• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Polaris performance numbers perf/$

R

razor1

[H]F Junkie
Joined
Jul 14, 2005
Messages
10,120
Radeon R9 490X 8GB News - AMD Polaris 10 GPU Reportedly Offers Near 980 Ti Performance For 300 USD

So going by this its going to be slower than the FuryX and 980ti, but at a price that is really damn good. Looks to me, this is what AMD has been saying along, its not going up against the performance cards. Its just a bit under the performance cards. These numbers were given or seen by website journalists.

But yeah 3d mark scores used in ratio of performance per dollar......

website went off line but here is a pic of it
xIEbYLk.png
 
Last edited:
Website is offline, they must be running that site on an Nvidia GPU ;)

Also "$300" doesn't make any sense for a x90X card, that leaves little room for additional card between $150 and $300 where the majority of GPU sales are.
 
lol put the picture of the page up now. Well price with standing, I don't think there is a final price for the product yet. The 3dmarks score is around 10% less than the 980ti and FuryX, so if that is anything to go by, its a damn good performer, for $300 bucks.

Link is working again.
 
Last edited:
Sorry, but I don't buy it.

Rule number one of business: Never sell your product for LESS than you can get away with.

If the product performs like a 980Ti, It will be PRICED like a 980Ti. If anything, if AMD REALLY wanted to cause a big huffle, they would sell it for $50-$100 less than a 980Ti. $300? Get out of here.
 
If they're a nut hair under the 980ti, why would they undercut it by some absolutely insane amount like that
 
I remember the so called FuryX benchmarks - when reality hit - it didn't match to well. There is enough differences that I wouldn't be surprised that on some stuff it will be faster than a FuryX and other times slower then a 390x. We will see.
 
socK said:
If they're a nut hair under the 980ti, why would they undercut it by some absolutely insane amount like that
Click to expand...

VR games look like sh*t because cards like 980ti, Fury X, etc are too slow. You need that type of speed minimum at the mainstream. Then mGPU takes care of the high end. Remember DX12, LiquidVR, whatever the Nvidia VR spec is.
 
TaintedSquirrel said:
The 970 performed like a 780 Ti but was priced like a 770.
Plus Polaris is quite a bit smaller.

$300, <150W, 10% slower than Fury X / 980 Ti. I believe it. VERY optimistic ofc.
Click to expand...

The 970 performed similar to a 780 vanilla AND had the 980 flagship ahead of it. The 980 was only marginally better than the 780Ti, and some OC'd 780Tis actually ran faster than stock 980s. The 980 was priced higher than the 780 Ti from what I recall, and that is the MAIN reason the 970 was sitting in its position of price.

If Polaris comes out as fast as a 980Ti, or even 10% slower, expect it to be priced accordingly UNLESS it is released at the same time as a faster card in the same series, even then, AMD would not want to cannibalise sales of 6 of their products, the 380X, the 390, the 390X, the Fury, the Fury X and the Nano by releasing something more powerful, more efficient and more affordable than ALL of them.
 
I would say that everything mainstream must be able to max out the Valve VR test. GTX 950 is the new 750ti.
 
Doing the quick napkin math,
770 -> 970 was a performance increase of 32% (1080p, TechPowerup September 2014)
970 / 390 -> 980 Ti is an increase of 38% (1080p, TechPowerup March 2016)

So in order for this rumor to be true, Polaris will have to match the GTX 970's performance leap over the 770, along with an additional 18% increase.
Tbh if Nvidia can do it on 28nm I believe AMD can replicate, and surpass, on 14nm.

If it comes in ~10% under the 980 Ti then it's even more plausible.
 
KazeoHin said:
The 970 performed similar to a 780 vanilla AND had the 980 flagship ahead of it. The 980 was only marginally better than the 780Ti, and some OC'd 780Tis actually ran faster than stock 980s. The 980 was priced higher than the 780 Ti from what I recall, and that is the MAIN reason the 970 was sitting in its position of price.
Click to expand...

GTX 980 had a launch MSRP of $550. GTX 780ti had a launch MSRP of $699 although street prices had fallen to ~$600 at GTX 980 launch time.
 
This exactly like the wccf article posted yesterday, leaving performance aside, at 175w and $300 this will bump heads with the smaller gp104

I wasn't sure about 3dmark 4k scores so I checked. Fury X hits 3900.

Fury x is 8.6tflops

Granted 3dmark isn't just about shading, but still 2560 cores hitting 8.6tflops requires 1.6ghz +

Taking this with a grain of salt
 
Last edited:
yes these were the figures that WCCF used for their article, these figure were what AMD showed reporters (of course they wouldn't say anything out right), but in any case, it is what it is.
 
AMD's Radeon Polaris | Performance Per Dollar & The Immersive Era

Raja Koduri said “What you’ll see us do is completely different with Polaris 10 and 11. We are really focusing on trying to bring FinFET technology with it’s amazing performance per watt to as many segments as possible. As many ranges of performance as possible. I can tell you Ryan, you and your readers, you’ll be pleased at what we’re going to do with this thing and you’ll be surprised.
Click to expand...

Funny thing is Raja did say the perf/watt boost is coming from the node, and not directly form the architectural changes.
 
Weird how performance per watt was a completely irrelevant metric for Hawaii and Tahiti (both to AMD itself and their fanbase) but now with finfet it's the selling point. AMD continue their hypocritical streak.

Keep it at or below 250W stock and let's get higher frame rates, please.
 
Well Vega will do that. The only reason its coming out later is cause of HBM2 delays. I'm sure nV p100 launch with HBM2 in June is going to limited due to this as well.
 
razor1 said:
Well Vega will do that. The only reason its coming out later is cause of HBM2 delays. I'm sure nV p100 launch with HBM2 in June is going to limited due to this as well.
Click to expand...
Yea Vega seems like it will be powerful chip if Polaris holds true to the rumors. Unfortunate it got delayed I have no doubt chip is ready because it was suppose to be launched this year. We will see what it brings in early 2017. So far I am looking forward to this. It's amazing no one is talking about pascal AMD has gotten the hype train going pretty damn well this time. All about Polaris perf/watt and price. They are already gaining market share by all this interest people are taking in Polaris lol.
 
Well I just talked to someone that is saying Polaris will not use GDDR5x...... I don't know if he is confused as in the lower end Polaris's or Polaris 10, he wouldn't go into more details........

I don't see how it will go above a r390 performance if that is true due to the smaller bus size of Polaris 10.

Any case ya, Vega seems to be people would want if they really want a nice upgrade from high end and enthusiast cards of today.

Well the hype train thing, is because AMD has been talking about Polaris, nV the only thing they have sated so far is about p100, and that has little to with gp104 outside of architecture.
 
AMD’s Polaris will be a mainstream GPU, not high-end
The reason Polaris is a big deal, is because I believe we will be able to grow that TAM [total addressable market] significantly," said Taylor. "I don't think Nvidia is going to do anything to increase the TAM, because according to everything we've seen around Pascal, it's a high-end part. I don't know what the price is gonna be, but let's say it's as low as £500/$600 and as high as £800/$1000. That price range is not going to expand the TAM for VR. We're going on the record right now to say Polaris will expand the TAM. Full stop."
Click to expand...

And this is Roy Taylor, no hype at all lol. Guess they muzzled him good.
 
razor1 said:
Well I just talked to someone that is saying Polaris will not use GDDR5x...... I don't know if he is confused as in the lower end Polaris's or Polaris 10, he wouldn't go into more details........

I don't see how it will go above a r390 performance if that is true due to the smaller bus size of Polaris 10.

Any case ya, Vega seems to be people would want if they really want a nice upgrade from high end and enthusiast cards of today.

Well the hype train thing, is because AMD has been talking about Polaris, nV the only thing they have sated so far is about p100, and that has little to with gp104 outside of architecture.
Click to expand...

If Polaris doesn't beat 390 performance, AMD should just pack up their shit and send everyone home. That architecture launched three years ago, they should have multiple SKUs under $300 blowing it out of the water.
 
Well I don't know what he was trying to tell me as he didn't clarify, I'm thinking he is just mistaken.
 
KazeoHin said:
Sorry, but I don't buy it.

Rule number one of business: Never sell your product for LESS than you can get away with.

If the product performs like a 980Ti, It will be PRICED like a 980Ti. If anything, if AMD REALLY wanted to cause a big huffle, they would sell it for $50-$100 less than a 980Ti. $300? Get out of here.
Click to expand...

They need market share. If they can afford it, then this is what I and others have been saying AMD should, will, and is likely to do.

People will stampede to AMD if the card can deliver at that price. Then AMD can worry about increasing their margins in subsequent generations. For now, they need market share, and this is an aggressive strategy that will allow them to do just that.
 
razor1 said:
Well I just talked to someone that is saying Polaris will not use GDDR5x...... I don't know if he is confused as in the lower end Polaris's or Polaris 10, he wouldn't go into more details........

I don't see how it will go above a r390 performance if that is true due to the smaller bus size of Polaris 10.

Any case ya, Vega seems to be people would want if they really want a nice upgrade from high end and enthusiast cards of today.

Well the hype train thing, is because AMD has been talking about Polaris, nV the only thing they have sated so far is about p100, and that has little to with gp104 outside of architecture.
Click to expand...


It will use GDDR5, not GDDR5x.

Maybe there is something to the new architecture that will allow it to be above an r390 even with a smaller bus size?
 
I can't wait to see these cards come out. Because until then this is all unfounded speculation and I can't buy unfounded speculation.
 
Zion Halcyon said:
It will use GDDR5, not GDDR5x.

Maybe there is something to the new architecture that will allow it to be above an r390 even with a smaller bus size?
Click to expand...


hard to say, compression techniques and better use of the bandwidth is pretty much the only thing to overcome the bandwidth deficit. Even with the fastest GDDR5 it still puts a 256bit bus less than the bandwidth of a r290x, quite a bit less, around ~30%, will those two thing over come that and capable of more?
 
razor1 said:
hard to say, compression techniques and better use of the bandwidth is pretty much the only thing to overcome the bandwidth deficit. Even with the fastest GDDR5 it still puts a 256bit bus less than the bandwidth of a r290x, quite a bit less, around ~30%, will those two thing over come that and capable of more?
Click to expand...

As you said, really hard to say.

But it would be a really impressive feat of engineering if they did pull it off. And it would also explain how the price could be cheaper.
 
How can they keep it on GDDR5 and not 5X? I thought with the collapse of HBM2 stocks, both companies best gaming cards this year would bne GDDR5X powered? We should at least get a little bit of an upgrade here. Okay, its great that You can get the power of a 980 Ti today costing $500-600 for 300-400 (and I'd love to see the 490X sub $350, but if they're saying the rumor is the 300 range, than I expect 399.99, sadly), that's great for the mid-ish range, but what will that mean for 4K gaming and VR until the next group of cards? Likewise, won't Nvidia expect their either 980 or more likely 970 replacements have a price and performance in the same range as what AMD just released...or perhaps higher? The rumor for Nvidia is they are using GDDR5X, so while this AMD rumor looks like a big win in terms of price/performance in the current market, but I worry that once both releases are launched this year it won't be impressive for performance or for price!

I really hope this isn't the case and AMD puts GDDR5X on at least the highest end line of cards released this year. I want to see AMD pick up a win, but depending entirely on price isn't going to cut it without going insanely low (ie 490X for $250 and everything else lower)
 
The GTX 1070 is confirmed to use regular GDDR5, the 1080 is confirmed to use GDDR5X so it is very unlikely any Polaris cards will use it.
Which makes AMD's choice to go with the 256-bit bus even more odd, and also contradicts the "near 980 Ti" rumors since you don't put a weak bus on a card like that. They're also sporting 8 GB.

As to why GDDR5X is missing, maybe too expensive? Or low availability. The 1080 might not even be launching in June.

Frankly I don't care about GDDR5X or HBM. Most people don't. These aren't 4K cards, hell they're potentially not even 1440p cards.
 
When is Polaris rumored to by buy able? I figure they be behind nvidia launch as usual, but how bad? Pascal is supposed to hit some time in june right? So how long for Polaris?
 
Should be around the same time. Doesn't take long once they have working silicon in hand to get them to mass production.
 
primetime said:
When is Polaris rumored to by buy able? I figure they be behind nvidia launch as usual, but how bad? Pascal is supposed to hit some time in june right? So how long for Polaris?
Click to expand...

The Announcements for both Pascal and Polaris are expected to be at Computex this year, which is end of May to the first day of June.
 
razor1 said:
hard to say, compression techniques and better use of the bandwidth is pretty much the only thing to overcome the bandwidth deficit. Even with the fastest GDDR5 it still puts a 256bit bus less than the bandwidth of a r290x, quite a bit less, around ~30%, will those two thing over come that and capable of more?
Click to expand...

Right, Tonga/Fiji has Delta Color Compression, but it's not as aggressive as Nvidia's version (that's why the 380 has to have twice the memory bus as the 960 to compete). That leaves some holes for improvement.

And Hawaii had zero frame buffer compression, so there's a lot of room for it's replacement to improve.

Reality: Nvidia was able to make a card that competed fairly well with the 390X with only slightly faster clocked 256-bit ram. If AMD can bump up the speed to 8Gbps, and introduce the same level of compression as Maxwell, they could do the same exact thing, or slightly faster! Add in a little more L2 cache over Maxwell, and you can make up the bandwidth difference enough to get within 10% of the 980 Ti!

Of course, they would lose their advantage of scaling well at higher resolutions, but that's what you lose when you give up that extra memory bandwidth. Would be an amazing card at $330 though :D
 
It's always possible they added a single stack of HBM1 with GDDR5 and use it like EDRAM. Use it for the framebuffer and a L3 cache alongside a bunch of slow GDDR5. Not to different from the rumored Zen setups. Really only need the GDDR5 to bypass the PCIE bus as a limitation and it wouldn't have to be that fast. Bit convoluted, but would get them another 128GB/s of bandwidth, an extra gig of VRAM, and let them make cutbacks to the GDDR5.
 
Anarchist4000 said:
It's always possible they added a single stack of HBM1 with GDDR5 and use it like EDRAM. Use it for the framebuffer and a L3 cache alongside a bunch of slow GDDR5. Not to different from the rumored Zen setups. Really only need the GDDR5 to bypass the PCIE bus as a limitation and it wouldn't have to be that fast. Bit convoluted, but would get them another 128GB/s of bandwidth, an extra gig of VRAM, and let them make cutbacks to the GDDR5.
Click to expand...

So you think they'd eat the cost of mounting HBM1 (not cheap, even with a single stack) PLUS GDDR5 in the same system, and building a memory controller than can handle both at the same time, rather than just going 384-bit wide GDDR5? Cause I think you should share the crack you're smoking :D

Plus, I already came up with a REAL way they could get that level of performance with low cost. It's a lot easier to reverse-engineer your competitor's more advanced compression algorithm, rather than developing it on your own.
 
You must log in or register to reply here.
Back
Top