RX480 power useage vs nvidia...

Are you purposely trolling and leaving out the most important part?

and the speed of the memory interface, which in this case is an unprecedented 8Gbps for GDDR5

The 1070 says Hi and LOL.

The 1070 offers almost twice the performance of the 480, has the same 8 GB of GDDR5 with 256-bit interface delivering 256 GB/s of memory bandwidth, while drawing less power.
 
The 1070 says Hi and LOL.

The 1070 offers almost twice the performance of the 480, has the same 8 GB of GDDR5 with 256-bit interface delivering 256 GB/s of memory bandwidth, while drawing less power.

That's really good for the 1070 but what does that have to do with the point I was making? And again, no thread crapping please...
 
That's really good for the 1070 but what does that have to do with the point I was making? And again, no thread crapping please...

The point is "the speed of the memory interface" is a really bad excuse, because there was a card with the same speed interface released a month earlier which has no power issues.

The problem has been found to be that for some unknown reason AMD cut corners in hardware design and 2 of the power phases which power the memory draw directly from the PCI-e slot which causes use to go over the 75W spec.

AMD is promising a driver fix which will likely be underclocking the memory to stay in spec, which will probably be a performance loss. I suspect there is no way to work around this hardware design flaw in software without affecting performance but we will see what this fix is soon enough.
 
The point is "the speed of the memory interface" is a really bad excuse, because there was a card with the same speed interface released a month earlier which has no power issues.

The problem has been found to be that for some unknown reason AMD cut corners in hardware design and 2 of the power phases which power the memory draw directly from the PCI-e slot which causes use to go over the 75W spec.

AMD is promising a driver fix which will likely be underclocking the memory to stay in spec, which will probably be a performance loss. I suspect there is no way to work around this hardware design flaw in software without affecting performance but we will see what this fix is soon enough.

What you just wrote here is fine, the previous post wasn't. That was MY point. This isn't hardforum and we don't want it to get out of hand here just to get a reaction. Thanks...
 
Just to note I was not trolling nor have I trolled this thread what-so-ever in previous pages.

I just didn't notice they were talking about 8Gbps on that small of a bus.
 
Just to note I was not trolling nor have I trolled this thread what-so-ever in previous pages.

I just didn't notice they were talking about 8Gbps on that small of a bus.

Sounds like they were talking about the speed of the 8GB on the 256bit memory bus. Which still doesn't make a hech of a lot of sense? This just doesn't look good and being the 4th weekend news is at a stand still on this.
 
Sounds like they were talking about the speed of the 8GB on the 256bit memory bus. Which still doesn't make a hech of a lot of sense? This just doesn't look good and being the 4th weekend news is at a stand still on this.

Yeah I don't understand what they're talking about. 8Gbps as a unit of measure doesn't make much sense in terms of memory speed..
 
It does make sense, it has 8Gbps transfer speed, which is just another way of describing clock speed when talking about memory.

Both the GTX1070 and 480 run GDDR5 at 8Gbps, which is about the limit for GDDR5. The GTX1080 runs GDDR5X at 10Gbps.
 
It does make sense, it has 8Gbps transfer speed, which is just another way of describing clock speed when talking about memory.

Both the GTX1070 and 480 run GDDR5 at 8Gbps, which is about the limit for GDDR5. The GTX1080 runs GDDR5X at 10Gbps.

Wow, I totally misread that as 8GB memory instead of speed. Now Nunz's post makes more sense....oops. :):(

Yeah, I think that is at the upper limit. Not sure why they can't saturate the 6-pin first then use the pcie as needed, but I guess its a balancing act?
 
The point is "the speed of the memory interface" is a really bad excuse, because there was a card with the same speed interface released a month earlier which has no power issues.

The problem has been found to be that for some unknown reason AMD cut corners in hardware design and 2 of the power phases which power the memory draw directly from the PCI-e slot which causes use to go over the 75W spec.

AMD is promising a driver fix which will likely be underclocking the memory to stay in spec, which will probably be a performance loss. I suspect there is no way to work around this hardware design flaw in software without affecting performance but we will see what this fix is soon enough.

Or they can do a small core undervolt and remind everyone of ocing which does push vids of any brand out of spec and voids warranties.
 
Or they can do a small core undervolt and remind everyone of ocing which does push vids of any brand out of spec and voids warranties.

There is a huge difference of burning out your card vs burning out your motherboard due to overclocking. You should not be allowed to overclock that card. Those cards should have an 8-pin PCIe power connector or both 8 and 6 pin PCIe power connectors.

Exceeding your power budget is not part of overclocking and anyone that thinks that way is reckless and delusional.
 
There is a huge difference of burning out your card vs burning out your motherboard due to overclocking. You should not be allowed to overclock that card. Those cards should have an 8-pin PCIe power connector or both 8 and 6 pin PCIe power connectors.

Exceeding your power budget is not part of overclocking and anyone that thinks that way is reckless and delusional.

Overclocking almost always results in higher power usage what are you on about?
 
There is a huge difference of burning out your card vs burning out your motherboard due to overclocking. You should not be allowed to overclock that card. Those cards should have an 8-pin PCIe power connector or both 8 and 6 pin PCIe power connectors.

Exceeding your power budget is not part of overclocking and anyone that thinks that way is reckless and delusional.

Do you mean the specs of the slot and or power cables? OC'ing always increases power usage.
 
[yt]E_E2eqtm4Yw[/yt]

The 6 pin is setup as an 8 pin and can pull 250w...

I think the fix will get at that but the ram is fed by the pcie and is separate from the 6 pin which feeds the gpu so how they deal with that will be interesting.

If anything I wouldnt oc the ram. I dont think it makes sense on a 480 anyways as the 1070 has the same ram and mem bw.
 
[yt]E_E2eqtm4Yw[/yt]

The 6 pin is setup as an 8 pin and can pull 250w...

I think the fix will get at that but the ram is fed by the pcie and is separate from the 6 pin which feeds the gpu so how they deal with that will be interesting.

If anything I wouldnt oc the ram. I dont think it makes sense on a 480 anyways as the 1070 has the same ram and mem bw.

The fix they will make is simple. Set a hard cap for power usage. Once it reaches 150w it will then throttle the card. But this in turn will probably cause lower GPU speeds thus probably lowering performance. So sites like Guru3D, Tomshardware and TPU will then redo the testing of the cards and they will probably score lower.

The issue they may run in to is how the the card is throttled. Is the power system independent form PCIE and PSU power? So when the PCIe power reaxhs 75w it then throttles or does it throttle the entire power system? This can in turn effect both memory and GPU at the same time. The throttling could be severe.

It just shows that this card was not meant to run at the speeds AMD came up with. They had to release something and it was probably alot lower performer initially.
 
I dont think that will be the case. If the 6 pin can draw a lot more power I think they will adjust it if they think they need to hit a power use target or leave it alone if they want gpu ocing to be possible on the reference card. As the video shows the power from the pcie is only feeding the 8 gigs of gddr5 and the overdraw on that only happens when theres a lot of mem usage. They may put a hard cap on the pcie to the ram power and would likely prevent you from ocing the ram which imo is a waste anyway but it shouldnt prevent you from ocing the gpu itself.

Now being hardwired that way you cant feed the ram from the 6 pin but you dont need to.
 
Back
Top