Monday, March 20th 2017

AMD's Upcoming RX 500 Rebrands to use LPP Process - Higher Clocks, Lower Power

AMD's upcoming RX 500 series of graphics cards is not going to set the world on fire with its feature-set. Essentially rebrands of AMD's mainstream Polaris GPUs used in current-generation RX 400 series, these have recently seen a slight delay on its time to market - now set at April 18th.

While architecture-level adjustments to this new series of cards so as to improve performance seem to be off the table, AMD is apparently looking to take advantage of manufacturing maturing and process improvements. The original Polaris 11 and Polaris 10 chips were manufactured using the Low Power Early (LPE) process, which looks to balance availability, yields, and time-to-market with performance and power. New reports peg the new dies to carry the Polaris 21 and Polaris 20 monikers, and will feature higher clocks on account of the new Low Power Performance (LPP) process.
As to the higher clocks, these apparently are only responsible for bridging the gap between the RX 480's reference and custom boards. The RX 580 will reportedly carry a 1340 MHz clock (74 MHz more than the reference RX 480), with the RX 570 carrying a much less significant 38 MHz increase over its RX 470 counterpart. The Radeon RX 560 will apparently make do with a clock speed of 1287 MHz.

These clock improvements only go so far as to allow AMD to claim a measure of increased performance comparing to their previous-generation, same architecture, one-year-old graphics cards. Vega is the only product from the company which will have some semblance of originality. A shame AMD didn't adopt some of Vega's refinements to its mainstream graphics cards.
Source: BenchLife
Add your own comment

62 Comments on AMD's Upcoming RX 500 Rebrands to use LPP Process - Higher Clocks, Lower Power

#26
efikkan
I don't care too much whether a new product is a new binning of an old or a new chip, but rather how it performs and if it improves the market situation by lowering the price of performance. RX 580 will target the same price range as RX 480, and offer a ~6% performance improvement which aligns it with GTX 1060, which is just AMD trying to catch up, not leading the way forward.
Posted on Reply
#27
Kanan
Tech Enthusiast & Gamer
Intel did the same with "7000" line of Core processors that offered zero (0%) IPC gain over Skylake, eg. i7 7700K, with mildly increased clocks too. Nvidia did 99% the same with Maxwell, they were just a inch smarter and renamed it to "Pascal", clocks got increased through the 16nm process while shader count was reduced (that's also why a high overclocked GM200 can beat a GTX 1080, 3072 shader @ 1500 MHz vs 2560 at higher clocks).
Posted on Reply
#28
efikkan
Pascal was not a node shrink of Maxwell, as evident by the increase in FPS per GFlop.
Posted on Reply
#29
cdawall
where the hell are my stars
behrouzSo according to you , there is still a problem , Still AMD hasn't fixed PCIE power........:rolleyes:
According to me an owner of 480's yes. Multiple different systems on a weekly basis.
Posted on Reply
#30
esrever
cdawallIt wasn't fixed it was bandaided. The fix would be to release a card that was promised. 95w of total draw, we never saw that.
Who the hell promised you that?
Posted on Reply
#32
esrever
cdawallNo one important.

www.amd.com/en-us/who-we-are/corporate-information/leadership/raja-koduri

Ever heard of him? His quote was and I am paraphrasing "more efficient than a 1080" and referenced a pair of 480's in crossfire.
So you take your own paraphrasing as promises made by other people? I wonder how you even function normally with all the broken promises in you must feel every day...
Posted on Reply
#33
Kanan
Tech Enthusiast & Gamer
efikkanPascal was not a node shrink of Maxwell, as evident by the increase in FPS per GFlop.
That's a rumour, in reality Pascal is 90-99% what Maxwell is, it has minimal changes in it, mostly regarded to VR and a few DX12 tweaks that didn't really change anything compared to Maxwell (both don't do any additional scaling under DX12).
Posted on Reply
#34
cdawall
where the hell are my stars
esreverSo you take your own paraphrasing as promises made by other people? I wonder how you even function normally with all the broken promises in you must feel every day...
I'm not going to go find his exact quote for you. Not my fault you haven't been following the Polaris sku's since release. If you want the exact wording it's on YouTube.

If that bothers you, well no skin off my back.
Posted on Reply
#35
ShurikN
KananThat's a rumour, in reality Pascal is 90-99% what Maxwell is, it has minimal changes in it, mostly regarded to VR and a few DX12 tweaks that didn't really change anything compared to Maxwell (both don't do any additional scaling under DX12).
Yeah but, maxwell to pascal had a die shrink, reduction in power consumption, some minor architectural changes and a huge increase in clock speeds.
RX400 -> 500 transition brings no actual improvements other than the process refinement. Clocks will be rather the same, nothing new will be added, and maybe we will get some better efficiency, but nothing groundbreaking.
Posted on Reply
#36
Kanan
Tech Enthusiast & Gamer
ShurikNYeah but, maxwell to pascal had a die shrink, reduction in power consumption, some minor architectural changes and a huge increase in clock speeds.
RX400 -> 500 transition brings no actual improvements other than the process refinement. Clocks will be rather the same, nothing new will be added, and maybe we will get some better efficiency, but nothing groundbreaking.
Maybe so, but I expect custom GPUs to be clocked much higher than the current ones. Ref clocks of RX 580 mean jack, I wouldn't buy them anyway. I expect custom ones to be clocked as high as 1400 MHz or even higher, making a decent improvement over RX 480 possible (like with R9 290X vs 390X one of the best GPUs AMD ever made that hold their ground against 2 generations of Nvidia GPUs). Polaris scales very good with higher clocks, the difference is quite noticable.
Posted on Reply
#37
bug
cdawallLike G92!!!
Yeah, that's not a reason to be proud of either. Though various chips at least came with some updates to the video engine. I hope RX 500 will incorporate some minor updates, too.
Posted on Reply
#38
Liviu Cojocaru
eidairaman1Nv does it aswell.
I know, I was talking about all rebrands :)
Posted on Reply
#39
Jism
Ferrum MasterThanks. I am currently sitting on nvidia, before I used a trusty 7970.

But truly they don't get it. Okay a software fault... like we casuals here don't know it.

But does not undo the fact that some boards still suffer.
Did it ever came across your mind, that 90% of who owns a RX480 does'nt have any clue on how to update the BIOS of that graphics card?

What more easier it is to offer a driver update which prevents the card taking too much power out of the PCI-E bus? It is only stated that the too much of a power from the PCI-E bus suffers a minority amount of motherboards with poor power designs.

The RX295X2 pulls way more then that RX480 does over PCI-E. Some OC'ers had burned traces on their 24pins ATX connector by going nuts on LN2.
Posted on Reply
#40
Ferrum Master
JismDid it ever came across your mind, that 90% of who owns a RX480 does'nt have any clue on how to update the BIOS of that graphics card?
So what does it have to do with me? And the issue here? In normal practice user must RMA his card if he has issues. And service/seller must solve the issue.

Seconds concerning top RX295X2... LN2 etc... are you high?
Posted on Reply
#41
medi01
RejZoRWhen someone says "higher clocks" and for new series despite being a rebrand, you think 200+ MHz. Not crappy 74 MHz.
Mhz bump is meaningless without taking into account base clock.

74 is 6% of 1200
efikkanoffer a ~6% performance improvement which aligns it with GTX 1060, which is just AMD trying to catch up
AMD caught up with 1060 a while ago (that's DX11). In DX12 it is ahead.
+6% will give it comfortable lead.
Posted on Reply
#42
HD64G
behrouzo_Oo_O Seriously ? So much Fanboy.....
Nah! Just a troll...
Posted on Reply
#43
danbert2000
medi01Mhz bump is meaningless without taking into account base clock.

74 is 6% of 1200




AMD caught up with 1060 a while ago (that's DX11). In DX12 it is ahead.
+6% will give it comfortable lead.
www.techpowerup.com/reviews/ASUS/GTX_1080_Ti_Strix_OC/29.html

Like it or not, the performance summary still shows 1060 ahead. I know you really want the 480 to be a faster card but on average that's not the case.
Posted on Reply
#44
cdawall
where the hell are my stars
danbert2000www.techpowerup.com/reviews/ASUS/GTX_1080_Ti_Strix_OC/29.html

Like it or not, the performance summary still shows 1060 ahead. I know you really want the 480 to be a faster card but on average that's not the case.
Well to be fair it shows them within 1% of each other at all resolutions on all games DX11 and DX12...That is using reference cards as well which I am curious which setting W1z is using as far as testing on the 480 goes. They have two options now with the whole power fiasco.
Posted on Reply
#45
danbert2000
cdawallWell to be fair it shows them within 1% of each other at all resolutions on all games DX11 and DX12...That is using reference cards as well which I am curious which setting W1z is using as far as testing on the 480 goes. They have two options now with the whole power fiasco.
It's 1% difference with the scale relative to the 1080 Ti, so it's closer to 2% stock to stock or 5+% with Pascal's superior overclocking ability.

www.techpowerup.com/reviews/MSI/GTX_1060_Armor/30.html

The 480 is usually a better buy because of Nvidia's stupid pricing, but the 1060 is generally faster at anything but the handful of DX12 or Vulkan games, and Nvidia has been working on the performance there in the last few driver updates.
Posted on Reply
#46
bug
danbert2000It's 1% difference with the scale relative to the 1080 Ti, so it's closer to 2% stock to stock or 5+% with Pascal's superior overclocking ability.

www.techpowerup.com/reviews/MSI/GTX_1060_Armor/30.html

The 480 is usually a better buy because of Nvidia's stupid pricing, but the 1060 is generally faster at anything but the handful of DX12 or Vulkan games, and Nvidia has been working on the performance there in the last few driver updates.
I don't think the performance difference enables you to play at higher settings on the 1060. But I still got one because of Linux support. AMD is getting there, but they're still not good enough for my needs.
Posted on Reply
#47
cdawall
where the hell are my stars
danbert2000It's 1% difference with the scale relative to the 1080 Ti, so it's closer to 2% stock to stock or 5+% with Pascal's superior overclocking ability.

www.techpowerup.com/reviews/MSI/GTX_1060_Armor/30.html

The 480 is usually a better buy because of Nvidia's stupid pricing, but the 1060 is generally faster at anything but the handful of DX12 or Vulkan games, and Nvidia has been working on the performance there in the last few driver updates.
That is fair also remember the 480 leaves the xfire upgrade path that the 1060 lacks completely and no matter what nvidia does the 1060 is still not parts wise a more powerful card. I doubt it will ever equal the 480 in DX12/vulkan games. That is design. My big thing is I don't buy for yesterday, that is unluckily pascal.
Posted on Reply
#48
danbert2000
cdawallThat is fair also remember the 480 leaves the xfire upgrade path that the 1060 lacks completely and no matter what nvidia does the 1060 is still not parts wise a more powerful card. I doubt it will ever equal the 480 in DX12/vulkan games. That is design. My big thing is I don't buy for yesterday, that is unluckily pascal.
Unless you're buying for VR performance, power and heat efficiency, or CUDA support, then it's the 480 that is behind with it's 30-50 watt higher TDP and lack of simultaneous multi projection. By the time DX12 or Vulkan is the only supported API for new games, both cards will be outdated.

www.techpowerup.com/reviews/MSI/GTX_1060_Armor/12.html

Half a frame behind at 1080p, looks like the difference between the cards is that the 1060 is faster at every DX11 game, and just behind in DX12 games. Just how long are you willing to stick to a 480 to make that jump in performance really worth it while everyone else is playing DX11 games that are still coming out in 2017?
Posted on Reply
#50
cdawall
where the hell are my stars
danbert2000Unless you're buying for VR performance, power and heat efficiency, or CUDA support, then it's the 480 that is behind with it's 30-50 watt higher TDP and lack of simultaneous multi projection. By the time DX12 or Vulkan is the only supported API for new games, both cards will be outdated.

www.techpowerup.com/reviews/MSI/GTX_1060_Armor/12.html

Half a frame behind at 1080p, looks like the difference between the cards is that the 1060 is faster at every DX11 game, and just behind in DX12 games. Just how long are you willing to stick to a 480 to make that jump in performance really worth it while everyone else is playing DX11 games that are still coming out in 2017?
I mean if the argument is features... Umm well there is a reason people don't bitcoin on nvidia. Amd supports 6 video streams per gpu etc.

I also don't normally stick with cards that long out side of the 7950's I have. I play at 4k and with two cards that were purchased at separate times. The 480's fit my needs better. I am personally waiting to see what Vega does and to see if nvidia can release something with a short pcb that will fit in my case. If the 1080Ti hit with a water block and sub 10" pcb I would own one, but it sounds like I'll be waiting for Vega to do that.
Posted on Reply
Add your own comment
Apr 27th, 2024 00:46 EDT change timezone

New Forum Posts

Popular Reviews

Controversial News Posts