets ironed out and the 1060 is as overpriced as the 1070.
I may look into getting a 480 for my secondary rig if this issue gets ironed out and the 1060 is as overpriced as the 1070.
Anarchist4000
Veteran
Isn't that type of throttling, reducing performance to maintain a power limit, exactly what WattMan is doing? If it's set at 150W instead of 200W those stock values are no longer an issue.
That is exactly the problem at hand: Speaking from our review, up to 6.4 amps non-overclocked in a regular gaming load. No one in their right state of mind would open this can of worms over an OC scenario though.
D
Deleted member 2197
Guest
AMD possibly will be using XX5 revisions e.g. AMD Radeon RX 495
http://www.guru3d.com/news-story/amd-possibly-will-be-using-xx5-revisions-e-g-amd-radeon-rx-495.htmlAMD could be moving towards xx5 donominators say 480 and then later on a 485 for their graphics card series, indicating faster models compared to 0 suffix based models. The company also reaffirms the arrival of the Radeon RX 490 model.
These xx5 revisions may actually be released after the initial batches and this are released after the current line of available products, thus these would be updated GPUs with say optimizations on 14nm FinFET chips for higher clock speeds and lower consumption. While AMD hasn't confirmed this is the case it certainly is indicative. So inevitable you might see say a Radeon RX 485 with the 4 for Generation, the 8 is the tier and the 5 being the revision. Striking is the mention of the '9', which AMD again the arrival of the RX 490 card confirms indirectly.
Tier versus performance things wil look like this, and if you look closely at the slide you can see the xx5 shown under 'Revision':
You can also spot the Tier 9 in there indicating Radeon RX 490, it is listed with a 256-bit wide (or higher) memory bus and tagged as a 4K capable graphics cards. AMD has not officially confirmed of the revisions, but pointed to the possibility of it with this slide.
Unfortunately it is not just OC but even changing the power target while leaving clocks alone.
Who knows how many gamers will mess around even with that parameter and assume they are not exacerbating the power consumption characteristics because they leave the clocks alone.
Cheers
I'm not sure that's really different than setting the clock or voltage levels differently.
It's effectively the same as changing both.
It's a factory setting, and Overdrive seems to void warranties.
It's effectively the same as changing both.
It's a factory setting, and Overdrive seems to void warranties.
Which in itself is even more overclocking than normal overclocking. In the sense of operating more parts of the machine out of spec.
Yeah makes sense but comes down to interpretation.
Did AMD clarify like Nvidia what can still be done under warranty?
For Nvidia increasing power target does not affect warranty, I assumed maybe AMD was the same.
Thanks
Last edited:
Just to put it into perspective.
This is how AMD overcame the PCI Sig certification and PCI Express specification with the 295X2.
Older chart so can be a bit more confusing as they do not use the term mainboard, but apply PEG specifically to the mainboard PCI Express slot.
Here it can be seen they only used average 28W for the slot, while putting all the pressure on the auxiliary 2x8pin connectors, which is designed to handle this stress more than the PCI-Express slot with ATX12V 24-pin connector.
As can be seen the 295x2 power demand is 400W average via the two auxiliary 8-pin connectors; by PCI Express spec they are rated at 300W combined.
Going with the ATXv12-Molex spec rating 8A for this type of connector and that figure increases to 576W combined and why AMD was pretty indifferent about the draw, albeit emphasising decent PC system.
Anyway notice they put all the pressure on the auxiliary connections and zero on the mainboard PCI Express x16 slot (in their chart named PEG).
Cheers
This is how AMD overcame the PCI Sig certification and PCI Express specification with the 295X2.
Older chart so can be a bit more confusing as they do not use the term mainboard, but apply PEG specifically to the mainboard PCI Express slot.
Here it can be seen they only used average 28W for the slot, while putting all the pressure on the auxiliary 2x8pin connectors, which is designed to handle this stress more than the PCI-Express slot with ATX12V 24-pin connector.
As can be seen the 295x2 power demand is 400W average via the two auxiliary 8-pin connectors; by PCI Express spec they are rated at 300W combined.
Going with the ATXv12-Molex spec rating 8A for this type of connector and that figure increases to 576W combined and why AMD was pretty indifferent about the draw, albeit emphasising decent PC system.
Anyway notice they put all the pressure on the auxiliary connections and zero on the mainboard PCI Express x16 slot (in their chart named PEG).
Cheers
- Status
