• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Poll: ** The AMD VEGA Thread **

On or off the hype train?

  • (off) Train has derailed

    Votes: 207 39.2%
  • (on) Overcrowding, standing room only

    Votes: 100 18.9%
  • (never ever got on) Chinese escalator

    Votes: 221 41.9%

  • Total voters
    528
Status
Not open for further replies.
As a user of a 1gb card with GTA5 I can tell you that streaming open world is a disaster for frame rates currently if you dont have the big space available to buffer it all currently and feed the GPU. GTA is probably well coded but they cant overcome that choke and apparently it applies in some amount less dramatically all the way upto top spec machines

I get like 45% gpu usage when going top speed through the middle of a city. I can tell it must be size related as if the LOD factor is increased by me maybe flying 100 feet overhead it all becomes manageable and frame rates double.
Thats the obvious effect that leads us into ever bigger vram counts. If they are using smart prediction to overcome rather brute force then great, I think that could be the best direction. I'm a big fan of realistic open worlds, I dont want them to be limited to the idea of just keep doubling up the memory and its fine.

Game engines are always directed in their backbone to the largest part of gaming population and that is pleb spec machines not enthusiast. Hopefully Vega can see a benefit somewhere maybe only in cost: performance by not needing the 16gb it might have required to do well otherwise


What is the minimum spec for GTAV? If you don't have enough Vram you don't have enough, not much you can do and even HBCC will need a reasonable buffer.

AMD's HBCC could be a good solution on such scenarios but as I said before, it first requires the game to use more VRAM that the card has but the card still needs sufficient vram.for the current scene. As.you witnesses games already exist that stream resources without requiring something like HBCC and if you had a 8GB card you wouldn't have had issues in GTA. So the value of HBCC is sandwiched between cards that have plenty of vram to spare and cards that don't have enough regardless of HBCC, and only applies to so subset of games that need to use giant data sets.

One of the reason megatexture didn't really take off was the resources required to create massive unqiue textures was prohibitive. Instead of an artist creating say a wall texture tiled across all.walls in. Room and repeated in different rooms they needed to create vast unqiue textures for each surface to really exploit the technology. That took lots of time and made for giant game files. The costs of art production dwarfs programming in many modern games. So even if there was technology to stream huge data sets most games probably won't go that route due to production costs.
 
Lol they r already doing that, classic pc gamers always try to find something to hate amd ...

how are we trying to hate amd? this is an objective look at their newest product.

we can safely say this isn't a professional card right? from the numbers above it doesn't remotely compete in professional workloads.

so that makes it a prosumer card, like the titan.

now, you've got 4gb extra vram, and fp16 over the titan, which is good.

but, you've got 125w+ tdp higher, less memory bandwidth (which is also important in tasks like deep learning)

the higher tdp means you NEED the aio version, ad 375w doesn't have a hope in hell of being cooled by the blower design.

so, for a workstation you're comparing a £1150 titan vs a £1700 vega, the vega has better fp16, but worse fp32 (the titan runs at higher than normal boost clocks even with stock blower) and the titan has better memory bandwidth.

it just seems like a very odd card to release, and I think most people will agree £500 for an AIO added is insane pricing.
 
Or the AIO one will run faster than the blower hence the higher wattage? It does state that these results are for the blower cooler and not the AIO.

do you think amd would actually quote lower performance figures than what they actually have?

never,has any company ever done this, except surprisingly Nvidia has higher tflops out of the box than quoted because boost 3.0 has higher clocks than advertised.

but other than that, I doubt amd would willingly say it took then 375w to get 13tflops of fp32...
 
Nvidia can sell their HPC/deep learning GPUs for 10K a pop but AMD can only muster a fraction.

Those benchmarks are just bogus because they aren't using the correct the drivers for the TitanXP. They need to use a quador with Quadro drivers.


But the whole table is stupid. They are claiming the TitanXp only has 190GFlop half precision which is bogus, all nvidia cards will do half precision at least the same as the single precision, that is self evident. They must be using double precision figures for the Titan and half for Vega, what complete nonsense.

Yeah, now makes sense that they are priced much lower that I expected. They just don't have the performance. 980ti performance with 375w, oh dear


Anyone else seeing a pattern?
 
Well there we have it, the Vega Frontier Edtion can barely beat the Radeon Pro WX7100 ( £660 ) aka RX 480. :p

Vega FE
Catia = 135.75
CREO = 83.94
SolidWorks = 114.88
Cinebench OpenGL = 183.28

Radeon Pro WX 7100 ( £660)
Catia = 104.34
CREO = 82.08
SolidWorks = 111.55
Cinebench OpenGL = 197.11

Titan Xp
Catia = 107.29
CREO = 65.2
SolidWorks = 67.75
Cinebench OpenGL =169.72

Clearly only ONE person on the forum predicted Vega's performance correctly! That man is @LoadsaMoney :eek:
I believe he's already pre-ordered two to replace his Furys.


wdWIN8C.png


5LuslYx.png
 
Last edited:
Looks like I won't be selling my RX 480 after all then. :(

Looks like the FE and TitanXp were run using gaming drivers.

Unless AMD somehow sliced their IPC in half with Vega, a 13.1 TFLOP card shouldn't be that close to a 5.7 T FLOP one in similar workloads and with similar drivers.
 
Looks like the FE and TitanXp were run using gaming drivers.

Unless AMD somehow sliced their IPC in half with Vega, a 13.1 TFLOP card shouldn't be that close to a 5.7 T FLOP one in similar workloads and with similar drivers.


simple, the vega FE isn't a professional card, it uses the conventional drivers just like the titan XP.

basically it's a more expensive titan with a much higher tdp, although I know you can flash the quadro drivers to the titan to give it higher performance in professional programs, not sure if you can do the same with the vega.
 
So far nothing about Vega is impressing me. I will probably wait till August and if it barely beats a 1080 and has a stupid high TDP, I think I will just get Volta at that point
 
VEGA(1).jpg


Under performing!!!!!


The figures above look very dodgy indeed.

Here is my old Pascal Titan at stock running Cinebench, it would score a lot more but it is CPU bottlenecked. I have not bothered to use my Titan Xp cards as they are even faster lol.

ZaLewqN.jpg

If someone runs a 1080 Ti on a good quad core like a 7700k they should be able to do even better than my Titan.:)
 
The figures above look very dodgy indeed.

Here is my old Pascal Titan at stock running Cinebench, it would score a lot more but it is CPU bottlenecked. I have not bothered to use my Titan Xp cards as they are even faster lol.

If someone runs a 1080 Ti on a good quad core like a 7700k they should be able to do even better than my Titan.:)

Here's what Hot Hardware found running with a 5960X

Looks like Cinebench is all over the place for GPU testing.

QUpSnSfjQ-a_l0HhHfNZXg.png
 
So far nothing about Vega is impressing me. I will probably wait till August and if it barely beats a 1080 and has a stupid high TDP, I think I will just get Volta at that point

I'm waiting for the 'gaming' focused Vega's, proper reviews, & I expect performance to improve with drivers. Ryzen has shown that even if the architecture is excellent other factors can rather significantly impact performance (BIOS ins Ryzen's case).
 
I wonder where that report got the TDP values as I see nothing of the sort on AMD's published article. Other than that though they're not fantastic numbers. Let's hope these numbers are not indicative of gaming performance.
 
How do you NEED the AIO if they're selling one clearly without it? The blower is a lower TDP and won't run at 300W.


Because it runs at far lower clocks and won't have the same performance figures.

I'm guessing 1600MHZ for the 375w water cooled,. 1200W for for the 300w blower. Gaming Vegas could be 1000-1200w. 300w is.srillmtoonhigh for a gaming card really.

Hope I'm.wrong but something needs to explain that power difference
 
Status
Not open for further replies.
Back
Top Bottom