3d Mark score and GPU specs for the gaming RX Vega

[Liviu Cojocaru]

This is what I found on the WCCFtech site:

http://wccftech.com/amd-rx-vega-ben...nfirmed-1630mhz-8gb-hbm2-484gbs-of-bandwidth/

I want to see more benchmarks

 

[FreedomEclipse]

So long as they can keep the TDP down, It would make a decent upgrade

 

[Liviu Cojocaru]

So long as they can keep the TDP down, It would make a decent upgrade

I agree, the price will play a big role in this as well

 

[Vya Domus]

If this is true , it looks like they did gimp the FE after all. That or they somehow made all the optimizations work in less than a month.

 

[Liviu Cojocaru]

If this is true , it looks like they did gimp the FE after all. That or they somehow made all the optimizations work in less than a month.

They probably already had optimised drivers for the gaming version

 

[Vya Domus]

They probably already had optimised drivers for the gaming version

Yeah , Vega is after all much different from previous GCN architectures despite what people believe. It must require a whole different approach to driver optimization. I am sure AMD was working on this for a while now.

 

[Liviu Cojocaru]

I hope they did and I also hope that this will be a very competitive product otherwise I will have to pay more and more money for my new Nvidia gpu

 

[Vayra86]

So long as they can keep the TDP down, It would make a decent upgrade

We're talking 250-275W TDP here, so sorry to disappoint.

Vega is just Fury X v2 and this is clear as day by now. All the leaks confirmed it, and all the chatter about revolutionary new GCN is a pile of BS. The gains are highly situational, look at the 3dmark bench in the article, its all over the place. Some gains, some equal, net gain 15% over Fury X and best case average about 5-10% above the 1080 non OC.

Its horrifying and exactly what I've been predicting all the time.

Also keep in mind that the OC headroom will be limited, if not nonexistant. There is only one reason to be so late and so weird with the release scheme with the Frontier Edition and all that bs, and that is that your real world performance metrics are shit.

HBM2 is a saving grace here to meet a somewhat acceptable TDP and perf/watt and still hit GTX 1080 performance levels. About similar to how Fury X needed HBM to not go completely overboard and still match a 980ti - that is if you played 4K back then.

Vega will confirm that GCN needs to go the Ryzen path to further scale up in performance, in other words, soldering the dies together and put Infinity Fabric in between to make it scalable, ie Navi.

 

[Vya Domus]

We're talking 250-275W TDP here, so sorry to disappoint.

Vega is just Fury X v2 and this is clear as day by now. All the leaks confirmed it, and all the chatter about revolutionary new GCN is a pile of BS. The gains are highly situational, look at the 3dmark bench in the article, its all over the place. Some gains, some equal, net gain 15% over Fury X and best case average about 5-10% above the 1080 non OC.

No one thought it was going to be revolutionary , unless you were delusional. What wasn't BS however is that it's new and different , otherwise you wouldn't see these results.

HBM2 is a saving grace here to meet a somewhat acceptable TDP and perf/watt and still hit GTX 1080 performance levels. About similar to how Fury X needed HBM to not go completely overboard and still match a 980ti - that is if you played 4K back then.

HBM doesn't save that much power , they most certainly did not chose it because of that. HBM gets you more bandwidth and lower latency , nothing more. And that's why those gains are as you say situational , sometimes you need all the memory bandwidth you can get , sometimes not. Same thing about the massive core count. It hasn't been confirmed but I am certain the drivers are single threaded as well , a massive handicap that AMD is still ignoring , they had the chance now to fix it.

So wrong assumption , the hardware is OK. The TDP is what it is , 4096 cores at 1600+ mhz , what was everybody expecting ? You can still run one of them on a good 550W PSU. What more do you want ?

 

[ratirt]

We're talking 250-275W TDP here, so sorry to disappoint.

Vega is just Fury X v2 and this is clear as day by now. All the leaks confirmed it, and all the chatter about revolutionary new GCN is a pile of BS. The gains are highly situational, look at the 3dmark bench in the article, its all over the place. Some gains, some equal, net gain 15% over Fury X and best case average about 5-10% above the 1080 non OC.

Its horrifying and exactly what I've been predicting all the time.

Also keep in mind that the OC headroom will be limited, if not nonexistant. There is only one reason to be so late and so weird with the release scheme with the Frontier Edition and all that bs, and that is that your real world performance metrics are shit.

HBM2 is a saving grace here to meet a somewhat acceptable TDP and perf/watt and still hit GTX 1080 performance levels. About similar to how Fury X needed HBM to not go completely overboard and still match a 980ti - that is if you played 4K back then.

Vega will confirm that GCN needs to go the Ryzen path to further scale up in performance, in other words, soldering the dies together and put Infinity Fabric in between to make it scalable, ie Navi.

Why people cheering for different team always say they have predicted and foreseen stuff with the team they think of as crap? I don't get it. There's still driver optimizations there which I can bet can change the performance(NEW GCN NEW DRIVER). Fury is different than Vega and that's been said by AMD some time ago. New GCN. Similarities can occur that's obvious but it is different. Fury may have been a try version for HBM and some solutions that's also obvious.
I never say I was right till I see with my own eyes I was right. We've seen one benchmark and for you that confirms everything? Good luck with that attitude
And that NAVI. It's not out. We don't know what it is. There's no information about anything yet and still you try to predict and foresee how will this one perform. For me I need more data to even predict anything but you can go without it great. It's like shooting eyes closed with your shotgun up in the air hoping you get a duck. Good luck

No one thought it was going to be revolutionary , unless you were delusional. What wasn't BS however is that it's new and different , otherwise you wouldn't see these results.



HBM doesn't save that much power , they most certainly did not chose it because of that. HBM gets you more bandwidth and lower latency , nothing more. And that's why those gains are as you say situational , sometimes you need all the memory bandwidth you can get , sometimes not. Same thing about the massive core count. It hasn't been confirmed but I am certain the drivers are single threaded as well , a massive handicap that AMD is still ignoring , they had the chance now to fix it.

So wrong assumption , the hardware is OK. The TDP is what it is , 4096 cores at 1600+ mhz , what was everybody expecting ? You can still run one of them on a good 550W PSU. What more do you want ?

Well the die of Vega is bigger a bit. Would that mean it would use slightly more power putting aside HBM2 mem? It is possible to improve power consumption not changing the die size. I guess we will have to wait and see. I'm still waiting with my 4k monitor :/ crap

 

[FreedomEclipse]

I guess Its the price that matters then.... if its priced just over the 1070 or 1080 but offers close to 1080Ti performance then that could be its main selling point. Though if its a card that runs hot and drinks like a V8 then id have to have a serious think about it.

 

[jboydgolfer]

im a bit reluctant to believe ANY scores unless someone i know personally as credible posts them......nothing to lose if you wait to form an opinion based on fact, mostly humiliation if choose to be impatient and form an opinion based on conjecture.

 

[xkm1948]

Wccf, that is more than enough then. Along the way they have also "predicted" that Vega may challenge 1080Ti's gaming performance. 25 more days and we will see.

 

[Liviu Cojocaru]

Wccf, that is more than enough then. Along the way they have also "predicted" that Vega may challenge 1080Ti's gaming performance. 25 more days and we will see.

It's not only WCCFtech, there are others as well: https://www.overclock3d.net/news/gpu_displays/alleged_amd_rx_vega_benchmarks_have_surfaced_online/1

 

[ratirt]

Vega will confirm that GCN needs to go the Ryzen path to further scale up in performance, in other words, soldering the dies together and put Infinity Fabric in between to make it scalable, ie Navi.

or i.e Volta
http://wccftech.com/nvidia-future-gpu-mcm-package/

 

[Crap Daddy]

Though these results have to be taken with a grain of salt but after the FE results, it is getting clearer that this card will be at 1080 level, bigger and more expensive chip and memory, higher power consumption and launch after more than a year. Only the pricing might make these attractive which of course will be a big problem for the company.

 

[Steevo]

Though these results have to be taken with a grain of salt but after the FE results, it is getting clearer that this card will be at 1080 level, bigger and more expensive chip and memory, higher power consumption and launch after more than a year. Only the pricing might make these attractive which of course will be a big problem for the company.

This is very realistic, and typical of AMD.

I fully expect FE to have been forced to run ECC mode on memory, forced high accuracy on every calculation through the pipeline, it had to have worse timings as twice the memory in the same stacks, and other non-user changeable settings that make sure every pixel was rendered correctly.

The biggest win AMD could walk away with may not be the highest framerate, but not having the lowest framerate, which, at this point would effectively make it a better card. But time will tell.

 

[Vayra86]

Why people cheering for different team always say they have predicted and foreseen stuff with the team they think of as crap? I don't get it. There's still driver optimizations there which I can bet can change the performance(NEW GCN NEW DRIVER). Fury is different than Vega and that's been said by AMD some time ago. New GCN. Similarities can occur that's obvious but it is different. Fury may have been a try version for HBM and some solutions that's also obvious.
I never say I was right till I see with my own eyes I was right. We've seen one benchmark and for you that confirms everything? Good luck with that attitude
And that NAVI. It's not out. We don't know what it is. There's no information about anything yet and still you try to predict and foresee how will this one perform. For me I need more data to even predict anything but you can go without it great. It's like shooting eyes closed with your shotgun up in the air hoping you get a duck. Good luck



Well the die of Vega is bigger a bit. Would that mean it would use slightly more power putting aside HBM2 mem? It is possible to improve power consumption not changing the die size. I guess we will have to wait and see. I'm still waiting with my 4k monitor :/ crap

Call it experience or common sense, but at some point you really don't need to wait for the definitive review of a product to gauge what it'll be looking like. My prediction history shows some pretty good shotgun accuracy thus far

But of course, there is the off chance AMD has fooled us all and puts out the ultimate 1080ti killer with a TDP of 250w. Fingers crossed! /s

 

[HD64G]

Almost in the middle of performance between 1080 and 1080Ti is the expected performance level of the RX Vega when it goes on sale, depending on the game and the graphic api ofc. Drivers will get it very close to 1080Ti in a 5-6 months as happened with RX480 and 1060, since NCU cores of Vega arch are muchly altered from CGN ones and need time to optimise them for the new game engines. Price will make it a hit or miss imho. Clocks are great if possible to keep it stable close to 1600MHz. 8GB HDM2 will also help much in 4K.

 

[Vellinious]

I wish they'd quit using 3D Mark 11, and start using something at least a LITTLE bit updated. Timespy, Superposition....even Firestrike. 11 is so CPU dependent, it could easily skew numbers one way or the other.....get back to me when they show some REAL benchmarks.