AMD first designed bigger chip, and they named it P10.
Then they worked on smaller chip, and they named it P11.
AMD Polaris architecture to succeed Graphics Core Next
- Thread starter wachie
- Start date
AMD first designed bigger chip, and they named it P10.
Then they worked on smaller chip, and they named it P11.
Adry9
Member
Something between a 980-980Ti for $250-$300? That sounds really good.
D
Deleted member 465307
Unconfirmed Member
Err, nevermind. I think I misunderstood the intent.
DeepEnigma
Gold Member
Ah ok. So P10 pushes more power than the P11?
Meaning it is a more powerful part performance wise, or are they comparable?
chaosblade
Unconfirmed Member
I doubt we will see a 40CU/2560 shader part until yields are better. Probably next series as a 575 or something.
P11 is the small chip, P10 is the medium chip, Vega is the big one. Price, performance, and power consumption will all correlate.
I don't think that's necessarily true though. AMD was ahead of the node transitions for so many years, even with their flagship cards, and nVidia still came ahead. Even during the Fermi days when nVidia was like 6 months behind AMD still only managed to get like 50% marketshare and they had nVidia beat on every single metric except for raw performance. People like to perpetuate AMD's power consumption, but Fermi was way worse relatively speaking. So even when they have a competitive product with better price/performance and performance/W, they still get swallowed up by nVidia's marketing and mindshare. Either that or people just wait for nVidia's responce.
Now they're focusing on the mainstream where you move more volume and can get OEM/mobile wins. They just don't have the money to do a complete top to bottom refresh, so they get the smaller chips that are (I assume) cheaper to design and validate while they take their time to not botch their end high end cards like they did with Fiji and to an extent Hawaii.
I don't know whether this is a good strategy or not. I guess time will tell.
RenegadeXV8
Banned
AMD had some serious supply issues with the 5800 series. Unless you checked every single day, it was borderline impossible to get one for months. By the time that was sorted out, Fermi was close.
I want to believe. My guess is that the center result is somewhere around where stock P10 will realistically land for $249-$279, with a shit ton of OC headroom due to the node shrink allowing it to on average knock heads with the Fury X.
If it plays out like I just listed I'll buy two.
Unknown Soldier
Member
What? How are you supposed to start a price war at 1/2 the performance of the competitors top shelf part and presumably 30% lower than their mainstream part if 1070 really does match the 980 Ti?
You are literally in a lower performance bracket. You ain't starting a war with jack shit until the 1060 shows up .
SneakyStephan
Banned
What?
Maybe I misunderstood your previous post:
What I thought you were saying: "amd can't offer better performance/dollar at the midrange"
Now you're saying the opposite
So either I'm misunderstanding you now, or then
Suzushiiro
Banned
I'd imagine the market for PC gaming hardware as a whole may be growing but the market for $200+ video cards has stopped growing or maybe even shrank because minimum specs aren't inflating as quickly anymore. Hell, most of the more popular PC games (LoL, WoW, etc.) are playable on integrated Intel chipsets if you're okay with super-low settings. >$400 cards are pretty much only relevant if you insist on 60+fps and max settings and supersampling and what-not.
dgrdsv
Member
Ehm, not really. Puts them in exactly the same spot they've been occupying with some remissions since G80 launch basically - they'll provide a bit more performance for the same price or will ask a bit cheaper price for the same performance while completely abandoning some of market segments for the time being.
GCN 1.2 wasn't faring better than Maxwell in any metric imaginable. Unless Polaris is a big architectural change from GCN3 nobody should expect the situation to change significantly between it and Pascal compared to how it was between Maxwell and 200/300 series.
No, I'm not saying the opposite. AMD has no way of offering better perf/dollar in the mid range. Both companies are tied to production costs in the mid range and they are comparable for both Pascal and Polaris as both chips use new production lines of external foundries. The difference in perf/dollar in the mid range has never been big enough to justify a choice of one vendor over another. No reason why it suddenly will be with Polaris vs Pascal.
Top end is a different story as there are no price ceiling basically as the prices are regulated by demand and competition mostly and this is where AMD may easily undercut NV with their $1000 Titan or $800 1080Ti or even $600 1080. But that won't happen until Vega which is October at best, a year from now at worst.
michaelius
Banned
Fermi managed to take performance crown which gave Nvidia - performance at all cost market, and also it had quite nice compute capabilities which also helped when we had big boinc/folding boom. And most importantly Nvidia had gtx 460 which offered amazing value and overclocking potential.
icecold1983
Member
right now the situation is
380/380x > 960
390 > 970
390x > 980
furyx < 980ti
in the market segment they are addressing they have gone from faster(sometimes significantly) and cheaper to being significantly slower and slightly cheaper. amd being better on 3 out of the top 4 gpu tiers to me means gcn is faring better than maxwell
Yeah, I know I was being reductive, but my point mostly was that in a head on fight AMD just kept losing. They dropped their small die strategy to focus on compute and that's when GTX 680 vs 7970 happened. They got beat by a smaller die that used less power because nVidia decided to beef up their second tier chip that mainly focused on gaming while AMD had a more compute-heavy card that only managed to beat it after months. nVidia has the resources to keep making monster dies they mainly use for HPC (and halo cards), but now they're also focusing on making really big gaming-focused dies like GM200 and the now rumored GP102.
AMD doesn't have that kind of luxury to focus on compute and gaming while being competitive and also keeping power consumption in check. For all we know small Vega might be able to compete with GP104, but I wonder at what cost. If it's going to be bigger, more expensive (if it uses HBM2) and more power hungry, the Maxwell cycle will just repeat. And then there's this hypothetical GP102 that might be GM200's successor that will make matters worse. I can see it be similar to what they did with GM200 and GK210. Since GM200 had its DP cut, nVidia did a respin of GK110 for the HPC market and used GM200 for desktops.
I don't really know where I'm going with all this. I don't want to come across as another harbinger of doom and gloom as I'd love to see AMD do well. I guess they have to find their niche and execute well. If they price Polaris right, they'll probably be interesting products. Let's just hope Vega ends up doing the same thing.
Wowfunhappy
Member
Why do desktop users really care about power consumption anyway? It's hard to believe this type of thing could really make a dent in overall electricity bills.
Nikodemos
Member
Because marketing/advertising.
mugurumakensei
Member
Power efficient = low temps = more overclock capability
icecold1983
Member
no not really
people only care when its in favor of their preferred IHV. theres no logical reason
4870 vs 260? The 4870 was like $150 cheaper when it launched and offered better performance.
D
Deleted member 465307
Unconfirmed Member
Personally, if I have two options that both would achieve my performance goals in a similar price range, I'd prefer the more efficient option. In my case, it's less about the money and more about wanting to reduce how much electricity I'm using from a community power load and environmental perspective. Even if it's not a huge difference and just something minuscule, I'm of an "optimizer" mindset and often in the "every ____ counts" camp, so I make these kinds of decisions when I can. However, I'm sure the fact that I'm even thinking about it is due to marketing, and I bet I'm making terrible decisions elsewhere in my life.
Also, I might be wrong about this (I'll be building my first desktop later this year and haven't done too much research yet), but I thought that lower power consumption correlates to lower temperatures. If that's the case, that's also appealing to me.
Not everyone has the same size/capacity power supply in their desktops.
It was heavily cited in the hundreds of Hawaii vs Maxwell forum arguments, since Maxwell was more efficient.
qa_engineer
Member
more efficient to me means I am getting similar or higher perfomance with less energy use. And when you reduce power consumption you're likely to reduce heat. This in turn means I can increase core and memory clocks to much higher levels, increase the amount of cards within my case without fear of melting steel beams (which current AMD GPUs can do), or mount a high performance card within a smaller enclosure with a lower power PSU.
Because the market leader says that it's important. If AMD overtakes Nvidia in power efficiency again, power consumption won't matter anymore.
I know how bitter that sounds, but that's really what it comes down to.
Chiggs
Gold Member
Indeed. Crossfire those two up and it appears performance is fantastic.
I have a mini ITX case. A case that small is liable to heat up quickly, especially if you introduce a power-hungry graphics card that produces a lot of heat. More heat with nowhere to go would work the fans more, making it noisy, which I don't want.
This.
End of power or Dennard scaling (~2006) has forced IHVs to get better energy efficiency from hardware.
Unknown Soldier
Member
And amusingly just a couple generations before when it was Fermi, the same arguments occurred but in the other camp's favor. Oh well.
Yeah but then you have to deal with Crossfire. Nobody in their right mind picks 2 weaker GPUs to offer similar performance to 1 more powerful GPU. You just don't do this. Multi-GPU is just not very good. The only "good" multi-GPU that ever existed was the original 3dfx Voodoo SLI which actually interleaved the scan lines and gave the user exactly double the performance in all cases.
dgrdsv
Member
Right, so all old Maxwell cards are mystically "worse" than all new Radeon cards (although that's highly debatable in case of 380 vs 960 for example; and in case of GM204 vs Hawaii there are some cases where Hawaii's TDP is an issue as well). And the only "new" Maxwell card is better than the new Radeon card. What does that tell you? To me this tells that NV hadn't updated their lineup since 2014 while AMD has such update in 2015.
In general terms Maxwell is a better architecture as can be seen from 980Ti > Fury X and 960 = 380. The 900 series lacked clocks and VRAM because it was launched nearly a year before Radeon 300 series. With Pascal and Polaris launching rather close to each other this is unlikely to be the case again.
They haven't gone anywhere. AMD has launched 300 series.
And NV quickly reacted with 260-something-something which was on the same perf/dollar level if not even better. The space for such maneuvers is very limited in the mid range segment.
Nikodemos
Member
What do you mean by 'new'? Apart from the Fury, none of the listed Radeons are new. The 380/380X are Tonga, which first appeared as the 285 (in fact, the 380 is the 285), while the 390/390X are the old Hawaii chips (2013!) with a fresh coat of paint and improved coolers.
icecold1983
Member
this. amds 2013 chips beating nvidias 2014.
Tortahegeszto
Member
With much higher power consumption and heat and a bunch of games seems to prefer the maxwell cards. Even games like the dx12 Forza 6 Apex.
Rushersauce
Banned
You mean the Forza 6 Apex in which the r9 390 gets 10 more FPS than a 970?
icecold1983
Member
yep. very few games prefer maxwell. the overwhelming majority run better on GCN. and power consumption is meaningless. its also worth putting in perspective that the 10 fps improvement the 390 offers in forza is ~20%.
http://www.techpowerup.com/reviews/Sapphire/R9_390_Nitro/27.html
http://www.techpowerup.com/reviews/ASUS/R9_380X_Strix/27.html
http://www.techpowerup.com/reviews/AMD/R9_Nano/36.html
http://www.techpowerup.com/reviews/Sapphire/R9_Fury_Tri-X_OC/34.html
those super high temps tho.....what will you ever do with all that heat?
What?
You can imagine that world, and I will still prefer cards that consume less power, if I am able to do so. AMD has been very good on me so far.
Lactose_Intolerant
Member
Yeah, I never heard anyone complain about this until this past generation of cards. Maybe because AMD was so much higher though.
It is a consideration now if you aren't buying new everything. You probably had a power supply that can cover a nvidia product. There is a decent possibility you need to buy a new one to cover a 390 upgrade for example.
The GTX 480 was panned for its high power consumption.
Indeed. And the whole FX series was completely obliterated (rightfully so) for its temperatures and inefficiency.
Generally, I think many people who believe to perceive this overwhelming bias in the market simply lack a long-term perspective.
Personally I don't even care about power consumption much, but there have always been people who do, and with the spread of ITX and other SFF systems it would only make sense for this awareness to increase.
amstradcpc
Member
Well, we could consider (being features equal) an architecture with better perf/watt as more elegant. That´s a selling point for some, above all the tech geeks.
That´s why people still remember Kyro 2, AMD R300, Nvidia G80...or AMD Athlon 64 and Intel Conroe...
Of course, also in the end a more efficient architecture will be capable of giving you more powerful cards.
The Kyro 2 was amazing, too bad rendering tech developments kind of worked against it.
NeOak
Member
Oh man, I still remember the press laughing at the dustbuster.
But yes, the problem with the 290/390 was the really awful reference cooler. It does have more power consumption, but the only time that will be a problem is when you buy those $20 "750W" PSUs or you are doing SFF builds like others here. Then you can't use it, but that precludes other stuff from both camps.
Oh wow, I remember wanting one while reading about it.
dgrdsv
Member
What I've said. It doesn't matter that the chips aren't new, the cards are and the chips in them are upclocked and repriced to better fight off the 900 series. Nothing strange that they are a little bit faster for the same money as that was the whole point of updating 200 series a year ago. What matters is that Maxwell gives the same performance with a lot lower power consumption and a lot less transistors spent which means that it is factually better than GCN2 or GCN3 architecturally. So unless Polaris (and 14LPE to some degree) will provide a big architectural jump, Pascal can easily end up being faster for the same price not only in the high end but across the lineup.
Tortahegeszto
Member
Dang it, I only saw the 750Ti vs 360 DF video. My bad.
For the perf/watt debate: I really don't give a damn until my PSU can support it. But more power = more heat -> louder fans. And I really don't want a vacuum cleaner on my desk and even less I'd like to witness any thermal throttling.
icecold1983
Member
more power /= more heat as ive just shown. nor does it create more noise. every card i linked is just as quiet or quieter than comparable nvidia cards. a bad cooler = more heat and more noise
they are upclocked by a whopping 5% and even the original 2 series cards perform better in most games anyway
If you use 1440p or 4k Fury X = 980Ti, the die size is basically the same as is power consumption. Maxwell does not really seem that much better. In some DX12 titles Fury X is far faster than 980Ti too and I bet in a year's time GCN will keep chugging away as Maxwell does a Kepler and starts to look far worse by comparison.