Radeon Vega Frontier Edition LCS and 4096 Shader Processors

Published by

Click here to post a comment for Radeon Vega Frontier Edition LCS and 4096 Shader Processors on our message forum
https://forums.guru3d.com/data/avatars/m/243/243702.jpg
So, it runs at 1.42GHz. And has 2 HBM2 stacks of 8GB.
https://forums.guru3d.com/data/avatars/m/258/258664.jpg
As it looks that it does boost to 1.5Ghz I can easily imagine big Vega to boost to 1.6GHz with a proper aftermarket cooler (or anything witch liquid cooling of course).
data/avatar/default/avatar29.webp
So, it runs at 1.42GHz. And has 2 HBM2 stacks of 8GB.
13 tflops it has to run 1550 or 1600 for that one
https://forums.guru3d.com/data/avatars/m/245/245425.jpg
I hope that the card comes in black and red or atleast the the blue colour on the gold card needs to be rgb ( Cause RGB increase perf by 5% right :thumbup: )
https://forums.guru3d.com/data/avatars/m/223/223673.jpg
I hope that the card comes in black and red or atleast the the blue colour on the gold card needs to be rgb ( Cause RGB increase perf by 5% right :thumbup: )
or paint it yourself in Red for the extra 25kW :infinity:
https://forums.guru3d.com/data/avatars/m/227/227994.jpg
So they still havn't figured out the Heat problem, seeing it needs Watercooling ๐Ÿ˜›uke2:
data/avatar/default/avatar30.webp
it is just a reworked fury 9x or something completely new?
https://forums.guru3d.com/data/avatars/m/270/270718.jpg
Unfortunately I think the fact that not a single gaming benchmark was shown is telling. AMD love to toot their own horn when they've got something that can beat the comp. Remember all the shenanigans they pulled, trolling Nvidia during the R9 290x days? All we hear on Vega concerning gaming is...... crickets. That and at least on paper, it looks like an overclocked Fury X. If Fury X =8.6 TF and Big Vega= 13 TF that's right at a 50% increase. Most benchmarks Ive seen for games and synthetics like 3dMark show that the Ti is between 80% to 100% faster than the FuryX across the board. Numbers don't lie. To compete with the 1080Ti the Vega would have to be roughly twice as fast as FuryX, and the TFlops do not reflect that...
data/avatar/default/avatar35.webp
Unfortunately I think the fact that not a single gaming benchmark was shown is telling.
They showed no gaming benchmarks because these cards are for the HPC segment, they fight Tesla P100/V100, not GeForce. NVIDIA didn't show gaming benchmarks either when they announced the Tesla V100 based on Volta. But in general you are right, its awfully quiet around consumer/gaming Vega.
data/avatar/default/avatar36.webp
Different card, same tricks. In this space the competitors are the Nvidia Quadro lineup - P2000, P4000(approx. $850) , P5000, or P6000. Checking a site that has some Quadro performance numbers for SPECviewperf v12.1.1 and comparing to those on the Frontier website shows why they chose to compare against the TitanXP.
https://forums.guru3d.com/data/avatars/m/164/164033.jpg
Different card, same tricks. In this space the competitors are the Nvidia Quadro lineup - P2000, P4000(approx. $850) , P5000, or P6000. Checking a site that has some Quadro performance numbers for SPECviewperf v12.1.1 and comparing to those on the Frontier website shows why they chose to compare against the TitanXP.
My bigger issue with those Frontier numbers is how it compares to way inferior 2304 core Polaris 10.... I guess Nvidia gimps the Titan Xp vs P6000 somehow? They seem to be spec wise the same thing. Also those colors were horrible to read on the dark theme ๐Ÿ˜€
https://forums.guru3d.com/data/avatars/m/80/80129.jpg
My bigger issue with those Frontier numbers is how it compares to way inferior 2304 core Polaris 10.... I guess Nvidia gimps the Titan Xp vs P6000 somehow? They seem to be spec wise the same thing. Also those colors were horrible to read on the dark theme ๐Ÿ˜€
Quadro drivers have a bunch of performance enhancements for those kinds of workloads/software but the Titan XP uses the Geforce driver.
https://forums.guru3d.com/data/avatars/m/164/164033.jpg
Quadro drivers have a bunch of performance enhancements for those kinds of workloads/software but the Titan XP uses the Geforce driver.
Makes sense. So depending on workload it matters quite a lot. I guess AMD has similar drivers for their Pro cards with similar enhancements?
data/avatar/default/avatar06.webp
Difference between the GTX Titan X and Quadro P6000 Titan XP is a cut down variant of the GP102 core and the P6000 utilizes the full GP102 core, and has 24 gb gddr5x, 1 8-pin/250w, 12 TFlops with drivers specifically designed for this space. Based on the performance numbers I'll be curious to see what the Frontier's price is.
https://forums.guru3d.com/data/avatars/m/164/164033.jpg
Difference between the GTX Titan X and Quadro P6000 Titan XP is a cut down variant of the GP102 core and the P6000 utilizes the full GP102 core, and has 24 gb gddr5x, 1 8-pin/250w, 12 TFlops with drivers specifically designed for this space.
They did compare it to the Titan Xp not Titan XP. I know the difference is 12gb ram tho. But ye I think it is in the drivers. My question is does it make any difference for amd which drivers are used since they did use Crimson drivers for this test.
https://forums.guru3d.com/data/avatars/m/80/80129.jpg
Makes sense. So depending on workload it matters quite a lot. I guess AMD has similar drivers for their Pro cards with similar enhancements?
Yeah - IIRC Pro Duo is actually pretty cool because you can install either driver, workstation or gaming. Frontier Edition is a workstation card though, so those results should be with workstation drivers. Price matters a lot too though - if this thing is $1000, P6000 is ~$5000 and clearly not 5x the performance. Pharma says the P4000 is $850 but it's not in the benchmark he linked so idk how it compares to the Frontier.
https://forums.guru3d.com/data/avatars/m/164/164033.jpg
Yeah - IIRC Pro Duo is actually pretty cool because you can install either driver, workstation or gaming. Frontier Edition is a workstation card though, so those results should be with workstation drivers. Price matters a lot too though - if this thing is $1000, P6000 is ~$5000 and clearly not 5x the performance. Pharma says the P4000 is $850 but it's not in the benchmark he linked so idk how it compares to the Frontier.
I just checked the Frontier site which said the thing was tested on AMD graphics driver 17.20. Which left me wondering but could be nothing. I am guessing this will be way cheaper. But at least it should deliver in FP16 side ๐Ÿ˜€ And checking that SPECviewperf thing it seems AMD simply has been weaker for quite a while.
data/avatar/default/avatar22.webp
Google is your friend. P4000 review May be a good match for the P4000, but at the $1000 dollar mark people will be looking at the P5000 since performance has precedence over cost in this space.
https://forums.guru3d.com/data/avatars/m/164/164033.jpg
Google is your friend. P4000 review May be a good match for the P4000, but at the $1000 dollar mark people will be looking at the P5000 since performance has precedence over cost in this space.
Yup. I just think going from WX7100 to Vega in workstation environment is rather poor result considering how much more hardware the Vega seems to have. And how much more raw power. While the cards before that seem to scale somewhat.
https://forums.guru3d.com/data/avatars/m/267/267787.jpg
Huh? I'm confused. How is HBM1 faster than HBM2? R9 Fury X = HBM1 512 GB/s whereas Vega frontier = HBM2 480 GB/s?????