XFX GTR RX 480 running 1500 core / 2100 memory on AIR.

I like the XFX rep explaining their component choices in the thread. XFXSupport I hope there will be a water block that fits your Vega GTR lineup. Is there one that fits the RX 480 GTR cards? Will the reference ones fit?
 
Did they look at the chip to see if it is a new revision or software indicating a revision?

Edit: GPU Z says it is rev C7 - is that the same as launch?
 
Last edited:
Looks like Glofo ironed out the kinks in the process and are making more capable chips now. Or XFX has some secret voodoo in their PCB design.
 
Rvenger said:
Looks like Glofo ironed out the kinks in the process and are making more capable chips now. Or XFX has some secret voodoo in their PCB design.
Click to expand...

The XFX rep says that they are buying premium parts and binning the cores in the Reddit thread. He chimes in very early in the thread and answers a lot of questions.
 
That's how you sell more cards I guess :) Their product quality is getting better and better every year since the DD 7970s.
 
noko said:
Did they look at the chip to see if it is a new revision or software indicating a revision?

Edit: GPU Z says it is rev C7 - is that the same as launch?
Click to expand...

Its the same.
 
The interesting thing is that these high clocking 480s are only now just matching 290x's from two years ago, albeit they are doing it with much less heat and fan fare but require 200mhz more clock.
 
thesmokingman said:
The interesting thing is that these high clocking 480s are only now just matching 290x's from two years ago, albeit they are doing it with much less heat and fan fare but require 200mhz more clock.
Click to expand...

Yeah, the temps on that are about the same as my 290 running on a custom water loop. Almost want to side-grade just to cool my loop down a bit....
 
Not-bad-GIF.gif
 
cageymaru said:
Buildzoid has his new XFX GTR RX 480 running at 1500 core / 2100 memory on air. 15,099 graphics score. Guess JayzTwoCents XFX GTR RX 480 wasn't a fluke?



GTR 1500/2100 Firestrike run
Click to expand...
You did notice the 1.3V?
Good luck running that voltage all the time and not suffering a failure longer term, the power leakage-energy waste-draw must be very high (need one of the 4 sites that has the lab type measurements to accurately report this)
The absolute ceiling so far is around 1.4V-1.45V for Polaris according to some of the best extreme OCers in terms of voltage tolerances and that is with LN2 and only for benchmark world record type attempts, usual suggestion is a max of 1.2V to 1.25V (usually water and ensuring power stage is fully cooled).

This generation of 14/16nm are far from the 28nm in terms of the long term voltage abuse they can take, and even the short term; combination of density-silicon die thermal-electical characteristics-die size.
Case in point Maxwell could volt to 1.5V, now with Pascal its voltage tolerance with LN2 and only for benchmark world record type attempts is around 1.3V
The voltage tolerance is a bit higher with Samsung, but that is still only around what I mention in the 1st paragraph.

But it does show how good the XFX cooling system is, albeit having to run at 100%.
Cheers
 
Last edited:
Shintai said:
Its the same.
Click to expand...
Wow! Talking about a rather huge variance here. Looks like process enhancements if the case, hopefully the later chips are more closer in variations with lower power requirements allowing higher clock speeds. I wonder if TSMC process can also see improvements as well.
 
OK anyone with the card still needs to be careful OC if looking at 24/7 gaming type setup.
Seems half the Vcore goes to the PCIe slot and half to the auxiliary, with 1.3V that would be a hefty power draw and unfortunately until it is reviewed by Tom's Hardware or Hardware.fr or PCPer we would not see the power distribution breakdown.
I would need to check but I think the reference design is similar in that half is to auxiliary and half is to PCIe motherboard slot, when PCPer test OC that 480 it nearly reached 100W for the slot, bearing in mind that was only with 4% frequency OC (maybe 1.15V)

1.2V would be really pushing it if anything like the reference in behaviour (and no reason to think otherwise even allowing for less thermal related leakage/waste, so fingers crossed they also get to review to provide some clarity on what is a comfortable ceiling and to provide clarity on this.
Cheers
 
CSI_PC said:
OK anyone with the card still needs to be careful OC if looking at 24/7 gaming type setup.
Seems half the Vcore goes to the PCIe slot and half to the auxiliary, with 1.3V that would be a hefty power draw and unfortunately until it is reviewed by Tom's Hardware or Hardware.fr or PCPer we would not see the power distribution breakdown.
I would need to check but I think the reference design is similar in that half is to auxiliary and half is to PCIe motherboard slot, when PCPer test OC that 480 it nearly reached 100W for the slot, bearing in mind that was only with 4% frequency OC (maybe 1.15V)

1.2V would be really pushing it if anything like the reference in behaviour (and no reason to think otherwise even allowing for less thermal related leakage/waste, so fingers crossed they also get to review to provide some clarity on what is a comfortable ceiling and to provide clarity on this.
Cheers
Click to expand...


Where is that article on PCPER?
 
Rvenger said:
Where is that article on PCPER?
Click to expand...
https://www.pcper.com/reviews/Graph...ns-Radeon-RX-480/Overclocking-Current-Testing
They tested with Witcher 3.
The Metro Last Light benchmark also on that page is just a double check reference to Tom's Hardware.
Note this is a 'reference' 480, so the power draw figures would drop a little with custom models at same clocks due to thermals increasing leakage and waste energy with the reference blower design, however custom cooled models offset that advantage by having custom default clocks that are higher and make the draw even more as it then pushes closer to the top of the optimal performance envelope of the silicon-node.
Cheers
 
CSI_PC said:
https://www.pcper.com/reviews/Graph...ns-Radeon-RX-480/Overclocking-Current-Testing
They tested with Witcher 3.
The Metro Last Light benchmark also on that page is just a double check reference to Tom's Hardware.
Note this is a 'reference' 480, so the power draw figures would drop a little with custom models at same clocks due to thermals increasing leakage and waste energy with the reference blower design, however custom cooled models offset that advantage by having custom default clocks that are higher and make the draw even more as it then pushes closer to the top of the optimal performance envelope of the silicon-node.
Cheers
Click to expand...
Yup. just have to load up drivers older than 16.7.1 or turn off that compatability flag in the current drivers for this to be an issue again. :rolleyes:

How is this still coming up when it's been a non issue for months now?
 
jbc029 said:
Yup. just have to load up drivers older than 16.7.1 or turn off that compatability flag in the current drivers for this to be an issue again. :rolleyes:

How is this still coming up when it's been a non issue for months now?
Click to expand...

Some ppl love the doom and gloom. Its... halloween time!
 
CSI_PC said:
https://www.pcper.com/reviews/Graph...ns-Radeon-RX-480/Overclocking-Current-Testing
They tested with Witcher 3.
The Metro Last Light benchmark also on that page is just a double check reference to Tom's Hardware.
Note this is a 'reference' 480, so the power draw figures would drop a little with custom models at same clocks due to thermals increasing leakage and waste energy with the reference blower design, however custom cooled models offset that advantage by having custom default clocks that are higher and make the draw even more as it then pushes closer to the top of the optimal performance envelope of the silicon-node.
Cheers
Click to expand...


This has nothing to do with the XFX GTR. Completely different PCB/Design layout/VRM and also has an 8 pin. PCI-E bus will remain unaffected.
 
Unbelievable....
Can you guys read my posts in context????

I gave that as a reference pertaining to voltage and power demand, that PCPER is with 4% overclocks on a 'normal' reference 480.
Now the XFX in the OP is 1.3V.....compared to 1.1V to 1.15V of the reference card as OC by PCPER.

You guys do understand the relationship between voltage-frequency-thermals (mentioned this a few times and while it will save some power it is not going to save massive amounts)-power , which would be applicable to all 480s?
Those GPUs would need some mega binning, and then you would still be a much higher power demand with such OC.
Cheers
 
Last edited:
jbc029 said:
Yup. just have to load up drivers older than 16.7.1 or turn off that compatability flag in the current drivers for this to be an issue again. :rolleyes:

How is this still coming up when it's been a non issue for months now?
Click to expand...

Sigh,
if you are using close to the 75W for sure, even Ryan mentions motherboard manufacturers do not mind the spikes and their OC results is on the cusp of what was acceptable (if you are an engineer then no it is not acceptable), it would be a problem longer term IF it used sustained power over 95W on the slot (according to actual motherboard manufacturers).
BUT THE XFX OC AS IN THE OP IS AT 1.3V not 1.15V, while also having the same distribution of phases for the GPU core between auxiliary and PCIE slot.

Cheers
 
Last edited:
noko said:
Wow! Talking about a rather huge variance here. Looks like process enhancements if the case, hopefully the later chips are more closer in variations with lower power requirements allowing higher clock speeds. I wonder if TSMC process can also see improvements as well.
Click to expand...

1.3V for 1500Mhz that's not even stable outside of short benchmarks. And the power consumption? Looks to be around 225W on the card with little load. So how much did it draw when benched? 300W+? Plus the option that its simply golden.

What would the actual clock be for gaming usage?
 
Shintai said:
1.3V for 1500Mhz that's not even stable outside of short benchmarks. And the power consumption? Looks to be around 225W on the card with little load. So how much did it draw when benched? 300W+? Plus the option that its simply golden.

What would the actual clock be for gaming usage?
Click to expand...
Seriously you are in the wrong forum. This is an overclockers forum POWER USAGE BE DAMNED. So you can stop trotting out the sign boards claiming we must save mother earth.
 
Yeah lets just ignore the manufacturing (meaning TSMC and Samsung/GF) performance envelope spec and threshold-tolerance (voltage sensitivity) of the silicon-node and crank it to 11 for normal gaming, and lets ignore the PCIe SiG spec again as the 480 never gave anyone problems with the PCIe slot so lets also crank that baby to a fair bit above 100W.
How many here has their Skylake at 1.4-1.45V all the time for normal gaming and expect it to be fine for years?
Anyway there is a limit one should go and the extreme OCers who set world records/work for IHV give good advice on normal day workload V limits.

The OP test is interesting but does not mean it should be used permanently for normal day-gaming operation, some like to push this as far as possible just to see what can be achieved.

Cheers
 
CSI_PC said:
Yeah lets just ignore the manufacturing (meaning TSMC and Samsung/GF) performance envelope spec and threshold-tolerance (voltage sensitivity) of the silicon-node and crank it to 11 for normal gaming, and lets ignore the PCIe SiG spec again as the 480 never gave anyone problems with the PCIe slot so lets also crank that baby to a fair bit above 100W.
How many here has their Skylake at 1.4-1.45V all the time for normal gaming and expect it to be fine for years?
Anyway there is a limit one should go and the extreme OCers who set world records/work for IHV give good advice on normal day workload V limits.

The OP test is interesting but does not mean it should be used permanently for normal day-gaming operation, some like to push this as far as possible just to see what can be achieved.

Cheers
Click to expand...
first no one here is saying you can or should. Second, as far as CPUs FX chips handle voltage fine just have to control temps. Many a FX user has run >1.6v 24/7 as have I for over 3 years now. So I am not so sure you can parallel any arch against another.

if there is a take away here it is there is the existence of 480 that can easily surpass 1400 and by OPs example, be done safely.
 
Shintai said:
1.3V for 1500Mhz that's not even stable outside of short benchmarks. And the power consumption? Looks to be around 225W on the card with little load. So how much did it draw when benched? 300W+? Plus the option that its simply golden.

What would the actual clock be for gaming usage?
Click to expand...
Sounds like a good candidate for some water and some fun time OCing the $h.. out of it. For a desktop gaming card with that kind of OC air cooled, I would pass. Still looks like Polaris has some growing room left, will they eventually hit 1600mhz? The other side of the coin is ram bandwidth limitations unless AMD starts using DX5x in the future.
 
JustReason said:
first no one here is saying you can or should. Second, as far as CPUs FX chips handle voltage fine just have to control temps. Many a FX user has run >1.6v 24/7 as have I for over 3 years now. So I am not so sure you can parallel any arch against another.

if there is a take away here it is there is the existence of 480 that can easily surpass 1400 and by OPs example, be done safely.
Click to expand...
You avoided my question about Skylake and using it permanently at 1.4V to 1.45V, which is a similar context with regards to your post of
Seriously you are in the wrong forum. This is an overclockers forum POWER USAGE BE DAMNED
Click to expand...
, this thread in many ways is being oversimplified to make it sound like this is the norm.
How is going over 100W permanently on the PCIe slot be counted as done safely????
Safe would be if he actually measured accurately the current, we only know of the 95W draw over the PCIe slot for a 4% OC of the reference model, not the XFX that is using 1.3V.

Easily surpass 1400MHz... yeah 1.3V easily surpass if ignoring as I keep saying the silicon performance envelope in terms of voltage-frequency and power draw (both distributed and leakage/waste energy from an engineering perspective), and also running the best air cooling system to date at 100% fan speed.
You any idea how loud the XFX is at 100%?
At Overclockers.co.uk owners were sending their cards back because they were too loud even at lower fan speeds with more normal clocks.
Cheers
 
Last edited:
noko said:
Sounds like a good candidate for some water and some fun time OCing the $h.. out of it. For a desktop gaming card with that kind of OC air cooled, I would pass. Still looks like Polaris has some growing room left, will they eventually hit 1600mhz? The other side of the coin is ram bandwidth limitations unless AMD starts using DX5x in the future.
Click to expand...
Some of the world record OCers were recommending an absolute maximum due to silicon and voltage sensitivity of 1.4V to 1.45V for the 480 GPU even with LN2, and that is when using it short term for beating benchmark records.
For Pascal they are saying it is around 1.3 to 1.35, although they say realistically keep it at 1.3V on LN2 for benchmarks with the same reasons.
Cheers
 
Does anyone know how much the duty cycle was changed with the drive update that patched the power issue? Let's say they made it 3:2 in favor of the PSU phases. At 150w you're fine because you're at 60/90 on average but if you start OCing depending on how much the total power draw of the card is you could start hitting that unsafe draw again but I guess you could always alter the duty cycle further through i2c.

At 1.3v 1500mhz be very reasonably be looking at 250W (including an appropriate memory OC)

Still it's a spectacularly dumb decision decision, it's pretty funny that XFX's top model falls into a 'sins of the father' trap lol.

Another thing worth noting is that as you start OCing that high then you'll be placing more and more load on 3 of the 6 VRM phases and having the load distributed unevenly certainly can't be seen as an optimal solution.

Makes no sense.
 
Power delivery can be hard coded if I am not mistaken. I can't believe there are people comparing this to the reference card. Go find some references and get some actual proof this card is garbage otherwise, its senseless trolling and ruining the discussion of something positive for once for AMD.
 
Rvenger said:
Power delivery can be hard coded if I am not mistaken. I can't believe there are people comparing this to the reference card. Go find some references and get some actual proof this card is garbage otherwise, its senseless trolling and ruining the discussion of something positive for once for AMD.
Click to expand...

Why can't you believe it ? It has the exact same problem. GPU power split 50/50 between motherboard and PSU
 
btw quoting XFX rep

We don't claim top binned chips, but we do claim premium Inductors and voltage regulators... more on that to come. The silicone lottery is just that.
Click to expand...
 
You must log in or register to reply here.
Back
Top