ASUS ROG GeForce RTX 3050 STRIX OC review
Palit GeForce RTX 3050 DUAL OC review
Gigabyte RTX 3050 Gaming OC review
ASRock Z690 Taichi review
Radeon RX 6500 XT review
MSI Vigor GK71 Sonic keyboard review
MSI GeForce RTX 3080 Suprim X 12GB review
TeamGroup A440 PRO Special 5eries 4TB NVMe review
Core i5 12400 processor review
MSI MAG B660M Mortar WIFI review
More Vega Frontier Edition Performance Numbers Surface
Earlier today new benchmarks of the AMD Vega Frontier Edition have surfaced. This round Ekin Tech got his hands on a Vega Frontier Edition, and in a live broadcast session they outted game performance numbers.
First a bit of a warning as the photos that you see are screenshots from that live-stream, so that's not the best quality to look at. Secondly these are FCAT FPS plots (50% is roughly what would be an average FPS) and thus can be a bit harder to grasp.
I also want to state very clearly that the the Radeon RX Vega Frontier Edition is not meant to be a gaming card whatsoever. However everybody is jumping on it as hey .. it's a Vega powered graphics card and we've been waiting for a year on some results right ? That does not imply or mean the results you see can be compared to what the consumer product will be like. The Frontier Edition is not intended to be a gaming card, hence the drivers have not been developed for it either (despite showing a gaming mode).
The card shines in some yet lacks in other benchnmark tests, at a resolution of 2560x1440, Vega Frontier Edition is close to the GeForce GTX 1080 Ti in Fallout 4. Then in Grand Theft Auto V it actually gets beaten by a GTX 1070, the same in The Witcher 3 but shows higher performance than a GTX 1070 in Hitman and Dirt Rally. So that last bit isn't really good news.
In Ultra HD 4K we see a totally different picture. The Vega FE matches the GTX 1080 mostly and closely in Dirt Rally and Fallout 4. In Hitman, the Vega FE is slower than the GTX 1080, 50 FPS versus 60 for the 1080.You'll see similar results for the Witcher 3.
Also good to note is that power consumption was measured and rated at 280 Watts, that is not that bad really. In the end though, everybody is hyped and excited. We do know that the final consumer parts will be clocked higher and will get actual proper drivers. Time will tell how good (or bad) Vega really is, but this should get you through the weekend I guess, have a good browse, click the thumbnails to see the full-size screenshots :)
Sources: Ekin @ Linus Tech Tips, PC Perspective YouTube Channel
Noisiv
Senior Member
Posts: 8143
Joined: 2010-11-16
Senior Member
Posts: 8143
Joined: 2010-11-16
#5448222 Posted on: 06/30/2017 05:54 PM
Touche
Well it's stupid for AMD to show benches in computational stuff against the Titan Xp and not Quadro cards and then expect not to compare those to cards in gaming also.
Touche
wavetrex
Senior Member
Posts: 1444
Joined: 2008-07-16
Senior Member
Posts: 1444
Joined: 2008-07-16
#5448225 Posted on: 06/30/2017 06:02 PM
It's pretty clear why they kept delaying it, and delaying it more.
Not HBM2...
... but the lackluster performance.
I am an ATI (well, AMD) fan, have used all their generations since X850 Pro, all the way through X1950 later a big jump to R4870, then 6950, then 7970GE ... but I stopped after.
Because they simply lost, and continue losing to this day.
Sad, but true.
I was really hoping for a comeback with Vega, but seems that is not going to happen. Competitive price or not, it is a slower and more inefficient GPU than nVidia's offerings from last year.
It's pretty clear why they kept delaying it, and delaying it more.
Not HBM2...
... but the lackluster performance.
I am an ATI (well, AMD) fan, have used all their generations since X850 Pro, all the way through X1950 later a big jump to R4870, then 6950, then 7970GE ... but I stopped after.
Because they simply lost, and continue losing to this day.
Sad, but true.
I was really hoping for a comeback with Vega, but seems that is not going to happen. Competitive price or not, it is a slower and more inefficient GPU than nVidia's offerings from last year.
Exascale
Senior Member
Posts: 390
Joined: 2017-06-09
Senior Member
Posts: 390
Joined: 2017-06-09
#5448229 Posted on: 06/30/2017 06:13 PM
As a RX480 user , il say this , i kinda knew all this would hapend when they stoped to hype VEGA , was 90% shure it will be a fail product , i really hope AMD next time would stop developing stupid product's VEGA was nothing but a money sink , they literaly burned money , why they had to bother with all the HBM2 still make's me mad, in stead they should have just focused to make a bigger refined polaris chip to beat @ least the 1080 but cheaper , but no they got stuborn with all the HBM2 crap , while NVIDIA milks the market whil normal gddr5 , i even lost all hope for NAVI to , i dout GPU division from AMD can do anything @ this point.
We will for shure have the same problem but with gpus that happend with cpu's , nvidia will lead for some year's to come while AMD will try to keep it's head above the wather .
Nvidia has had GPUs with HBM2 for longer than AMD has. HBM and HMC are the future. No one serious is using anything but that in their highest end chips anymore. Fujitsu was first with HMC on the PrimeHPC FX100s SPARC XIfx, Intel has Knights Landing with HMC, Nvidia P100 and V100.
DIMM based DDR and DDR3 derived GDDR SGRAM are all old school. Fine for gaming but not for serious computing. AMD is being smart by having one big chip that goes into a bunch of things and uses HBM, particularly with the foresight to give it 49bit VA and 48bit PA for large HSA HPC installations as well as their HMCC caching technique.
People dont understand this GPUs architecture or appeal. It offers similar performance to GP100 in specific workloads, when used in MI25, for a fraction of the cost.
The HBCC for the HBM2 is a very sophisticated technology that can be used in conjunction with Infinity Fabric and HSA to work on massive datasets in memory.
It has a built in NVMe controller to directly talk to SSDs, which allows for holding even more massive datasets close to the GPU and having it be directly addressable.
This particular GPU kind of has its own niche. I am not just talking about the FE card. I mean the big Vega chip in general. The BIGGEST problem i can see with it is no ECC on the cache and registers. That means some workloads are off limits to it.
As a RX480 user , il say this , i kinda knew all this would hapend when they stoped to hype VEGA , was 90% shure it will be a fail product , i really hope AMD next time would stop developing stupid product's VEGA was nothing but a money sink , they literaly burned money , why they had to bother with all the HBM2 still make's me mad, in stead they should have just focused to make a bigger refined polaris chip to beat @ least the 1080 but cheaper , but no they got stuborn with all the HBM2 crap , while NVIDIA milks the market whil normal gddr5 , i even lost all hope for NAVI to , i dout GPU division from AMD can do anything @ this point.
We will for shure have the same problem but with gpus that happend with cpu's , nvidia will lead for some year's to come while AMD will try to keep it's head above the wather .
Nvidia has had GPUs with HBM2 for longer than AMD has. HBM and HMC are the future. No one serious is using anything but that in their highest end chips anymore. Fujitsu was first with HMC on the PrimeHPC FX100s SPARC XIfx, Intel has Knights Landing with HMC, Nvidia P100 and V100.
DIMM based DDR and DDR3 derived GDDR SGRAM are all old school. Fine for gaming but not for serious computing. AMD is being smart by having one big chip that goes into a bunch of things and uses HBM, particularly with the foresight to give it 49bit VA and 48bit PA for large HSA HPC installations as well as their HMCC caching technique.
People dont understand this GPUs architecture or appeal. It offers similar performance to GP100 in specific workloads, when used in MI25, for a fraction of the cost.
The HBCC for the HBM2 is a very sophisticated technology that can be used in conjunction with Infinity Fabric and HSA to work on massive datasets in memory.
It has a built in NVMe controller to directly talk to SSDs, which allows for holding even more massive datasets close to the GPU and having it be directly addressable.
This particular GPU kind of has its own niche. I am not just talking about the FE card. I mean the big Vega chip in general. The BIGGEST problem i can see with it is no ECC on the cache and registers. That means some workloads are off limits to it.
MorganX
Senior Member
Posts: 132
Joined: 2015-03-30
Senior Member
Posts: 132
Joined: 2015-03-30
#5448230 Posted on: 06/30/2017 06:14 PM
Optimistic here
Well, I guess I'm the only person that's pretty optimistic at this point. With results being inconsistent and all over the place (clearly driver related), the clock speed not actually reaching 1600 but averaging 1400, not having optimized gaming drivers, running software none of which written to take advantage of any of it's advanced features (alleged features anyway), and no DX12 or Vulkan, yet still manages to be in 1080ti territory here, 1080 there, 1070 over there, I'm looking forward to the gaming variant, optimized drivers, and modern games to take advantage of the architecture.
But most of all, price/performance. I don't have a 1080 now, so if I can flirt with 1080ti at a price point below current 1080 prices, plus get Freesync out of the deal, I'll be good.
Optimistic here
Well, I guess I'm the only person that's pretty optimistic at this point. With results being inconsistent and all over the place (clearly driver related), the clock speed not actually reaching 1600 but averaging 1400, not having optimized gaming drivers, running software none of which written to take advantage of any of it's advanced features (alleged features anyway), and no DX12 or Vulkan, yet still manages to be in 1080ti territory here, 1080 there, 1070 over there, I'm looking forward to the gaming variant, optimized drivers, and modern games to take advantage of the architecture.
But most of all, price/performance. I don't have a 1080 now, so if I can flirt with 1080ti at a price point below current 1080 prices, plus get Freesync out of the deal, I'll be good.
Click here to post a comment for this news story on the message forum.
Senior Member
Posts: 2983
Joined: 2004-12-17
Then call it game development mode and don't put this on your website:
https://pro.radeon.com/en-us/product/radeon-vega-frontier-edition/
Or better yet, don't release a card with no clear target for what it should be used for.
Seems like the decision for a HBM2 only memory controller has cemented Vega's price tag.