Poll: ** The AMD VEGA Thread **

[N19h7m4r3]

How about waiting for the official launch? Drivers clearly aren't ready for primetime yet. Without proper driver support most of the new features probably aren't working which is why we are seeing Fury X-levels of performance.

I get ya on the drivers; but this is the official launch. AMD hard launched, places in the UK actually had stock yesterday. I even tried to buy one to play with.

They also refused to Sample any reviewers with the cards. Anandtech and PcPer had to order their own.

Messy launch is messy.

 

[zornyan]

This can only be a good thing. Maybe the gaming mode driver doesn't do much/anything differently to the pro mode yet and that's why the performance isn't great. It could be that it's currently a placeholder while they actually finalise optimised gaming drivers.

pro-gaming drivers don't actually make much of a difference

Radeon pro duo, same as fury x crossfire in benchmarks.

quadro p6000, same as titan Xp in games.

at the very most switching the drivers around gains 5%, that's being very generous.

https://videocardz.com/65014/nvidia-quadro-p6000-outperforms-titan-x-pascal-in-gaming-benchmarks

here the p6000 vs the older titan XP, 7% faster, which is the exact same as the new titan Xp (both full gp102)

that's right, gaming vs pro drivers gained 0% here.

 

[Sargatanas2511]

Clearly drivers aren't optimised, i read somewhere that in one of the games tested half of the screen wasn't rendered correctly lol With proper drivers etc i expect gaming line to reach ti levels or be close to that

They demoed it in Doom about a year ago with higher performance. There has to be something wrong for it to actually drop rather than improve.

 

[CAT-THE-FIFTH]

Another AMD launch with some issue. Unless they are trying to be deceptive about the final gaming performance of the RX Vega gaming cards,I am not sure what they achieving by releasing the card in its current state outside hitting the 1H launch schedule.

 

[Lokken86]

Why wouldn't they optimise the drivers they've had long enough.

 

[Phixsator]

On the brightside folks, the dude test Frontier Edition for mining. It sucks for it. Getting 30-35 MH/s
Also he's getting a joke of a Cinebench OpenGL score, 97.39 FPS

He states no difference in performance between Pro and Gaming driver mode.

That DOOM Score WTF?! That's overclocked Fury X levels again!

Spoiler: Pics

I would take anything that is not backed up by hard evidence and video material to be fake/incorrect on arrival. Especially something relayed from one guy through another and only in text form. It stinks to high heaven.

 

[DragonQ]

Another AMD launch with some issue. Unless they are trying to be deceptive about the final gaming performance of the RX Vega gaming cards,I am not sure what they achieving by releasing the card in its current state outside hitting the 1H launch schedule.

When was the last nVidia or Intel launch without any issues? Not saying this isn't a big one or that they haven't bungled the release but this whole "AMD can't launch things without problems" is just confirmation bias. I wish they'd learn about releasing too early though. Would be better if they added on another quarter to the expected release date and possibly release it early, rather than being forced to release an unfinished product to satisfy shareholders of a met deadline.

 

[D.P.]

I think the1382MHz clock is telling typical clock speed,as AMD stated. You also have to realize Vega has less.memory bandwidth than Fiji, which is rare for a new GPU not to progress BW.
Given the typical clock speed stated by AMD the theoretical FP32 is slightly below 1080ti performance. In all recent GPUs, AmD have required significantly more more raw brute Force power to reach Nvidia's performance. And the max bandwidth is similar to 1080ti, and again AMD have recently required more raw BW due to lower compression and efficiency.

From very basic scaling given these figures you would from Fiji and Polaris, you would expect Vega to be around 1080 to slightly faster, but no where near ti.
The expectation being that architecture improvements would give the performance Delta required. There should be another 20-30% somewhere but that boost is not a simple linear response like increases jn clocks or core count.


AMD have also mentioned that they had a focus on increasing clock speed and IPC. Well,.one way to increase clocks speed is to break the most expensive t instructions into more cycles, which effectively lowers IPC but if the instructions per second increases then you get more performance. Nvidia did this with Pascal which allows their GPU to hit 2.0GHz quite easily.leadingbto a massive performance jump. If AMD did this but expected higher clocks then this would be limiting performance. Seems like and wanted Vega to run at 1600 at least bit it draws too much power and produces too much heat. Hence the water cooled Vega pulls an extra 75wbto try and keep closer to the 1600mhz.

 

[erazor51]

Well for £999 it compares well against the Quadro P5000 at least in those tasks.

Thanks for this. My mate was looking to buy the P5000 for his Maya work. Now I will recommend him to buy this instead!

 

[D.P.]

Clearly drivers aren't optimised, i read somewhere that in one of the games tested half of the screen wasn't rendered correctly lol With proper drivers etc i expect gaming line to reach ti levels or be close to that

They've had working samples for 6 months showed to the public, maybe ES for 9, and a lot of driver work can be done before receiving bfirst silicon.

If it has taken them this long to get this basic level performance then it might take years before they get close to Ti at this rate

 

[Gregster]

I really wouldn't put too much stock into that chaps testing methodology. And he is testing a workstation card. I implore you guys waiting to ignore those benches and wait for the official gaming card. No disrespect to the guy who done the tests but he clearly isn't the right guy for it and you really need to make sure that background processes are at the bare minimum, the PC is running well (I personally would never use a 550W PSU for benches but at the same point, it would either work or wouldn't, so it appears that it was suitable for it), make sure the RAM speed and timings are running efficiently and the same for the CPU.

In a nutshell, ignore his tests and wait for proper review tests or better still, wait for the gaming card reviews!

 

[five8five]

I would take anything that is not backed up by hard evidence and video material to be fake/incorrect on arrival. Especially something relayed from one guy through another and only in text form. It stinks to high heaven.

It's the data from the livestream last night in tabular form.

 

[five8five]

Clearly drivers aren't optimised, i read somewhere that in one of the games tested half of the screen wasn't rendered correctly lol With proper drivers etc i expect gaming line to reach ti levels or be close to that

Yes, but why have they released them in this state? Also without any form of disclaimer.

 

[TNA]

I really wouldn't put too much stock into that chaps testing methodology. And he is testing a workstation card. I implore you guys waiting to ignore those benches and wait for the official gaming card. No disrespect to the guy who done the tests but he clearly isn't the right guy for it and you really need to make sure that background processes are at the bare minimum, the PC is running well (I personally would never use a 550W PSU for benches but at the same point, it would either work or wouldn't, so it appears that it was suitable for it), make sure the RAM speed and timings are running efficiently and the same for the CPU.

In a nutshell, ignore his tests and wait for proper review tests or better still, wait for the gaming card reviews!

Yep, this is what I am saying also. Until RX Vega reviews come out in August with proper drivers and reviews, I will not take notice of any results.

Drivers can make a big difference in performance and price can make a big difference in whether a card is seen as a fail or not.

 

[D.P.]

They demoed it in Doom about a year ago with higher performance. There has to be something wrong for it to actually drop rather than improve.

I don't think it dropped.
Remember, according to the AMD official benchmarks the FuryX was going to be way faster the 980ti in their own benchmarks.

If AMD carefully chose specific games and specific settings that highlighted best possible performance and then someone test with other settings and performance changes, that doesn't mean the drivers got worse.

Rather it looks like performance hasnt changed much. Perhaps they found a hardware bug they had hoped to fix in drivers but couldn't.

 

[five8five]

I really wouldn't put too much stock into that chaps testing methodology. And he is testing a workstation card. I implore you guys waiting to ignore those benches and wait for the official gaming card. No disrespect to the guy who done the tests but he clearly isn't the right guy for it and you really need to make sure that background processes are at the bare minimum, the PC is running well (I personally would never use a 550W PSU for benches but at the same point, it would either work or wouldn't, so it appears that it was suitable for it), make sure the RAM speed and timings are running efficiently and the same for the CPU.

In a nutshell, ignore his tests and wait for proper review tests or better still, wait for the gaming card reviews!

There's also the guy who did the livestream last night. He had a better spec machine, Ryzen 1800X, 1200W, open case, extra fans added for cooling the GPU. N19h7m4r3 posted his results.

 

[hudson86]

As someone who actually wants a decent performing AMD card to replace my Fury ( have a freesync monitor) , what a complete and utter marketing cockup by AMD ... yet again. Regardless of how RX Vega turns out, poo sticks.

As i'm a bit lazy , i'll wait a bit longer before selling up and going green, but I bet many wouldnt.

 

[D.P.]

I would take anything that is not backed up by hard evidence and video material to be fake/incorrect on arrival. Especially something relayed from one guy through another and only in text form. It stinks to high heaven.

Lol,in denial I see.

That is 2 independent people that have found the same performance.

 

[D.P.]

I really wouldn't put too much stock into that chaps testing methodology. And he is testing a workstation card. I implore you guys waiting to ignore those benches and wait for the official gaming card. No disrespect to the guy who done the tests but he clearly isn't the right guy for it and you really need to make sure that background processes are at the bare minimum, the PC is running well (I personally would never use a 550W PSU for benches but at the same point, it would either work or wouldn't, so it appears that it was suitable for it), make sure the RAM speed and timings are running efficiently and the same for the CPU.

In a nutshell, ignore his tests and wait for proper review tests or better still, wait for the gaming card reviews!

There are 2 guys giving benchmarks,one of them is using a 1200w PSU

 

[JediFragger]

In a nutshell, ignore his tests and wait for proper review tests or better still, wait for the gaming card reviews!

PCPer and Anandtech are coming. This should get interesting...