[Rumor] RX 480 Overclocking 1500+Mhz

[maddie]

Didn't you know that $449 RX 470 and $699 480 were designed as 1070/1080 competitors but testing them 6 months ago at 1.7Ghz revealed 250W power usage? As the last resort, AMD had to drop clocks to low and mid-1200mhz range, reduce MSRP to $149-179 and $199-229. Then they took their 2018 next generation ultra high-end cards and pushed them forward to 2017 as a sign of desperation. They will now use a 600mm2 VEGA HBM2 just to compete with a 1080. Navi is AMD's dual-Vega card with a 280mm2 radiator aimed to compete with Big Pascal but will arrive 1 year later than scheduled due to lack of funds and sufficient amount of low-leakage VEGA chips. VEGA will also be a 2-3 months paper launch in 2017 because HBM2 has supply shortages.

^ I swear I heard this from reliable sources. You don't have to believe me but I am just looking out after you guys. 🙂

Thank you my lord. In thee I trust.

 

[frozentundra123456]

That style of marketing worked great for Apple. You don't need to be technically savvy or know about rams and processing any units, "It just works." Buy it and use it, never question it.

However, is there actual evidence of Kyle/Hard being paid to bash AMD and promote Nvidia? Poor reviews and bias are one thing, but some of the allegations being tossed around here make me uncomfortable.

Well said. One can disagree with reviews or point out that they dont agree with other sites, but this namecalling, personal bashing and accusations of being paid off is a bit, well I dont really know how to describe it without getting an infraction, but there really is no place for it.

 

[Slaughterem]

I will take bets that Kyle will compare 1080 Pascal arch with 480 Polaris arch for performance/ watt without mentioning the 1080 receiving a ~15 watt benefit from using gddr5x memory compared to gddr5.

 

[digitaldurandal]

I will take bets that Kyle will compare 1080 Pascal arch with 480 Polaris arch for performance/ watt without mentioning the 1080 receiving a ~15 watt benefit from using gddr5x memory compared to gddr5.

If you are comparing perf/w of two specific cards it doesn't matter how you acheived it.

 

[selni]

But RS, Kyle has "been told" AMD's 14nm process has a 30% advantage over Nvidia's 16nm.

I may have some dissatisfaction with Polaris, but at least I'm not delusional like Kyle who expected outright 1:1 competition with GP104. Also I'm looking forward to in what way RX 480 is hotter than GP104 as claimed by him, since it clearly has lower power consumption and leaks show lower temperatures too. If it can match or beat performance-per-watt then Kyle further loses any sense of credibility. Sadly even a minor ppw advantage for Nvidia I'm sure he would spin into validation of his claims.

I'm not 100% convinced AMD planned on the gap we're expecting between Polaris and GP104 - that potentially catching them a little off guard sounded like one of the more believable parts of that rant. The whole "the die size is smaller it was obviously never intended as competition" argument was nonsense, since when has die size ever been the only factor - particularly when you're not even comparing the same process?

Eg Hawaii was 438mm2 and was at the very least strongly competitive with the much larger GK110 (which yes was a compute die but still) and definitely intended as competition for it. Yeah Polaris was always a mainstream part, but this isn't GP100/102 either.

 

[Slaughterem]

If you are comparing perf/w of two specific cards it doesn't matter how you acheived it.

If you are comparing arch to arch then it is

 

[guachi]

Didn't you know that $449 RX 470 and $699 480 were designed as 1070/1080 competitors but testing them 6 months ago at 1.7Ghz revealed 250W power usage? As the last resort, AMD had to drop clocks to low and mid-1200mhz range, reduce MSRP to $149-179 and $199-229. Then they took their 2018 next generation ultra high-end cards and pushed them forward to 2017 as a sign of desperation. They will now use a 600mm2 VEGA HBM2 just to compete with a 1080. Navi is AMD's dual-Vega card with a 280mm2 radiator aimed to compete with Big Pascal but will arrive 1 year later than scheduled due to lack of funds and sufficient amount of low-leakage VEGA chips. VEGA will also be a 2-3 months paper launch in 2017 because HBM2 has supply shortages.

^ I swear I heard this from reliable sources. You don't have to believe me but I am just looking out after you guys. 🙂

A very good parody of Kyle at HardOCP.
Well played, sir. Well played. :thumbsup:

 

[crisium]

I'm not 100% convinced AMD planned on the gap we're expecting between Polaris and GP104 - that potentially catching them a little off guard sounded like one of the more believable parts of that rant. The whole "the die size is smaller it was obviously never intended as competition" argument was nonsense, since when has die size ever been the only factor - particularly when you're not even comparing the same process?

Eg Hawaii was 438mm2 and was at the very least strongly competitive with the much larger GK110 (which yes was a compute die but still) and definitely intended as competition for it. Yeah Polaris was always a mainstream part, but this isn't GP100/102 either.

I think the gap may be larger than they intended for too (going off of "leaks" of slightly slower than 390X). But I think the best hope of AMD was Pitcairn vs GK104. Smaller chip, not designed to compete, but not outrageously slower. It looks more likely with each leak that, at least at stock clocks, 480 vs 1080 will be less favorable for AMD than 7870 vs 680 was.

 

[Mercennarius]

I think the gap may be larger than they intended for too (going off of "leaks" of slightly slower than 390X). But I think the best hope of AMD was Pitcairn vs GK104. Smaller chip, not designed to compete, but not outrageously slower. It looks more likely with each leak that, at least at stock clocks, 480 vs 1080 will be less favorable for AMD than 7870 vs 680 was.

Why would anyone cross shop a single 480 vs a 1080?? That's like cross shopping a new Corvette Z06 and a Mustang GT. If there is anyone seriously cross shopping these cards they are doing so with two 480s in crossfire vs one 1080.

 

[zinfamous]

lol

Sorry, that is not true.

...are you still using those cheap Eveready batteries that this thing came with?

 

[crisium]

Why would anyone cross shop a single 480 vs a 1080?? That's like cross shopping a new Corvette Z06 and a Mustang GT. If there is anyone seriously cross shopping these cards they are doing so with two 480s in crossfire vs one 1080.

Hmm? That's not at all what I am talking about. I'm just trying to compare the performance gaps to what we saw with the last node jump (GK104/GP104 and Pitcairn/P10... obviously small Vega competes with GP104 just like Tahiti competed with GK104). Naturally the price gaps will be bigger this time around since the performance gap is larger too, and no one cross shopped 7870 and 680 back then either, so why would you think I am now?

 

[Mercennarius]

...are you still using those cheap Eveready batteries that this thing came with?

LoL

I realized the sarcasm after my post.

 

[Mercennarius]

Hmm? That's not at all what I am talking about. I'm just trying to compare the performance gaps to what we saw with the last node jump (GK104/GP104 and Pitcairn/P10... obviously small Vega competes with GP104 just like Tahiti competed with GK104). Naturally the price gaps will be bigger this time around since the performance gap is larger too, and non one cross hoped 7870 and 680 back then either.

I think AMD is thinking the price gap is more important this time around when it comes to actual GPU sales. The 1080 is significantly more powerful but for your average GPU buyer the 480 will likely seem more attractive.

390 level performance with very low power consumption for $200 starting price is going to grab a huge chunk of buyers.

 

[frozentundra123456]

If you are comparing arch to arch then it is

So what? What matters is the final result, not how it is achieved.

 

[krumme]

So what? What matters is the final result, not how it is achieved.

I think you agree and have a misunderstanding here. Fiji isnt as efficient as maxwell just because nano is.
He was referring to the arch not card.

 

[krumme]

Obviously there is a cost vs eff issue here. Standard ram and smaller die will naturally mean less efficiency.
We will see with vega but i think the issue is just the be sufficient efficient for the market. It doesnt really matter much if 480 uses 100 or 150w for desktop but 200 vs 300 usd is a major difference for the potential sale.

 

[Slaughterem]

I think you agree and have a misunderstanding here. Fiji isnt as efficient as maxwell just because nano is.
He was referring to the arch not card.

:thumbsup:

 

[Qwertilot]

I think AMD is thinking the price gap is more important this time around when it comes to actual GPU sales. The 1080 is significantly more powerful but for your average GPU buyer the 480 will likely seem more attractive.

Given how much it would have helped their market position (now and with Vega), I can believe they were originally aiming for say 10% more - just enough to tread very firmly on a stock 1070's toes - but didn't get it.

It is entirely impossible to tell of course 🙂

390 level performance with very low power consumption for $200 starting price is going to grab a huge chunk of buyers.

I do wonder slightly about this. New sales until the 1060 etc comes out yes.

Upgrade sales? Low/mid range upgrade cycles tend to be reasonably slow.

From a 390/(X) you'll need to hit someone who's into the massive efficiency upgrade, and they're probably not using a 390/(X) to start with! No chance from a 970/80 of course and there's a lot of those.

Amid all of this its well worth remembering that the R&D from Polaris has already worked quite nicely for them with the new console deals. Any decent amount of desktop sales is basically gravy.

 

[kraatus77]

Amd could have used 14ff advantage to the performance side, but since they were behind maxwell in p/w and got ridiculed for being power hungry. they used most of the advantages for p/w instead.

that's why pascal has only 1.6x more p/w but polaris is 2.8x more p/w than their previous archs. but as always goalposts have changed again.

 

[krumme]

Given how much it would have helped their market position (now and with Vega), I can believe they were originally aiming for say 10% more - just enough to tread very firmly on a stock 1070's toes - but didn't get it.

It is entirely impossible to tell of course 🙂



I do wonder slightly about this. New sales until the 1060 etc comes out yes.

Upgrade sales? Low/mid range upgrade cycles tend to be reasonably slow.

From a 390/(X) you'll need to hit someone who's into the massive efficiency upgrade, and they're probably not using a 390/(X) to start with! No chance from a 970/80 of course and there's a lot of those.

Amid all of this its well worth remembering that the R&D from Polaris has already worked quite nicely for them with the new console deals. Any decent amount of desktop sales is basically gravy.

85% of the desktop market is below 300 usd.
Now add mobile apple consoles. As p10 p11 is geared towards that.
Actually i dont understand why they are making vega. At the end of the life cycle i would be surprised if it ammounted to more than like what 5% of the sales? Add this expensive hbm2 and what is the purpose?. Nv will always have some fat ass faster card so its its imo crazy fight to go into as at that end nothing but perf counts.
The bs about amd aiming for 1070 whatever is nonsense. They wouldnt know how nv would price their cards anyway. Mainstream cards is 200 to 250 tops and not 300mm2 and with fancy ram.

 

[Qwertilot]

Are sure if that figure is installs or actual annual sales? You'd expect quite a difference, because the top end cards get replaced/upgraded a lot faster.

The top end of gaming is also definitely very profitable/unit sale - look at Titans(!).

As for why Vega? Well, with big Vega there's the attraction of competing in the potentially really profitable compute/workstation etc markets. Intel there as well as NV now of course, but its a very big pie so maybe they can carve a niche.

Small vega does seem like it might suffer from existential angst - pitched bang against a 1070/80 who have been out on sale for 6(likely +) months? Hard work!

 

[antihelten]

It looks more likely with each leak that, at least at stock clocks, 480 vs 1080 will be less favorable for AMD than 7870 vs 680 was.

To be fair it's not really reasonable to expect that small of a gap between the 480 and the 1080 anyway, seeing as GK104 is only about 5% faster than Pit Cairn these days (770 vs. 270X):

Now why GK104 is so bad, or inversely why Pit Cairn is so good is anyone's guess, but it would probably be unrealistic to expect a repeat.

 

[crisium]

To be fair it's not really reasonable to expect that small of a gap between the 480 and the 1080 anyway, seeing as GK104 is only about 5% faster than Pit Cairn these days (770 vs. 270X):

https://tpucdn.com/reviews/ASUS/GTX_1080_STRIX/images/perfrel_1920_1080.png

Now why GK104 is so bad, or inversely why Pit Cairn is so good is anyone's guess, but it would probably be unrealistic to expect a repeat.

Well, even at the launch period before AMD's optimization drivers became a thing of legend there was a gap that still seems closer than 480 vs 1080 may be.

https://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_680/27.html

7870 is 77% of a 680 at 1920x1200, and 75% of a 680 at 2560x1600.

https://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_1080/26.html

390X at 60% of a 1080 at 1920x1080, and 61% of a 1080 at 2560x1440 (just a filler for possible 480 if its close to this)

At the risk of performance-per-mm2 dismissers eating me alive again, much like other comparions, there is a die size parallel here. GK104 to Pitcairn is very nearly the same as GP104 vs Pascal 10. Obviously Maxwell was quite the stride forward, but it would have been nice to see AMD "reset" back to early 2012 where this gap wasn't so large.

 

[AnandThenMan]

I mean if the main editors/owner(s) cannot comprehend that P10 is an HD7850 successor and is not a 660Ti/670/680 (1070/1080) competitor, then what else is there to discuss? AMD told AnandTech in the spring that they will have Vega 10 and 11 as higher end cards but this flew over their heads too.

Kyle knows exactly what Polaris is all about and the market it is aiming for. He is purposely being disingenuous and claiming AMD aimed for a halo part and fell short even though the slightest bit of logic destroys that notion. To dignify what Kyle is saying we'd have to believe that AMD thought they could make a high end part using a tiny die with a 256bit GDDR5 memory bus this would require incredible breakthroughs in architecture. D:

The vastly saner suggestion is that he's publishing semi rants - against everybody note - to get attention and drive traffic.

Nailed it. Kyle is basically trolling his own forum to drive traffic and it works.

Anyway I'm wondering if AMD should move the NDA forward there have been so many leaks that by the time official reviews hit we'll already know most everything.

 

[Headfoot]

Why would anyone cross shop a single 480 vs a 1080?? That's like cross shopping a new Corvette Z06 and a Mustang GT. If there is anyone seriously cross shopping these cards they are doing so with two 480s in crossfire vs one 1080.

I'm cross shopping 480 and 1080. I want to move to single card. Don't forget that 480 @ $200 w/ roughly 390x performance will be a great stopgap card until the actual high end 16nm cards drop with Vega and GP102. Probably not many buyers know enough to take this approach, but anyone on this forum could. It would let me sell out my 290s while miners are still buying them up for a good price but keep similar single card performance for not a ton of cash.

But at the same time I'm considering 1080/1070 because it would boil my 290CF performance into a single consistent card. The problem is how pricey it would be, and i'll certainly be tempted to upgrade when big 16nm hits.