-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
Polaris refresh?
- Thread starter orbitalwalsh
- Start date
Well it looks like either the leak is wrong or XFX are not using the new revision, as they I cannot see how both can be correct.
Either the leak is correct and a new lower power Polaris will be the XX5 branded part or XFX is not using the new revision, which of course then points to XFX being cagy about their cards TDP.
Also to be honest this is the first leak that I have seen with news of a Polaris revision and XFX already have cards in the shops for sale, that in itself kinda makes the leak irrelevant.
Whether there will be a revised Polaris part or not I don't know, but personally I think XFX are doing something funny with their numbers.
Either the leak is correct and a new lower power Polaris will be the XX5 branded part or XFX is not using the new revision, which of course then points to XFX being cagy about their cards TDP.
Also to be honest this is the first leak that I have seen with news of a Polaris revision and XFX already have cards in the shops for sale, that in itself kinda makes the leak irrelevant.
Whether there will be a revised Polaris part or not I don't know, but personally I think XFX are doing something funny with their numbers.
Like what? lie?
It just so happens to score higher than any other 480 in benchmarks, overclock higher than anyother RX 480, 1475Mhz with 0 throttling Unigine on a loop....
At least study it before dismissing it.
Maybe the TDP is being misreported by the card, or something, we need to see more tests with this card.
I just don't see how if this is a new revision with much better power efficiency, we haven't heard more about it, before this card hit the shelves. Like AMD mentioning it themselves, seeing as it is quite a game changer for the RX480 in general.
I just don't see how if this is a new revision with much better power efficiency, we haven't heard more about it, before this card hit the shelves. Like AMD mentioning it themselves, seeing as it is quite a game changer for the RX480 in general.
No because he said in that said video that his MSI 480 which is watercooled can't overclock as far as the XFX which is air cooled and the MSI still draws more power using the same method of measuring the power. around 40 watts more power and it has a lower overclock.
Did u guys even watch the video?
Yes I did watch the video and if al the number are accurate then it is an excellent card, putting the RX480 right back in the game power wise.
I am just a bit surprised that with AMD touting the Polaris release as their most efficient card ever, that they haven't said anything about this new revision or that XFX have a card that uses considerably less power than every other rx480 out there.
I am just a bit surprised that with AMD touting the Polaris release as their most efficient card ever, that they haven't said anything about this new revision or that XFX have a card that uses considerably less power than every other rx480 out there.
exactly this, It's reported gpu die power not board power consumption . All that's going on is the xfx gtr at stock is fine tuned with a golden low leakage die, and a bios has been written for this particular gtr using 1.05v at 1288mhz, I'd imagine the memory controller is undervolted to 900/950mv too.
When he overclocks the card 1475mhz at 1.175v is fantastic scaling and the gpu die is seeing 130w-138w avg, 150w peak (160-170w average/ 200wpeak board consumption )with great temps.
Here's my measured board power of my rx470 during firestrike,
this includes mem ic, mem chips whole board power unlike jays.
1.225v at 1370 1850 250w
1.20v at 1350-1850 230w
1.175 190w
1.150 170w
1.125 140w
1.100 110w
1.000 95w
Amd's current yield problems are with the variance in leakage across the silicon.
They played it safe with gcn 1.0 on 28nm but still even then they applied too much voltage and were downclocked too much.
No doubt as time goes on the refinements in the silicon will see reduced power consumption and higher average clocks.
Is the Xfx gtr what the rx480 could have been? certainly but it's not representative of the average.
End of the day a 1450-1500 MHz rx480 is still way behind a Gp104 for performance per watt.
edit
Last edited:
Ok re-watched the whole thing listening as well as watching, and saw he mentioned the reference watercooled card he has a vid on.
Here mentioned:
And he's remembered it wrong because here's the test:
So it's a difference of ~20W on average (133W vs 153W), at 1475 MHz vs 1470 MHz. With the reference crashing after a little while.
Managing 5 MHz extra, plus stability, with 20W less power draw under heavy overclock is VERY MUCH within the realm of a golden sample card with a custom PCB.
This is clearly NOT a revision card. AMD is just experiencing a huge variance in die quality from GloFo. There seems to be something wrong with their first revision of 14LPP, and there's reports they've improved it and are unofficially calling the new one '14LPP+' (which coincides with this speculation of an RX 485).
Some more figures showing the variance in die quality, rubbish Gigabyte card:
Last edited:
I'm fairly sure the whole thing with the power setup is a symptom of this - with properties more like expected across the range it would have been entirely 100% a non-issue.
The card you are talking about was already another 480 that surprised him how little power it was using at 105 Watts, compared to the previous card he reviewed was pulling 125 to 135.
The XFX GTR was pulling 85 watts.
I don't think there is anything new about these cards, i just think the production of the dies from the node was not good, with them now maturing AMD may be getting the right power envelopes.
Last edited:
Soldato
- Joined
- 19 Feb 2011
- Posts
- 5,849
Problem with this now is you cant be sure if you are getting a first phase card or a newer phase refined card, if they indeed have matured the process.
AMD need to actually state if they have improved the chip and mark them up as so, so people buying know exactly what they are getting, otherwise you will have a spate of people selling cards as "the newer better version" when infact its just old stock.
There needs to be a whole lot of clarity if they have indeed refined their process and achieved better power envelopes and better clock speeds.
AMD need to actually state if they have improved the chip and mark them up as so, so people buying know exactly what they are getting, otherwise you will have a spate of people selling cards as "the newer better version" when infact its just old stock.
There needs to be a whole lot of clarity if they have indeed refined their process and achieved better power envelopes and better clock speeds.
The whole point is it hasn't happened yet, and when there's a significant difference they'll be launching the RX 485.
The current situation is a ~45W power draw difference, and a ~100 MHz increased overclock range, between 'perfect' golden sample cards and really really bad cards.
So it's not a great situation, but not insanely outside normal variance.
They'll only have something to write home about once a completely normal average card can hit 1450 MHz while consuming ~120W. And be sitting at ~75W at stock 1266 MHz (this being the GPU die draw afterburner shows, rather than total board power)
Associate
- Joined
- 9 Aug 2016
- Posts
- 60
Where on the card would I find this info?
Associate
- Joined
- 9 Aug 2016
- Posts
- 60
I'll go dig it out of the box I had put it back in for RMA.
Associate
- Joined
- 9 Aug 2016
- Posts
- 60
2.3, and there's an unmarked bios switch, very helpful.
This sod kept artifacting and was extremely noisy.
Soldato
- Joined
- 25 Sep 2009
- Posts
- 10,244
- Location
- Billericay, UK
That makes senese, you only have to look at what AMD and GoFlo have managed to achieve with 28nm with the latest Bristol Ridge CPU's to see what improvements in manufacturing (along with chip design) can achieve.
While nice for a select few consumers it would just add confusion to have another stock tier to 480, particularly if they plan to release a 485, and it would also gut the market for their older 480 cards giving retailers a hard time shifting that AMD stock.
Better to start producing the refined silicon and continue to sell it as a 480 at the original clock targets. Didn't Nvidia do similar where later Maxwell was able to clock/boost considerably better than early Maxwell dies?
