(Juvenile, off topic social media post from 2016)
A) its not off topic. Its exactly about VegaGrow up.
A) its not off topic. Its exactly about Vega
B) I stand by what I said and most people agree that AMD is clueless in marketing
Their countless teasers and venues for Vega will now come bite them. A lot less people will wait for any other AMD GPU whenever AMD tease something in the future and I hope they are aware of that. A lot fewer people will trust them and their promises
A) its not off topic. Its exactly about Vega
B) I stand by what I said and most people agree that AMD is clueless in marketing
Their countless teasers and venues for Vega will now come bite them. A lot less people will wait for any other AMD GPU whenever AMD tease something in the future and I hope they are aware of that. A lot fewer people will trust them and their promises
Are you kidding?Where did you get your marketing degree? What would you do different?
Countless? You can count the number of showings on 1 hand.
A lot less blah blah blah blah waiting. What forecasting data are you using?
A lot fewer blah blah blah promises. What promises have been broken?
Are you kidding?
They were behind this...,
So, they made expectations so high, that Vega could even beat Volta, and you don't see that?
News flash! Intel and Nvidia products can be dramatically undervolted at stock configurations too. This phenomenon is not exclusive to AMD, it's merely become a pretend benefit / talking point since most AMD products have no headroom and their graphics cards suffer in efficiency.
No one thinks its a pretend benefit. It's obviously a negative aspect of RTG's launch/binning process. It is a factual statement that RTG frequently launches over volted cards, and I don't see anyone here praising this.
IMO, if they can match 1080 performance they are better off matching 1080 at $500 price to sell it to loyalists and Freesync users. Gotta get whatever return you can. The good news is that if they can match 1080, then the cut down Vega should have no problem beating 1070 since AMD will cut the shaders 12.5% vs Nvidia's 25% cut. That one I would expect closer to $400, as even if is a power hog it can sell to some users if it sits above the 1070 consistently in benchmark charts.
Despite negativity, these can move units to the right customers. But they really need to at least match the 1080 performance to stand a chance. If 56NCU Vega cannot beat the 1070 I don't know what AMD can do that's profitable here.
Over volting is just as successful on Intel and Nvidia, it just so happens their products tend to have headroom so users opt to increase performance instead of dropping voltage.
This is one of the reasons why AMD is the worst hype marketing clueless company to ever exist. August 2016.
Like it was so amazing. Little did people know it would take a freaking year to get the product out
What the heck is the point with these stupid teasers looking back now?
Grow up.
A) its not off topic. Its exactly about Vega
B) I stand by what I said and most people agree that AMD is clueless in marketing
Their countless teasers and venues for Vega will now come bite them. A lot less people will wait for any other AMD GPU whenever AMD tease something in the future and I hope they are aware of that. A lot fewer people will trust them and their promises
Your post was like Trash WCCFTECH .that's why he said Grow up.Don't bring those things.
Where did you get your marketing degree? What would you do different?
Countless? You can count the number of showings on 1 hand.
A lot less blah blah blah blah waiting. What forecasting data are you using?
A lot fewer blah blah blah promises. What promises have been broken?
Are you kidding?
They were behind this...,
So, they made expectations so high, that Vega could even beat Volta, and you don't see that?
*edit
You know, I now know where the GPU marketing team came from.
It was the same people that marketed bulldozer.
They famously once said, IPC will be higher in these forums.
Well AMD was right about poor Volta. We are all going to be poor after what Nvidia will be able to charge for it with no competition
This reminds me that after Nvidia's last earnings call in May, a reporter asked Jensen if he was concerned about Vega. He said
"My assessment is that the competitive position is not going to change."
https://www.overclock3d.net/news/gpu_displays/nvidia_is_not_worried_about_amd_s_vega_architecture/1
What is Volta's performance? What is RX Vega's performance? Do you think marketing is literal? Even given you can see the paragraph below talking about voltage. It's just some snark. You can keep telling stories with Cloud.
Starting to think the "Poor Volta" with a Radeon sticker was just foreshadowing of the poor voltage settings future AMD GPUs will launch with. We saw it with Polaris, Polaris duex, and now most likely Vega.
Poor voltage indeed.
EDIT: I feel the need to point out this is a joke. Jimmies be rustling around here lately.
https://forum.beyond3d.com/posts/1973875/The Vega conversation become too heated for me. I know that people are very excited about this card. I assumed there would be lots of reviews of Vega FE at launch, but everybody started to ask me to do a review instead. I even got requests from press to come and benchmark to our office. There's no conspiracy or anything. I am just letting AMD and the press to do their work, while focusing on my own.
I am currently finalizing our tech and I have my hands full of tasks. I skipped E3 and SIGGRAPH this year to have more time to get things done. Our launch is getting closer and shipping to 5 platforms (base & upgraded consoles + PC) with a 3 people team is pretty hard.
Vega is working well and I use it daily. Async compute is also working fine. I had plans to test some fp16 code on it, but I found out that Unreal Engine doesn't yet support fp16 on PC. So I am waiting for Epic to add fp16 support before I continue with fp16 work on PC. Ran my buffer benchmark on Vega (https://github.com/sebbbi/perftest), and noticed no surprises. It behaves like any other GCN based GPU (same peak perf/clock on various L1$ buffer load operations). Our game is mostly ray-traced, so it has no very little overdraw. Tiled rasterizers shouldn't help much at all, so I haven't spend any time confirming gains on that regard. I will test DX 12.1 features (especially volume tiled resources) after our game launches.
The card is very quiet. Even when playing our game at 4K without vsync. Running it at stock settings (like our CPUs and other components). In gamedev hardware stability is always more important than slight perf increase. I'd assume AMD has plenty of headroom to run the fan faster in consumer release if they want to.
Just wanted to clarify that I meant AMD GCN2 (consoles) vs Nvidia's latest (Maxwell/Pascal). AMD PC GPUs have also improved since GCN2.
Improvements for general performance:
- GCN3 introduced delta color compression. Including ability to sample/load compressed textures without decompress step.
- GCN3 improved geometry tessellation performance
- GCN4 improved geometry performance in general (including fast strips, primitive discard, etc).
- GCN4 improved delta color compression.
- GCN4 added instruction prefetch (reduces pipeline latency, again helps with geom bottleneck).
- GCN4 improved async compute scheduling (GPU side)
GCN5 (Vega) adds these general performance improvements:
- L2 cache includes L2 ROP cache (L1 ROP caches under L2). Don't need to flush caches between pixel shader passes.
- Tiled rasterizer. Reduces overdraw, bandwidth and makes ROPs more efficient in general.
- Improved geometry pipeline (including proper load balancing, up to 2x higher peak throughput)
- General purpose memory paging system
(I didn't list features that don't bring performance improvements without programmer intervention)
All of these improvements mean that GCN5 should run general purpose pixel/vertex shader code much better than GCN2. GCN5 has most of the same tricks that are seen in modern Nvidia GPUs. There are nice compute improvements as well, but they need special programmer support (DPP, SDWA, FP16). We will see the real impact of these improvements when DX12 SM 6.0 becomes available. Doom is already using these features with Vulkan, resulting in nice gains.
Im using puffin web browser and opera max data compression on my phone, both awesome for speed and data savings but unfortunately sometimes the browser disconnects from the cloud and resets the page whilst writing, I had written a long post and somehow it reset and posted with a bit of writing, which was weird so I deleted the single line and couldn't be bothered to re type.Mind sharing that time travel design. I see you edited before you posted.
Actually, I think it IS reflected in reality.https://forum.beyond3d.com/posts/1992079/
https://forum.beyond3d.com/posts/1973875/
GCN5 (Vega) adds these general performance improvements:
- L2 cache includes L2 ROP cache (L1 ROP caches under L2). Don't need to flush caches between pixel shader passes.
- Tiled rasterizer. Reduces overdraw, bandwidth and makes ROPs more efficient in general.
- Improved geometry pipeline (including proper load balancing, up to 2x higher peak throughput)
- General purpose memory paging system
(I didn't list features that don't bring performance improvements without programmer intervention)
Why what is posted in second post is not reflected in reality?
Its not that AMD is clueless company. The hardware is there. So what is happening?
https://forum.beyond3d.com/posts/1992079/
https://forum.beyond3d.com/posts/1973875/
Two posts from the same person, that is very knowledgable about GPUs from different parts of time.
Why what is posted in second post is not reflected in reality?
Its not that AMD is clueless company. The hardware is there. So what is happening?
Most of the major titles are both Xbox and Playstation.I'm starting to seriously suspect the failure of MSFT's Xbox One catching dominance is what is hurting AMD more so right now. PS4 I believe uses a different SDK that doesn't (from my understanding) translate as well to direct DX12 implementation.
It really comes down to something being an AMD or Nvidia based feature and how they push(or don't) adoption. Nvidia is much more aggressive and consistent, while AMD is mostly passive with short campaigns but lacking persistence.We all remember the "wait for DX12" slogan that covered Fury. And when Polaris came around, it was still propagated. Now 4 years into these consoles lives and we're still seeing DX12 versions of game performing worse then their DX11 counterparts. It's only those that seem to have started in DX12 that show the most improvements. But those titles still seem to be the minority by a long shot.
But that's just my tinfoil hatting. I get the feeling AMD wasn't expecting such a slow adoption of DX12.
The tiled rasterizer only works on some workloads as I mentioned before, and, as I thought, it is transparent to the workload in question. (There is no API call for this).
Not flushing is usually a big saving, not sure why that savings isn't showing up on some workloads, perhaps AAA game engines do this internally as well, so, another feature that won't do much on modern AAA engines.
Not if it's dependent on the Draw Stream Binning Rasterizer working to achieve the mentioned throughput.For the geometry pipeline, we should be seeing good throughput here, I am guessing what is going on is, the batches are too small to make a tangible difference, and you will only see good throughput on big workloads (ie, CAD stuff with millions of polygons).
GTA V uses more than 4 GB for 1440p gaming on my system. I've had games use as much as 8 GB. Doesn't really matter much for me since my graphics card has 11 GB. HBCC is pointless on a flagship card unless 8k gaming takes off or developers start taking advantage of all that extra VRAM.For the memory paging system, I assume once again that most games are staying well within their limits of 2 or 4GB, so, again, no improvement until you go over what the VRAM the card has.
That's the reason everyone thinks that something is off. The Vega architecture overview claimed higher IPC vs Fury-X. There have been plenty of people arguing why, but I guess we'll find out in a few weeks.That means, out of all of these features that are supposed to bring Vega to the table aren't enough.
They need more ROPs & IPC.
Vega need more memory bandwidth not Rops or IPC.Btw if miners will start using vega we wont see vega in stores next 6-12months probably.
This mining craze wont crash this time.It will be here probably forever.