RX Vega 64 or GTX 1080 Ti

you are correct but OpenCL is not as fast as CUDA most of the time. and AMD cards work/worked better with OpenCL than Nvidia . ( have not looked at current comparisons so my info is out of date.) workloads really do determine which one better suits your needs as does time saved vs money spent. it is a tough ball to juggle and not devolve into fanboy problems.

Old thread but wanted to chime in with my experiences:

I have a 1050ti and a Vega 64. Used to have a Vega 56 as well.

Vega mines better than 1080ti, thatā€™s about it and depends on coin and algo at that. Good option for gaming if youā€™re locked into freesync and place a high priority on that aligning your monitor sync.

Also a good option if you want to support companies that make an open source effort, virtualize using Looking Glass (cannot do with Nvidia last I heard).

Iā€™ve successfully ran tensorflow on both cards, it runs faster on my 1050ti than on my Vega still. ROCm has made a lot of progress, and I am still very excited to see where it goes. However, at it currently stands your much better off getting a nvidia GPU if deep learning is a high priority. One exception I could see if if you are on a budget and need lots of ram for your ML model, you could maybe get by with Vega Frontier.

1 Like

I have a 1080ti, I did want to get a Vega64 but the Ti was cheaper at the time and smashed the V64 by a good percentage in 4k for all the games I play. Plus at lower watt draw and the ability to fit in my ITX cases, the 1080ti was the only choice if I was going for 4k gaming in a ITX form factor.

I think the Vega64 price is coming down now so maybe it will someday be a better choice but atm their only $120usd cheaper than 1080ti which is not a big enough gap, they need to be a good bit cheaper IMO.

I will be waiting for AMD to release a 1080ti killer card in ITX form factor before I move back to AMD. Also we are now seeing 4k 120hz screens appear on the market so only a matter of time before they have Freesync and HDR options together.

2 Likes

Where I live the Vega 64 is on par price wise with a 1080Ti. If you game, the 1080ti is the way to go. Im hoping the Vega 56 will come down to between 1070 - 1080 pricing but that is also priced not far behind the 1080ti :frowning:

At the original MSRP Vega was awesome. The V64 was aimed at the 1080 and Windows game wise was in the ballpark. The V56 was aimed at the 1070 and could overclock ok.

For value If I had to buy now would do passthrough with a 1080ti. I am waiting on pricing to drop or new AMD cards.

The 1080Ti is the faster card. If that is the whole question then this is a non-topic.

In my mind voting with your wallet isnā€™t a fanboy thing so Iā€™m just quickly gonna mention it.
I donā€™t like nvidia as a company, I donā€™t think I need to explain why these days.
Therefor I refuse to give them money. That is something of value to me personally.

3 Likes

This is the main reason I want Vega to succeed. However, I am a little disappointed by the Vega brand. I donā€™t the details but PC World Videos says at least one of the ā€œVegaā€ based laptops is actually Polaris?

I am disappointed that support for the 2200G and the 2400G isnā€™t here yet. Wendell said as much just two weeks ago I think. :frowning:

1 Like

Kaby Lake G is not the standard Vega chip anyway, itā€™s a semi-custom and speculation exists for who wanted it called Vega in the first place. Itā€™s sort of half-Vega as far as anyone can tell. Definitely not pure Polaris.

1 Like

AMD for what ever reason have always been behind on code release / development for Linux on their hardware, so it will always drag behind. If you need the latest and greated as soon as it comes out on Linux AMD isnā€™t for you. But Iā€™m not sure many brands are.

The plus side is the support does appear and itā€™s open sourced baked right in, thatā€™s been pretty consistent since there move to more open code. And also the reason I went with AMD hardware.

1 Like

Thatā€™s good. So I could see it as a blessing in disguise. I am forced to wait which means I donā€™t pay any early adopter tax. Thank you for the insight! Maybe by the end of the year (: Would be a very nice new year gift to myself x

I really hope that is not a thing. If AMD keep up the pace on open source to the point new cards run on linux out of the box. Thats my world. Now so far Nvidia does not give a fuck yet. Amd are going hard on a fringe thing that is linux and that got my respect 2 years ago and an RX480 sale.

2 Likes

I got a RX Vega for just over 600ā‚¬ including tax and I have to say itā€™s actually a pleasant experience. Fedora 28 works out of the box.

6 Likes

Looks amazing!

How much clearance to you have between the front intake fan and the GPU?

The Fractal Design Meshify C can handle GPUs with the length of 315mm and the Sapphire RX Vega 64 Nitro+ is 310mm long. In reality the clearance is like 2mm - 3mm. At first I was sceptical if this might impact cooling but the 2x 240mm Silent Wings 3 are working just as intended. I like that this came to your eye :smiley:

1 Like

Would call it ā€œperfect fitā€.

Am helping a friend with his hardline watercooling stuff, so attention to detail is on 100 right now^^

Cooling a R7 2700X and a RX Vega 64 under full load isnā€™t quite that easy. This is the maximum Iā€™d cool by heatsinks and air to be honest. While playing Ghost Recon Wildlands this setup almost draws 450W from the wall. Maybe a bigger case would also help the cooling a little.

Have you played with undervolting the card yet? I got really nice results from undervolting and using added percentage to the power limits from my 56 when it was still using the reference cooler.

Just a little yet. This card already boosts to about 1600MHz on itā€™s own. Iā€™m almost sure that this card will not hit 1700MHz even if undervolted and overclocked properly. I think the extra 100MHz might not be worth the effort. Saving energy and producing less heat might be the better goal by just undervolting the card. But if I raise the power limit the card will eventually draw even more power while undervolted. Vega is not easy to tame as it seams.

1 Like

It definitely takes some effort and time, and it doesnā€™t behave like Nvidia cards either, it wonā€™t sit at the boost clock if it doesnā€™t need to (which Iā€™ve come to like about it actually). Iā€™m running an OC of ~1650MHz and 1040MHz on the HBM2 but using a framerate limiter and I rarely see over 200w reported by the card. But thatā€™s on a 75Hz 1080p ultrawide so itā€™s not struggling to keep up in almost anything.

Benchmarks lie too, it wonā€™t be stable in real games as high as I can get it to run in Superposition or 3DMark.

I hardlined two Vega 64s and Threadripper 1950X. Threadripper never goes above 50c at 100% load @ 3.9 clock. That is full AVX load for hoursā€¦

I run the Vegas at 1692, 1100. They boost to 1700 -1750 and average 1710 never going over 45c core and 55c HBM2. That is at 100% load for sometimes days at a time doing rendering. Also 250w max.

These are compute cards for sure, that can game. I would still take them over a 1080 Ti because of HBCC. which comes in very handy.

If I was a gamer, I would go 1080 Ti all the way. There is no comparison when gaming, compute is a different matter entirely.

2 Likes