Because no one buys a PC to play 2 games only ever? Obviously the Vega is going to be able to sustain higher framerates in more demanding current and future games.
Because no one buys a PC to play 2 games only ever? Obviously the Vega is going to be able to sustain higher framerates in more demanding current and future games.So, why would you spend "$500-$700" (or more?) on Vega when you can spend $250-$300 on a RX580 if you are only going to play BF1 & Doom?
Freesync would disguise any dips in performance.
Yes, so doing the blind test on only 2 games "where you can't tell the difference" makes no sense.Because no one buys a PC to play 2 games only ever? Obviously the Vega is going to be able to sustain higher framerates in more demanding current and future games.
Yep agreed. It's all marketing fluff.Yes, so doing the blind test on only 2 games "where you can't tell the difference" makes no sense.
If they would have allowed a much broader range of games to be 'blind' tested, then, they would have had a more truthful picture of things, not that gimmick line they said.
Yep agreed. It's all marketing fluff.
Of course it won't be since they specifically state in the AMD blog article that the difference was $300. Which places the standard RX Vega at $499.Somehow, I don't think that will be the case.
http://www.pcgamer.com/doom-benchmarks-return-vulkan-vs-opengl/2/
DOOM 1440p Ultra, Vulkan. 5930k @4.3Ghz and GTX 1080 - Average 136.1fps
So we can presume the GTX 1080 was probably sitting at the 100Hz cap the entire time. Question is now, was Vega.
I don't like this.
https://gaming.radeon.com/en-us/vega-freesync-budapest/?sf101934658=1
"Even if we pitted Radeon RX Vega against the mightier GeForce 1080Ti, my money would still bet on a similar outcome: that you wouldn’t really be able to tell the difference with variable refresh rate."
So AMD are saying the 1080Ti is mightier then.
I agree, I wonder how far ahead Nvidia is right now.
At the time of this post:
Vega consumes 400W to achieve 1.6Ghz clock.
At 1.6Ghz clock, it is still slower than a reference gtx 1080, which consumes 200W.
15 months late compared to a gtx 1080.
The only saving grace for Vega at this point is price, which Nvidia adjust accordingly at any time.
Raja please go.
Someone on HardOCP forums pointed out that the real-world bandwidth of Vega FE is probably limited by the Synopsys memory controller they are using for HBM. Testing from PCGH.de provided only a bit over 300 GB/sec of bandwidth in testing, despite the fact that the stacks should be able to put through 480 GB/sec at the clock rate they are run at. The Synopsys HBM controller supports only up to 307 GB/sec bandwidth.
This may be part of the explanation for why Vega underperforms so badly in gaming. They were originally going for 512 GB/sec memory bandwidth, but ended up with less than 60% of that. Whatever simulations and other stuff they ran assumed nearly twice as much memory bandwidth as they actually got. Hynix/Micron/Samsung/whoever they're getting HBM2 from underperformed, and apparently Synopsys, from whom they're licensing their memory controller, underperformed as well (or they didn't read the specs carefully enough?) This is one of the drawbacks of running an operation on a small R&D budget: you have to rely on third-party vendors and can't always control the quality of their work.
Someone on HardOCP forums pointed out that the real-world bandwidth of Vega FE is probably limited by the Synopsys memory controller they are using for HBM. Testing from PCGH.de provided only a bit over 300 GB/sec of bandwidth in testing, despite the fact that the stacks should be able to put through 480 GB/sec at the clock rate they are run at. The Synopsys HBM controller supports only up to 307 GB/sec bandwidth.
They are likely talking about aggregate bandwidth per controller. etc. Which would be half a stack if they are using two controllers per HBM stack like they did with Fiji. or it could be a single controller solution per stack this time.
AMD has always struggled with memory controllers
From the article I linked above...
With this logic...
They could have pitted a Radeon RX Vega vs the Radeon RX 580, my money would still bet on a similar outcome: that you wouldn’t really be able to tell the difference with variable refresh rate.
So, why would you spend "$500-$700" (or more?) on Vega when you can spend $250-$300 on a RX580 if you are only going to play BF1 & Doom?
This is why this whole "blind" test is a load of hooey.
Wait for Volta is definitely something I wouldn't blame any person for doing right now. This is 1 year late.All this. Volta is going to be completely unanswered. AMD has really been churning out turds the past 2 years. I rocked a 7950 and 7970 along with a 390 and all were terrific cards, especially the 7000 series so I'd love to see AMD catch up with NVIDIA but it seems like the gap just keeps getting wider and wider. Hopefully the HBM gamble pays off one day because we're 2 years in and it's still a turd in a punch bowl while NVIDIA's GDDR5 and GDDR5X cards crush both clock speeds and power consumption.
I could of swore AMD was using Rambus tech for their memory controller for HBM2, but, yeah, according to http://www.nasdaq.com/press-release...gbs-bandwidth-for-graphics-and-20170725-00873 Synopsys got the win....Whatever simulations and other stuff they ran assumed nearly twice as much memory bandwidth as they actually got. Hynix/Micron/Samsung/whoever they're getting HBM2 from underperformed, and apparently Synopsys, from whom they're licensing their memory controller, underperformed as well (or they didn't read the specs carefully enough?) This is one of the drawbacks of running an operation on a small R&D budget: you have to rely on third-party vendors and can't always control the quality of their work.
That IS my point.You know this isn't true so I don't even know why you bother with it. There's a reason we want the fastest GPU possible to sync up with Freeysnc.
I just hope the "Launch" at Siggraph, puts an end to nonsense blind testing, and it is not some kind of paper launch and only a few outlets get test cards on condition that only have them testing doom, in the dark...
AMD is waiving their hands so furiously, trying to show everyone that they can give a similar game play experience to a GTX 1080. This all means that, generally speaking, GTX 1080 is going to be faster. Anyone intent on buying a 300w card for the same price as an 180w equivalent (or better) that has been out for 13+ months, that is some serious dedication to reward piss poor engineering.
Volta is going to drop in the December '17 - March '18 time frame with a GTX 2070 shaping up to be 15-20% faster for half the power consumption. If you waited 13+ months for Vega to reward you with nothing, why not wait another 6 months for Volta to actually deliver?
The marketing team is also viral so I wouldn't really say that the marketing is "working" on quite a few people here...I would say to you that a lot of these people are actually AMD viral.What is surprising to me is how AMD's marketing seems to be working on quite a few people here... Guess the marketing team isn't as bad as some say.
The marketing team is also viral so I wouldn't really say that the marketing is "working" on quite a few people here...I would say to you that a lot of these people are actually AMD viral.
Whoever wrote that has probably reached a wrong conclusion.Someone on HardOCP forums pointed out that the real-world bandwidth of Vega FE is probably limited by the Synopsys memory controller they are using for HBM. Testing from PCGH.de provided only a bit over 300 GB/sec of bandwidth in testing, despite the fact that the stacks should be able to put through 480 GB/sec at the clock rate they are run at. The Synopsys HBM controller supports only up to 307 GB/sec bandwidth.
This may be part of the explanation for why Vega underperforms so badly in gaming. They were originally going for 512 GB/sec memory bandwidth, but ended up with less than 60% of that. Whatever simulations and other stuff they ran assumed nearly twice as much memory bandwidth as they actually got. Hynix/Micron/Samsung/whoever they're getting HBM2 from underperformed, and apparently Synopsys, from whom they're licensing their memory controller, underperformed as well (or they didn't read the specs carefully enough?) This is one of the drawbacks of running an operation on a small R&D budget: you have to rely on third-party vendors and can't always control the quality of their work.