AMD RX Vega Leaked Benchmark Shows It Ahead Of GTX 1080 – Specs Confirmed, 1630MHz Clock Speed, 8GB HBM2 & 484GB/s Of Bandwidth

[cj09beira]

Just now, ONOTech said:

At worked so blocked.

 

If you're referring to the GN video, I've already seen it. I'm still not exactly sure what you're trying to prove though. The Fury X is slower than Vega clock for clock which goes back to my original reply that the cause of this may not just be drivers.

the thing is, there are perf ipc improvements around 30-60% on pro apps, but in games its 0, it makes no sense unless the drivers aren't firing on all cylinders yet

[Drak3]

1 hour ago, Master Disaster said:

Like how Quadro is made for WS work and GeForce is made for gaming yet when Linus tested a Quadro in gaming workloads it performed exactly the same as its cheaper "gaming" equivalent?

 

Vega is Vega.

Except, in recent generations of NVidia cards, the differences between the Quadro and the GTX equivalent boiled down to ECC VRAM, the helpings of it, additional resources/controllers  (such as NV Link vs SLI) that don't impact gaming performance at all, and drivers. There's nothing that should grant GTX an advantage over a Quadro with similar or the same drivers.

 

Whereas we still don't know if the VEGA stack has hardware optimizations on card that yields the MI25 to deep learning, FE and Radeon Pro/WX to 'workstation' tasks, and RX to gaming. There's also the issue that apparently tile rasterization isn't working correctly or well on Frontier, perhaps due to drivers, perhaps to firmware, perhaps to use those resources for something else.

 

But until we have WX/Pro and/or RX land, we don't know VEGA gaming performance 100%. We don't know if there is a difference between an RX and equivalent WX besides drivers.

[Trixanity]

1 hour ago, Master Disaster said:

Like how Quadro is made for WS work and GeForce is made for gaming yet when Linus tested a Quadro in gaming workloads it performed exactly the same as its cheaper "gaming" equivalent?

 

Vega is Vega.

Did the Quadro run GeForce drivers or Quadro drivers?

[Terryv]

16 minutes ago, cj09beira said:

the thing is, there are perf ipc improvements around 30-60% on pro apps, but in games its 0, it makes no sense unless the drivers aren't firing on all cylinders yet

This is my thinking as well. It's clearly going to be much faster than the fury in gaming eventually. It can't both match the fury and utterly destroy it, it's one or the other once drivers are settled.

 

Did anyone else also notice that the under clock almost matched full speed gaming results on the FE? Clock scaling was great on pro workloads but near nonexistent on gaming loads.

[MageTank]

26 minutes ago, MageTank said:

Well. I ran my 1070 at 2050mhz, my CPU at 5ghz, and used my ram's 3200 XMP profile for an "average overclocked" run. My results: http://www.3dmark.com/3dm11/12258063

 

26,775. This is over 4k higher than the 1070 result in the OP. A massive 20% difference in graphics score, and for what? To make the Vega look better than it really is? This is absolutely sad, and yet another reason why WCCFTech should never be trusted. They create this massive hype by skewing the results intentionally, and ride the hypetrain they generate for more clicks.

 

If anyone would like me to underclock my 1070 to see exactly how crippled it needs to be in order to achieve their graphics score, I'll gladly do so. 

Well, I did it anyways. Underclocked my core clock with a -200 offset, unclocked VRAM by 500mhz , and hit 1750mhz boost under load. This is 300mhz lower than the 2050mhz I ran my original test at (which scored 26,775): http://www.3dmark.com/3dm11/12258104

23,690. 300mhz difference in clock speed, nearly a 3k difference in score (give or take whatever the memory clock speed did to the result as well). This means at 1750mhz, my graphics score was still 82 points higher than the one WCCFTech used. To put this into perspective, my EVGA 1070SC boosts to 1950mhz by itself at stock, without me downloading ANY overclocking software. They either used a card that was thermal throttling hard, or they crippled their own clock speeds and ran the test with full intent to deceive. I have NEVER seen any 1070 do 1750mhz boost by itself, unless it's a laptop 1070 (which has a higher core count by roughly 6.2%, and has the notebook denotation). I've had my hands on several 1070's, and the worst one still hit 1900mhz with relative ease, while the best hit 2228mhz. Far above 1750.

 

I don't know what the real Vega will look like, but no matter what it is, it's not fair to use crippled, unrealistic results when comparing against it's performance. It's only going to lead to massive disappointment. 

[Taf the Ghost]

21 minutes ago, Terryv said:

This is my thinking as well. It's clearly going to be much faster than the fury in gaming eventually. It can't both match the fury and utterly destroy it, it's one or the other once drivers are settled.

 

Did anyone else also notice that the under clock almost matched full speed gaming results on the FE? Clock scaling was great on pro workloads but near nonexistent on gaming loads.

A guy did some Game vs Pro mode testing, while doing OC testing. Funny thing, the Pro Mode was better in most games.  Even the Pro Mode is looking at the current Game Mode driver and thinking, "slow poke!". 

[ATFink]

7 hours ago, dexT said:

That's what they have to do or only AMD enthusiasts will buy. A little more might be OK.

Not true. If I could get gtx 1080 performance on par (since I have a freesync monitor) or cheaper than gtx 1080 price with Vega I would take that option provided I wanted to buy a new high end GPU (which I don't want to do ATM). $450 or even $500 would sell me on Vega If it launched with gtx 1080 performance (again, if I were in that market this would be the case) because I know the history of AMD driver developments.

[Terryv]

10 minutes ago, Taf the Ghost said:

A guy did some Game vs Pro mode testing, while doing OC testing. Funny thing, the Pro Mode was better in most games.  Even the Pro Mode is looking at the current Game Mode driver and thinking, "slow poke!". 

Just further evidence that their gaming drivers are bad right now.

 

I haven't seen that yet, how much faster?

[Valentyn]

3 minutes ago, Terryv said:

Just further evidence that their gaming drivers are bad right now.

 

I haven't seen that yet, how much faster?

Just a smidge, but still shows gaming drivers are rubbish.
 

 

[Taf the Ghost]

3 minutes ago, Terryv said:

Just further evidence that their gaming drivers are bad right now.

 

I haven't seen that yet, how much faster?

 

3-4%. 

 

My comments there were mostly about the DX12 + Ryzen performance relative to the 1080. There's "problems" with Ryzen + DX12 + 1080 class right now.  That's down to Nvidia's driver. If RX Vega comes in roughly 10% over the 1080, it's going to toast the 1080 Ti on Ryzen + DX12 in a few titles because of the driver issue.

[Terryv]

4 minutes ago, Valentyn said:

Just a smidge, but still shows gaming drivers are rubbish.
 

 

Those OC results were a surprise to me. A lot of gains there.

[ATFink]

4 hours ago, L.Lawliet said:

I would sell it though for VEGA!!!

400 hell what a rip off lol

Sold mine for $360, bought at $240. Effin miners... Thanks for the cash! Glad I had my old GTX 760 (4GB model) sitting around.

 

Yes I am gambling on a cryptocurrency crash and picking up a nice GPU for cheap (can't be more than a year old though). Maybe Vega if I sell some of my AMD stock? I don't know. Depends on pricing. If Vega is $400, maybe (I find this highly unlikely). Lower than that, probably. Above $450, definitely not.

[Lord Nicoll]

9 hours ago, Master Disaster said:

I just don't believe these numbers.

 

Unless someone can explain how halfing the RAM and boosting the core clock by 30Mhz can push performance so far ahead of Vega FE I'm sorry but I just don't see these numbers as being real.

 

Linus said on last weeks wan show, Vega FE & Vega RX are the exact same die, sure there might be a few small changes but essentially Vega is Vega and basic physics tells us you can't magically make something out of nothing which leaves 3 scenarios

 

1) AMD is infact Hogwarts

2) They gimped Vega FE

3) These numbers are not real

 

Which one do you think is more likely?

The FE card has a different silicon chip with a different pipeline most likely, as well as atm poor drivers (they seem to be fury drivers just modified). The RX vega cards probably have more texture units and ROPs and other gaming shit, 

[Terryv]

41 minutes ago, ATFink said:

Sold mine for $360, bought at $240. Effin miners... Thanks for the cash! Glad I had my old GTX 760 (4GB model) sitting around.

 

Yes I am gambling on a cryptocurrency crash and picking up a nice GPU for cheap (can't be more than a year old though). Maybe Vega if I sell some of my AMD stock? I don't know. Depends on pricing. If Vega is $400, maybe (I find this highly unlikely). Lower than that, probably. Above $450, definitely not.

I sold my fury x for 450$ (also what I paid for it when i got it new) and pick up a local 1060 for 180$ to hold me over.

[MageTank]

45 minutes ago, Lord Nicoll said:

The FE card has a different silicon chip with a different pipeline most likely, as well as atm poor drivers (they seem to be fury drivers just modified). The RX vega cards probably have more texture units and ROPs and other gaming shit, 

That is extremely unlikely. Unless yields are absolutely perfect, they are going to be cutting down the FE cards to make these gaming Vega cards. They are the same silicon. Whether or not @Master Disaster believes the Vega results in the OP, he certainly shouldn't believe the 1070, 1080, or 1080 Ti results in that WCCFTech post. They are all off by 20%+ at least. I already proved the 1070 result can only be obtained by underclocking a 1070 down to 1750mhz, and the 1080 result is still magnitudes slower than the GTX 1080 that is in 6000th place on 3dMark 11's GTX 1080 results. 

 

This thread is up in arms over the Vega score, but can't see that it's not even close to it's competition due to WCCFtech's cherry picked results. People are expecting around GTX 1080 or better performance, but if this result is anywhere near close to what Vega will deliver, then it won't even deliver that. 

[Valentyn]

7 minutes ago, MageTank said:

This thread is up in arms over the Vega score, but can't see that it's not even close to it's competition due to WCCFtech's cherry picked results. People are expecting around GTX 1080 or better performance, but if this result is anywhere near close to what Vega will deliver, then it won't even deliver that. 

People never learned their lesson after wccft fucked around and made people believe Polaris was going to be Fury performance on launch, and overclocked 980Ti. They faked benchmarks like 3D Mark, and said you could OC on reference cooling to 1500Mhz to beat a 980 Ti.

They need to be ignored for everything. Lately they've even started going back months to years and retroactively editing their posts so their fuckups aren't around for too long.

[MageTank]

Just now, Valentyn said:

People never learned their lesson after wccft fucked around and made people believe Polaris was going to be Fury performance on launch, and overclocked 980Ti. They faked benchmarks like 3D Mark, and said you could OC on reference cooling to 1500Mhz to beat a 980 Ti.

They need to be ignored for everything. Lately they've even started going back months to years and retroactively editing their posts so their fuckups aren't around for too long.

Sadly, they will always get away with it. People want to see things that confirm their way of thinking. Just look at the WCCFTech comments in that article. Sort them by best. You have people that genuinely believe AMD is going to beat the 1080 at both performance AND price, and by a large margin. They are disillusioned by the misinformation WCCFTech constantly spews, and are beyond redemption at this point.

 

Even in this thread, I see people that are still clinging to the notion that this card is going to be more than it can be, and setting themselves up for disappointment by clinging to that way of thinking. Yes, we don't know it's price. Yes, we don't know it's real-world performance, but we do know what it's top GPU (from the same silicon) is capable of, and extrapolation only goes down from there. Driver maturity may help, and perhaps there are hidden architectural features that may sway performance to some degree, but it's not going to be enough to topple the Titan XP or 1080 Ti, and as it currently sits, beating the 1080 is going to be a challenge in and of itself. 

[Drak3]

3 minutes ago, MageTank said:

I see people that are still clinging to the notion that this card is going to be more than it can be, and setting themselves up for disappointment by clinging to that way of thinking.

And then there's Drak3, whose long list of expectations and criteria for getting a VEGA are:

 

It's better than an RX580.

It costs less than a P5000.

 

Super hard targets to hit, I know. But damn it, I can dream.

[MageTank]

Just now, Drak3 said:

And then there's Drak3, whose long list of expectations and criteria for getting a VEGA are:

 

It's better than an RX580.

It costs less than a P5000.

 

Super hard targets to hit, I know. But damn it, I can dream.

Honestly, VEGA doesn't even need to beat the 1080 Ti. If it can offer healthy competition at the 1070/1080 market, then it's going to be great. Nvidia has gone unanswered for way too long in that price segment. The problem is, Vega is launching so late, that Volta will likely come and answer Vega at an even more competitive price. If a Volta x60 card can deliver on Pascal x70 performance for x60 prices, Vega is going to be in trouble again. This is AMD's problem. It's not their performance that will be their downfall, it's their timing and always being one step behind on the release schedule.

[Drak3]

Just now, MageTank said:

Honestly, VEGA doesn't even need to beat the 1080 Ti. If it can offer healthy competition at the 1070/1080 market, then it's going to be great. Nvidia has gone unanswered for way too long in that price segment. The problem is, Vega is launching so late, that Volta will likely come and answer Vega at an even more competitive price. If a Volta x60 card can deliver on Pascal x70 performance for x60 prices, Vega is going to be in trouble again. This is AMD's problem. It's not their performance that will be their downfall, it's their timing and always being one step behind on the release schedule.

I think it's fairly safe to say that VEGA prices will be in the right spot until VOLTA, and we'll see a seesawing in the GPU space moving forward, or AMD refines VEGA and can lower prices/rebrand/refresh at lower prices.

 

At any rate, the chances of me not getting VEGA are small. Volta, I might skip until an x80Ti.

[Jito463]

3 hours ago, Master Disaster said:

Like how Quadro is made for WS work and GeForce is made for gaming yet when Linus tested a Quadro in gaming workloads it performed exactly the same as its cheaper "gaming" equivalent?

 

Vega is Vega.

 

I didn't say software isn't important, I said I doubt driver optimisation could give that much of a boost but it what others are saying about FE being launched on Fiji drivers is true then that could explain the huge difference.

Of course, that was also a $5,000 card.  For that price, it should perform well (it's just awful in price/performance, unless you're doing professional work).  I'd wager if you used a sub-$1,000 card like the P4000 (a quick search on Google showed it as $819.61), then the results would be drastically different.

[Taf the Ghost]

36 minutes ago, Jito463 said:

Of course, that was also a $5,000 card.  For that price, it should perform well (it's just awful in price/performance, unless you're doing professional work).  I'd wager if you used a sub-$1,000 card like the P4000 (a quick search on Google showed it as $819.61), then the results would be drastically different.

A lot of people are defaulting to the "Because Nvidia does X, AMD has to do X as well!" argument, when Raja specifically said wasn't the case. Once we get the optimized drivers, though, I do expect the FE's results to pick up.  There's also the point the FE is clocked lower than the RX Vega, but that's not up for discussion either.

 

It's in the ballpark, but it's not the same as Nvidia's approach, especially if there's anything interesting going on with that ability to switch Modes so easily.

[cj09beira]

2 hours ago, Lord Nicoll said:

The FE card has a different silicon chip with a different pipeline most likely, as well as atm poor drivers (they seem to be fury drivers just modified). The RX vega cards probably have more texture units and ROPs and other gaming shit, 

at most FE is using an older stepping that has bugs on some of the new features, more texture units and rops is a no no, 

1 hour ago, Drak3 said:

I think it's fairly safe to say that VEGA prices will be in the right spot until VOLTA, and we'll see a seesawing in the GPU space moving forward, or AMD refines VEGA and can lower prices/rebrand/refresh at lower prices.

 

At any rate, the chances of me not getting VEGA are small. Volta, I might skip until an x80Ti.

we know here is another vega coming, called vega 20, it will have more memory bandwidth and fp64 on 7nm to test the waters before navi,

because vega has fp16 support maybe they can find a way to use it without dev support .

any way i will love to see what can be done using fp16, it can't be used for everything as some calculations have errors that come from the equation it self (because pcs cant use real cos, sin tan etc they have to be approximated) and other high precision tasks like calculating position of far away objects but things like pixels can be done in fp16 as you only use 8 bits for each color

[Colonel_Gerdauf]

Didn't a few people here say something to the effect of nVidia being frightened by Vega being the one and only reason the 1080TI exists at all?

What they are saying now?

 

<crickets>

 

Figured as much.

[Drak3]

2 minutes ago, Colonel_Gerdauf said:

Didn't a few people here say something to the effect of nVidia being frightened by Vega being the one and only reason the 1080TI exists at all?

To be fair, they might not exactly be wrong. Any 'fear' on Nvidia's part would now be proven a waste. In the end, we got the 1080Ti so it don't really matter.