AMD RX 480M, RX 470 Benchmarks?

[HardOCP News]

The folks at Tweakers have posted a handful of benchmarks and a few leaked slides that supposedly show performance numbers for the AMD RX 480M and RX 470. The site isn't in English and Google translate only does so much so, as usual, take these with a grain of salt until we see some real world benchmarks. I snagged a few images for those of you unable to access the site:

 

[alamox]

480M with 35watt ? isn't that supposed to be 460M polaris 11 ? because thats alot of cutting down and underclocking to go from 150watt to 35watt

 

[tybert7]

480M with 35watt ? isn't that supposed to be 460M polaris 11 ? because thats alot of cutting down and underclocking to go from 150watt to 35watt

I think the 480m is the polaris 11 version of the 480.

It's performance in firestrike is right around the current nvidia 960m, except with a power envelope of 35w instead of 60w. So similar power envelope to a 940m with 960m power. Before amd gpus were not even an option on the lower end because they used far too much power, it will no longer be a clean sweep this generation on the mobile side.


Next year is where things get much more interesting. AMD will have zen based apus, presumably with polaris graphics built in. As long as zen is respectible, you will have dual gpu configs by default for anyone going with a zen mobile apu + some polaris 11 equivalent. If amd throws in some chip combo discounts to oems this could be a very popular configuration for a lot less money and bulk to get decent performance.

 

[razor1]

480M with 35watt ? isn't that supposed to be 460M polaris 11 ? because thats alot of cutting down and underclocking to go from 150watt to 35watt

The 480m is P11 but the problem with that is, its going up against a tough bracket, mainstream ( cut down) nV GPU's, 1060 or a 1050, I don't think its going to be a pretty ending.

Now Desktop cards the rx470 is the cut down version of the P10, at least that what it looks like from the slides.

So what we have here is 2 P10 GPU's, the rx 470 and rx 480.

470 is at 110 watts, the r480, will be higher, if we are talking about 20% more performance per bracket, which is fairly typical, 132 watts if there is a linear increase in performance to wattage which we know doesn't happen, so around 140 watts? Damn its getting up there to ~ a 1070 power usage at 40-50% less performance.

The 1060's if they come out soon enough will have a field day in the desktop market.

I think the 480m is the polaris 11 version of the 480.

It's performance in firestrike is right around the current nvidia 960m, except with a power envelope of 35w instead of 60w. So similar power envelope to a 940m with 960m power. Before amd gpus were not even an option on the lower end because they used far too much power, it will no longer be a clean sweep this generation on the mobile side.


Next year is where things get much more interesting. AMD will have zen based apus, presumably with polaris graphics built in. As long as zen is respectible, you will have dual gpu configs by default for anyone going with a zen mobile apu + some polaris 11 equivalent. If amd throws in some chip combo discounts to oems this could be a very popular configuration for a lot less money and bulk to get decent performance.

And lets not talk about next years, for the past 5 years its been AMD lets see they will come back next year. This is becoming a broken record.

 

[NKD]

The 480m is P11 but the problem with that is, its going up against a tough bracket, mainstream ( cut down) nV GPU's, 1060 or a 1050, I don't think its going to be a pretty ending.

Now Desktop cards the rx470 is the cut down version of the P10, at least that what it looks like from the slides.

So what we have here is 2 P10 GPU's, the rx 470 and rx 480.

470 is at 110 watts, the r480, will be higher, if we are talking about 20% more performance per bracket, which is fairly typical, 132 watts if there is a linear increase in performance to wattage which we know doesn't happen, so around 140 watts? Damn its getting up there to ~ a 1070 power usage at 40-50% less performance.

The 1060's if they come out soon enough will have a field day in the desktop market.




And lets not talk about next years, for the past 5 years its been AMD lets see they will come back next year. This is becoming a broken record.

I will wait until someone actually does actual power usage testing. That seems like they are copy pasting right there. like if they had tested rx 480 they would have basically said 150W taking straight from TDP without testing it.

I am taking that power figure with a grain of salt. Looks like they took that number from 480 150w board power and came up with 110w from there. We know AMD won't limit the board to 150w if it was actually using 150w.

 

[noko]

The 480m is P11 but the problem with that is, its going up against a tough bracket, mainstream ( cut down) nV GPU's, 1060 or a 1050, I don't think its going to be a pretty ending.

Now Desktop cards the rx470 is the cut down version of the P10, at least that what it looks like from the slides.

So what we have here is 2 P10 GPU's, the rx 470 and rx 480.

470 is at 110 watts, the r480, will be higher, if we are talking about 20% more performance per bracket, which is fairly typical, 132 watts if there is a linear increase in performance to wattage which we know doesn't happen, so around 140 watts? Damn its getting up there to ~ a 1070 power usage at 40-50% less performance.

The 1060's if they come out soon enough will have a field day in the desktop market.




And lets not talk about next years, for the past 5 years its been AMD lets see they will come back next year. This is becoming a broken record.

The past does not always indicate the future other wise businesses would never fail if that is the case. As for Nvidia wiping the floor so to speak with the 1060 and more so with the 1050 also means TSMC can deliver high volumes of GPU's (in the millions). This round it maybe a Foundry limitation for Nvidia for awhile. Right now it looks like Nvidia has no answer to the oncoming onslaught of less than $300 gpu solutions where what ~80% of the market is. If you remember when Maxwell came out it came out small (Ti750) and it was a huge landmark change for Nvidia - that did not start out on the top where AMD 290x was indeed better than a 780.

I think it is too early to tell, if AMD has limited quantities of these cards (not just selling out because most likely they will sell well) then Nvidia will have some time. Plus Nvidia may have an inventory or seller inventory issue (lower end stuff just not selling) - Things that don't sell also hinders a company perception as a whole. This is only a start - If AMD takes the performance crown as well later with their halo product, unique configurations and can sustain the volumn unlike what Nvidia is doing now with the 1070 and 1080 it will be a new ball game.

A lot of if's - hence we just don't really know.

I also agree though that the 1050/60 should be some very nice chips - as long as Nvidia can deliver them and when they do, it will also be close to Zen APU's which can also put the 1050 out on left field.

 

[NKD]

The past does not always indicate the future other wise businesses would never fail if that is the case. As for Nvidia wiping the floor so to speak with the 1060 and more so with the 1050 also means TSMC can deliver high volumes of GPU's (in the millions). This round it maybe a Foundry limitation for Nvidia for awhile. Right now it looks like Nvidia has no answer to the oncoming onslaught of less than $300 gpu solutions where what ~80% of the market is. If you remember when Maxwell came out it came out small (Ti750) and it was a huge landmark change for Nvidia - that did not start out on the top where AMD 290x was indeed better than a 780.

I think it is too early to tell, if AMD has limited quantities of these cards (not just selling out because most likely they will sell well) then Nvidia will have some time. Plus Nvidia may have an inventory or seller inventory issue (lower end stuff just not selling) - Things that don't sell also hinders a company perception as a whole. This is only a start - If AMD takes the performance crown as well later with their halo product, unique configurations and can sustain the volumn unlike what Nvidia is doing now with the 1070 and 1080 it will be a new ball game.

A lot of if's - hence we just don't really know.

I also agree though that the 1050/60 should be some very nice chips - as long as Nvidia can deliver them and when they do it will be close to Zen APU's which can also put the 1050 out on left field.

I really don't think AMD is going to have any shortage here. Looks like they knew this would be a problem, getting the best yields from the new process. Look at 1070/1080, Nvidia announced them over a month ago and they are still not readily available. Looks like AMD is making a hard launch and decided to start off with smaller chips first to wait until they can ramp up the bigger chips. This might actually play out good in the long run. Sell what you can make bunch of now and wait for the big chips later strategy..

They literally have their entire mid range lineup coming out on 29th with a hard launch and with news about stores having bunch of stock already. So looks like they are going hard on midrange.

 

[noko]

I will wait until someone actually does actual power usage testing. That seems like they are copy pasting right there. like if they had tested rx 480 they would have basically said 150W taking straight from TDP without testing it.

I am taking that power figure with a grain of salt. Looks like they took that number from 480 150w board power and came up with 110w from there. We know AMD won't limit the board to 150w if it was actually using 150w.

Bios switch on cards? I hope that continues and AMD adds a software selection of which bios - so you could have low power and one with a more performance orientated configuration. Anyways I agree, we just need to see the results (good reviewers doing a good job) and also see if AMD can deliver the large volume in this area. If they can then that will be great for AMD.

 

[jbc029]

470 is at 110 watts, the r480, will be higher, if we are talking about 20% more performance per bracket, which is fairly typical, 132 watts if there is a linear increase in performance to wattage which we know doesn't happen, so around 140 watts? Damn its getting up there to ~ a 1070 power usage at 40-50% less performance.

I don't know about 40-50%. While the firestrike benchmarks show the 1070 being 38% faster than a 980, the benchmarks here on this site show it being around 18-22% faster, where common settings exist. It is, or eventually will be, about a 980ti for $400. It is not faster. A 980ti is about 20-25% faster than a 980, not 40%.

So which do you believe? The synthetic benchmark or the actual reviews here on the site? (Reference the 980 performance used in the Fury review on May 24th against the stock and overclock reviews of the 1070. The two common test beds are for RotTR and Fallout 4, each showing the 1070 being 18-21% faster than the corresponding 980. I'm assuming that something is missing in the Witcher 3 documentation of the 1070 overclock review, because while they show the same settings in the chart, apparently the OC'd 980 is faster than the OC'd 1070...and that I just can't believe).

 

[tybert7]

The 480m is P11 but the problem with that is, its going up against a tough bracket, mainstream ( cut down) nV GPU's, 1060 or a 1050, I don't think its going to be a pretty ending.

Now Desktop cards the rx470 is the cut down version of the P10, at least that what it looks like from the slides.

So what we have here is 2 P10 GPU's, the rx 470 and rx 480.

470 is at 110 watts, the r480, will be higher, if we are talking about 20% more performance per bracket, which is fairly typical, 132 watts if there is a linear increase in performance to wattage which we know doesn't happen, so around 140 watts? Damn its getting up there to ~ a 1070 power usage at 40-50% less performance.

The 1060's if they come out soon enough will have a field day in the desktop market.




And lets not talk about next years, for the past 5 years its been AMD lets see they will come back next year. This is becoming a broken record.

This year is zen desktop. zen apus were known to be 2017 for some time now. But we won't have to wait until next year to see if both polaris and zen is worth anything.

 

[noko]

This year is zen desktop. zen apus were known to be 2017 for some time now. But we won't have to wait until next year to see if both polaris and zen is worth anything.

Yep, if the 1050 came out around Christmas time, Zen Apu's should be out less than a year afterwards which will probably kill the 1050 product line, instead of a 3 year run it will last ~1 year. That is if Zen and subsequent APU are good.

 

[razor1]

I don't know about 40-50%. While the firestrike benchmarks show the 1070 being 38% faster than a 980, the benchmarks here on this site show it being around 18-22% faster, where common settings exist. It is, or eventually will be, about a 980ti for $400. It is not faster. A 980ti is about 20-25% faster than a 980, not 40%.

So which do you believe? The synthetic benchmark or the actual reviews here on the site? (Reference the 980 performance used in the Fury review on May 24th against the stock and overclock reviews of the 1070. The two common test beds are for RotTR and Fallout 4, each showing the 1070 being 18-21% faster than the corresponding 980. I'm assuming that something is missing in the Witcher 3 documentation of the 1070 overclock review, because while they show the same settings in the chart, apparently the OC'd 980 is faster than the OC'd 1070...and that I just can't believe).

NVIDIA GeForce GTX 1070 8 GB Review

Well ya lets use games then *was too lazy to look it up*, its 29% faster at 4k overall.

 

[razor1]

The past does not always indicate the future other wise businesses would never fail if that is the case. As for Nvidia wiping the floor so to speak with the 1060 and more so with the 1050 also means TSMC can deliver high volumes of GPU's (in the millions). This round it maybe a Foundry limitation for Nvidia for awhile. Right now it looks like Nvidia has no answer to the oncoming onslaught of less than $300 gpu solutions where what ~80% of the market is. If you remember when Maxwell came out it came out small (Ti750) and it was a huge landmark change for Nvidia - that did not start out on the top where AMD 290x was indeed better than a 780.

I think it is too early to tell, if AMD has limited quantities of these cards (not just selling out because most likely they will sell well) then Nvidia will have some time. Plus Nvidia may have an inventory or seller inventory issue (lower end stuff just not selling) - Things that don't sell also hinders a company perception as a whole. This is only a start - If AMD takes the performance crown as well later with their halo product, unique configurations and can sustain the volumn unlike what Nvidia is doing now with the 1070 and 1080 it will be a new ball game.

A lot of if's - hence we just don't really know.

I also agree though that the 1050/60 should be some very nice chips - as long as Nvidia can deliver them and when they do, it will also be close to Zen APU's which can also put the 1050 out on left field.

Yeah tell that to AMD thinking they were months aheads of nV of releasing new products on FinFet nodes. If anything it tells us AMD's ability to predict what nV's up to is not very good. Even though many people outside of nV knew nV had samples of certain Pascal chips late last year in one form or another.

 

[NKD]

Yeah tell that to AMD thinking they were months aheads of nV of releasing new products on FinFet nodes. If anything it tells us AMD's ability to predict what nV's up to is not very good. Even though many people outside of nV knew nV had samples of certain Pascal chips late last year in one form or another.

Nvidia paper launched the shit out of 1080/1070. It will be a month and half soon and we still have people complaining about short supply and high prices. Its the same thing, AMD went low to mid range high supply of smaller die and nvidia launching a new node on a bigger die and not being able to supply enough in the beginning.

We will see, if AMD can't supply enough than yea they will hurt themselves big time but It doesn't look like thats the case as reports are stores have bunch of stock already.

 

[KazeoHin]

I would hope AMD would not be so stupid as to approve a silly old-fashioned naming scheme like this. I'm SICK AND TIRED of notebook GPUs being named after desktop GPUs that are easily 2-5x more powerful.

Hey look at this notebook GPU that has the equivalent performance of a GTX 950...



Lets call it the TITAN XXX GTX DESTROYER Ti M.


if the performance is equivalent of the GTX 950, call it the GTX 950M. I'm tired of having to explain to people that , no, their GTX 970m is nowhere even CLOSE to HALF the STOCK performance of an actual GTX 970.

 

[tybert7]

I would hope AMD would not be so stupid as to approve a silly old-fashioned naming scheme like this. I'm SICK AND TIRED of notebook GPUs being named after desktop GPUs that are easily 2-5x more powerful.

Hey look at this notebook GPU that has the equivalent performance of a GTX 950...



Lets call it the TITAN XXX GTX DESTROYER Ti M.


if the performance is equivalent of the GTX 950, call it the GTX 950M. I'm tired of having to explain to people that , no, their GTX 970m is nowhere even CLOSE to HALF the STOCK performance of an actual GTX 970.

I'm ok with simply having the "m" suffix denote a lower powered variation of the chip.

/tangent

I vastly prefer the nomenclature of rx 480 to r9 280x/r9 290x

rx ___ just sounds like a nicer way to say things, it flows better and should be the model of future series of gpu names to build on.

 

[KazeoHin]

I'm ok with simply having the "m" suffix denote a lower powered variation of the chip.

/tangent

I vastly prefer the nomenclature of rx 480 to r9 280x/r9 290x

rx ___ just sounds like a nicer way to say things, it flows better and should be the model of future series of gpu names to build on.

I agree with the RX naming. It's quite pleasing to read and say.

I work as a professional PC repairer and builder (with a specialisation as gaming and high-performance PC design) and the amount of times people with piss-poor Dellienware notebooks brag to me about how their 970m will be able to max-out the newest games because it meets the recommended spec, when I have to tell them that the 970m is more like a GTX 950 on the desktop. Then arguments, then anger...

You don't know what it's like, MAN. You weren't THERE.

 

[jbc029]

And links like that are why I don't use aggregators like that. Go to the actual games themselves, not the overall page. Check the resolution similar to the tests presented here (since that's all you have...no settings are given at all, not even presets used).

The RotTR tests for the 980 match up pretty well for the A2A test in the 1070 review. But somehow their 1070 was 11% faster than the [H] 1070 card in that A2A comparison? More likely, they aren't using maxed out settings like [H] was for the 1070 and they recycled 980 results from a couple weeks after TR's release and called it good.
Edit: TPU testing is on DX12, while [H] was DX11. This could account for the difference in the 1070 performance, I suppose, but no data exists to correlate yet.


Swing over to the Fallout 4 page. The 1070 is showing 100+ fps at 1440p. What are they using, medium presets? Because top end here was showing barely 60 fps at 1440p. That entire page isn't remotely indicative of what someone with that card would be using it for.

If you aren't willing to get into the details, you shouldn't vouch for a source's reliability. TPU seems to keep a large log of test results that are gathered within a month or so of each major hardware/game release. After that time has passed, they never refresh them, the aggregates become skewed and they don't reflect current performance. Since they don't keep track of improvements via patches or driver releases, their data isn't good for anything beyond rough estimates and those will almost always skew to the high side because card performance tends to gradually increase over time, not decrease. (meaning that older card/resolution/game combinations remain at their initial unoptimized/non-patched performance, which is then presented as current against the new card/resolution/game metric. This presents a greater difference between combinations of items than would actually be found in current testing)

 

[noko]

Yes, HardOCP I think should have had a 980 for the 1070 OC review for those with 980's.

Anyways by those benchmarks looks like about a 290/290x performer and not a 390x performer. It will be good at $200 if those represent gaming, we have to see how the 480 plays in games to get a real feel for it. So for us Enthusiast - this will probably be rather dull and boring but for the Mainstream and the under $250 buyers these will be much better then anything on the market now. So I don't expect too much cheering around here in the end.

Where is the SLI 1070 review pitted against 2 CFX Fury's using mostly DX12 games
-Hitman
-Quantum Break
-WarHammer
-Rise of the . . .

Wait - most of those don't support dual cards

 

[Stoly]

I rather wait for the [H] full review from an AMD provided sample...

 

[Ieldra]

Yes, HardOCP I think should have had a 980 for the 1070 OC review for those with 980's.

Anyways by those benchmarks looks like about a 290/290x performer and not a 390x performer. It will be good at $200 if those represent gaming, we have to see how the 480 plays in games to get a real feel for it. So for us Enthusiast - this will probably be rather dull and boring but for the Mainstream and the under $250 buyers these will be much better then anything on the market now. So I don't expect too much cheering around here in the end.

Where is the SLI 1070 review pitted against 2 CFX Fury's using mostly DX12 games
-Hitman
-Quantum Break
-WarHammer
-Rise of the . . .

Wait - most of those don't support dual cards

lold.

I don't know how people can be so positive about mGPU in the future, the responsibility now being in the hands of developers does not inspire confidence to me.