[Rumor] RX 480 Overclocking 1500+Mhz

[VulgarDisplay]

https://forums.overclockers.co.uk/showpost.php?p=29646312&postcount=4094

^ Sapphire Italy pretty much confirms 1.5ghz custom models.

So using past history as a guide, about how much headroom do AIB partners leave on their oc models? Could we see another 100-150mhz on top of what they are putting out?

 

[Silverforce11]

So using past history as a guide, about how much headroom do AIB partners leave on their oc models? Could we see another 100-150mhz on top of what they are putting out?

1607mhz is the highest so far.

https://postimg.org/image/hxavd2r5n/

This is from the Chiphell user who did a livestream playing on the RX480 earlier.

 

[Silverforce11]

^ These guys.. such tease.

 

[Raising]

So.. if it ends up being equal to a 1070 will nvidia drop the prices ?

 

[Silverforce11]

So.. if it ends up being equal to a 1070 will nvidia drop the prices ?

No. Because these rumored OC RX 480 models are $299.

NV won't need to drop prices, 1070 MSRP is $379. Cards will suddenly be priced at MSRP instead of $449. No price drops necessary! JHH wins by milking early adopters, while not losing face to be forced into a price drop.

 

[William Gaatjes]

The 6 pin its not going to limit OC, at worst you gona melt the plug and the cables, but its hightly likely you gona blow up the VRMs first(or the board circuits), VRM is what limit the power, and 6 pin is able to carry more than 75W whiout melting everything down, out of spec...

6+8 will drive max power way beyond the 300W capacity, thats just crazy as hell, a 8pin alone is more than enoght.

I agree.
The 6 pin plug plug is not going to melt. The contacts are specified at 8A each and there are 3 12V contacts in the six pin connector. That allows for 12*3*8 = 288W of power. Sometimes not all 3 of the 12V pins are occupied but only 2. So that would be a limit to 192W(12*2*8). I guess the 75W number has been chosen with a lot of headroom or it may be part of the PSU spec and the wire gauge. I think 16 to 12 AWG is common.

The smps VRM would be the limit but the real limit is of course the chip itself.

 

[sirmo]

No. Because these rumored OC RX 480 models are $299.

NV won't need to drop prices, 1070 MSRP is $379. Cards will suddenly be priced at MSRP instead of $449. No price drops necessary! JHH wins by milking early adopters, while not losing face to be forced into a price drop.

Yup, as soon as I saw the FE prices during the Pascal launch event, this was the first thing that popped in my mind. Dual MSRP to milk the early adopters.

 

[nkdesistyle]

Seems about right. Nvidia never gave an actual date for standard 1070 just founders edition. It think they didn't know a whole lot about amd strategy but left room to drop the price to 379 if needed. So instead of looking stupid they decided to launch with 2 editions. You will see more and more standard cards showing up as rx 480 cards start showing up at 250-300. As always early adopters get bent over and f'ed

 

[JDG1980]

Some people are skeptical of the overclocking claims, but they seem reasonable enough. Clock speed improvements are one of the major advantages of FinFET. Apple got over 40% clock speed improvements from A8->A9 (20nm planar -> 14/16nm FinFET) at about the same power usage. Nvidia saw similar gains on the GM204 -> GP104 transition; default boost clocks went up from 1216 MHz on the GTX 980 to 1733 MHz on the GTX 1080, over a 42% increase. Overclocking was even better; the reference 980 had a max stable overclock of 1350 MHz, while the 1080 could overclock to 2114 MHz. That's over 56 percent max clock speed improvement.

AMD should be able to get the same 40% clock speed gains from the 28nm to FinFET transition. That would mean that the P10 chip will be about as efficient at the default 1266 MHz boost clock as Pitcairn was at 900 MHz. And just as virtually every Pitcairn chip can do 1100 MHz (and some AIB cards were clocked this high by default), pretty much every P10 should be able to do >1500 MHz. I expect the real-world limit for non-cherry-picked chips without excessive voltage or extreme cooling to be somewhere above 1500 and below 1600 MHz.

 

[IntelUser2000]

Some people are skeptical of the overclocking claims, but they seem reasonable enough.

Remember the claims Intel made with DC, Broadwell and Skylake? They fell short of the claims established by "leaks". Regardless of whether it was Intel guys or others the average real world overclocks did not pan out to be as much.

Good reason for people to be skeptical. Let's see how it turns out in about, 1 month.

 

[boozzer]

even if the 480's oc ability is as bad as the 1070 and 1080, it will be at least 10%.

 

[brandonmatic]

Didn't you read the whole post.

I did. I'll reserve judgment until we see benchmarks. Also the post specified that this comment was in reference to DX12, where GCN has done better than in DX11.

Here's the part you missed:

Keep in mind this is Dx12 games, this is where GCN 4 truly shows off.

 

[Qwertilot]

No. Because these rumored OC RX 480 models are $299.

NV won't need to drop prices, 1070 MSRP is $379. Cards will suddenly be priced at MSRP instead of $449. No price drops necessary! JHH wins by milking early adopters, while not losing face to be forced into a price drop.

Also pushing a smaller tier that hard to try and match a higher tier card never tends to work out that well. Destroys any power efficiency you had to start with etc.

 

[desprado]

Also pushing a smaller tier that hard to try and match a higher tier card never tends to work out that well. Destroys any power efficiency you had to start with etc.

This................... The purpose of 480 is destroyed.

 

[zentan]

Come on,have patience. We are just 10 days away from the new value King likely.

 

[AtenRa]

This................... The purpose of 480 is destroyed.

It will still have higher perf/watt than last year models (Fury X, GTX 980ti) at the same or higher performance. Not to mention 2x perf/$.

 

[flopper]

This................... The purpose of 480 is destroyed.

Not really, in the normal usage patterns the main usage is desktop browsing.
that you can boost the card with gaming using OC tools seems to me to keep the purpose of a low wattage and performance while surfing and then be able to artifically boost the card with AMD Value when your gaming are in perfect harmony with the purpose of the 480.

That the boost is made by you instead of the company selling you the card in the bios, whats the difference really m8?

 

[flopper]

No. Because these rumored OC RX 480 models are $299.

NV won't need to drop prices, 1070 MSRP is $379. Cards will suddenly be priced at MSRP instead of $449. No price drops necessary! JHH wins by milking early adopters, while not losing face to be forced into a price drop.

200 euro will be the difference here between a 480 OC to 1500mhz+ even out and surpassing a 1070 at 200 euro less. so dont judge the card from one market in the world as there are markets that have a worse priced 1070 that will be needed to go down or they loose vs the 480 priced performance.

AMD just crushed the 1070 and they will have to drop prices on that card.

 

[psolord]

Overclocking (for me) is not meant to be run 24/7. Of course its anyone's prerogative, but personally I prefer to have the option to overclock and use that option in games when my target framerate is not met. Or in the benchmarks I do from time to time for my Youtube channel.

Let's say I play witcher 3 and for the settings I am using, I am around 50ish most of the time. If I don't feel like fiddling around with the settings, I can try to get +20% performance through overclocking. Now if for the length of 2-3-5 games, I need to overclock to get the performance I want and doing so I lose the power efficiency, so be it.

If I need to overclock all the time for the performance I want, then I have gotten the wrong card, but usually I don't.

 

[VulgarDisplay]

Overclocking (for me) is not meant to be run 24/7. Of course its anyone's prerogative, but personally I prefer to have the option to overclock and use that option in games when my target framerate is not met. Or in the benchmarks I do from time to time for my Youtube channel.

Let's say I play witcher 3 and for the settings I am using, I am around 50ish most of the time. If I don't feel like fiddling around with the settings, I can try to get +20% performance through overclocking. Now if for the length of 2-3-5 games, I need to overclock to get the performance I want and doing so I lose the power efficiency, so be it.

If I need to overclock all the time for the performance I want, then I have gotten the wrong card, but usually I don't.

With the new crimson software I think you can set up oc profiles on a per game basis. It's a great feature. You could even set the card to down clock for older games to get the best power usage.

Sent from my XT1575 using Tapatalk

 

[amenx]

Also pushing a smaller tier that hard to try and match a higher tier card never tends to work out that well. Destroys any power efficiency you had to start with etc.

This................... The purpose of 480 is destroyed.

No, the advertised purpose is not destroyed, at stock clocks efficiency is maintained. If AIBs OC the cards they are not bound by power efficiency, which a good majority of buyers do not care about anyway. AMD is covered either way.

 

[Yakk]

No, the advertised purpose is not destroyed, at stock clocks efficiency is maintained. If AIBs OC the cards they are not bound by power efficiency, which a good majority of buyers do not care about anyway. AMD is covered either way.

Yup, it's a win-win scenario; AMD aims for price and efficiency, while also giving their AIBs room to differentiate their products and increase performance. Looks like a very well setup card indeed.

 

[3DVagabond]

Wut?!?!?! You mean if I O/C it'll use more power! That's it. I'm never O/C'img anything again. I wish I would have known this before.

Seriously, it's a given that heavy O/C'ing is going to kill efficiency. Everyone knows this. It's the price you pay for pushing your performance to the next level.

 

[Abwx]

Wut?!?!?! You mean if I O/C it'll use more power! That's it. I'm never O/C'img anything again. I wish I would have known this before.

Seriously, it's a given that heavy O/C'ing is going to kill efficiency. Everyone knows this. It's the price you pay for pushing your performance to the next level.

Overestimating the thing each % higher frequency will reduce perf/Watt by the same amount, so 1.25x overclock would reduce the 2.8x perf/watt improvement down to 2.8/1.25 = 2.24x.

And as said this is an overestimation since i assumed that RAM is also overclocked by 25% and that both the GPU and RAM voltage margins are kept unchanged..

 

[Qwertilot]

Isn't it normally almost entirely non linear? Some small gains for kind of 'free' then it gets less and less efficient to get more performance out. You only need to look at say Skylake at 35 vs 65 vs 90(+).

No one has used this process for anything quite like this before, architecture new too so a total guess how it'll behave