TR Forums

Ok so now that cards are at least in the realm of a decent price which card should I go with.
I am looking at the Vega 64 vs the GTX1080. The price is within $50.00 of each other. And when looking at benchmarks, not just ones from here, it seems these cards are basically back and forth.

I know most people would go GTX1080, but the one thing I want you to take into consideration is I have a Asus 35" UWQHD (3440 x 1440) 1800R curved monitor with 100Hz and Adaptive Sync/FreeSync. So I would only get sync with the Amd Cards.

So knowing this would the cards(if priced similar) have any advantage over each other?

FreeSync is pretty nice in my experience. I'd personally get the one with FreeSync. From a gaming perspective, I'd pay a bit extra for FreeSync. I think its definitely worth it.

The other consideration is CUDA. There are more 3d renderers and accelerators that use CUDA (although OpenCL's base is still strong in Adobe products). So if you have any CUDA-accelerated applications (Octane Renderer for example), then you'd want to go with NVidia.

---------

Another consideration: what is the fan situation on both cards? Many Vega64 cards have the crappy stock cooler (whose only real purpose seems to be... rip it off and throw a liquid cooler onto it). If you have a double-fan or triple-fan setup, its much better for most use cases. A proper fan is worth at least $20, maybe more.

dragontamer5788 wrote:

FreeSync is pretty nice in my experience. I'd personally get the one with FreeSync. From a gaming perspective, I'd pay a bit extra for FreeSync. I think its definitely worth it.

The other consideration is CUDA. There are more 3d renderers and accelerators that use CUDA (although OpenCL's base is still strong in Adobe products). So if you have any CUDA-accelerated applications (Octane Renderer for example), then you'd want to go with NVidia.

---------

Another consideration: what is the fan situation on both cards? Many Vega64 cards have the crappy stock cooler (whose only real purpose seems to be... rip it off and throw a liquid cooler onto it). If you have a double-fan or triple-fan setup, its much better for most use cases. A proper fan is worth at least $20, maybe more.

I am looking at the Asus Strix version of both cards. So they both have higher clock speeds and the triple fan design. Also the reason I wanna stick with the Strix versions is because my monitor is a strix monitor so they can do color linking as far as their LEDS go. (Though this is not a huge deal to me I just figure may as well make them match.)

Having a Freesync monitor seems like a reasonable reason to choose the Vega 64 over the 1080.

1070Ti or Vega 56 and flash it to 64 BIOS.

Vega architecture has some legroom left in it while Pascal is fully matured and will not see any significant improvements.

Freesync vs fan noise.

You can obviously get quiet Vegas and noisy 1080s, but there's a roughly 100W difference in power consumption under load.

I decided on Freesync over fan noise - no regrets. Additionally, I have a lot of experience trying to get Nvidia cards to behave reliably at refresh rates between 60 and 120. They work fine at either, and they work fine at 144, but for some reason all of the many Nvidia cards and monitor combinations I've encountered seem to frameskip at 75Hz, 85Hz, 100Hz. That applies to LG, Samsung, Achieva, Acer, and Dell monitors, using a 760, 660Ti, 960, 970, and 1060. It might not be a universal problem, but it does seem widespread.

Krogoth wrote:

1070Ti or Vega 56 and flash it to 64 BIOS.

Vega architecture has some legroom left in it while Pascal is fully matured and will not see any significant improvements.

The price difference between the vega 56 and vega 64 is 50 dollars. And a very similar price difference between the 1070ti and 1080 is even less.
It just doesn't seem worth it to save a few bucks.

Well, a Vega56 runs cooler and quieter, even when you flash a 64 BIOS and overclock the HBM.

If you're into tinkering, that's probably the best overall solution, since the 56 performs very closely to the 1080 at 945MHz HBM2 (technically that's not even overclocking the HBM2 modules). It's failsafe too, since the power-save BIOS on Vega cards is read-only, meaning you can't possibly* brick the card with a bad flash.


* - It may be that Asus have deviated from the dual-BIOS 1x locked/1x flashable standard that most cards adhere to. I would double check the exact specific model and version before BIOS flashing anything.

If a Vega 56 with a 64 BIOS runs the same as Vega 64, why would it be cooler?

Radeon RX Vega56 has 7/8 as many cores enabled as Radeon RX Vega64. With the original stock cooler and default Radeon Wattman settings, Vega64 may be limited by its cooler. With fewer cores, Vega56 generates less heat per cycle, so there's enough headroom in Vega56 to overclock it enough to reach a similar performance level before the cooler can't keep up. With the beefy coolers on the Asus Strix cards that you're looking at, the Vega64 may be able to clock higher than the stock card does.
https://www.asus.com/us/Graphics-Cards/ ... 8G-GAMING/

Of course, the argument goes the other way, too. If you underclock the Vega64 to Vega56 performance levels, it runs even cooler and quieter.

I would not go with either, they are too old (10.5 and 13 months old).

Rumours are strong for the GTX 1180 based on Turing to land next month, and AMD is already sampling 7 nm Vega 20 for launch this year (only talking so far about selling it in compute parts, not gaming ones, but if Turing kills the ability to sell Radeon RX Vega 56 and 64 above production cost, its hand will be forced).

Do you need to game at super high settings right now, that only those cards will support?

Topinio wrote:

I would not go with either, they are too old (10.5 and 13 months old).

Rumours are strong for the GTX 1180 based on Turing to land next month, and AMD is already sampling 7 nm Vega 20 for launch this year (only talking so far about selling it in compute parts, not gaming ones, but if Turing kills the ability to sell Radeon RX Vega 56 and 64 above production cost, its hand will be forced).

Do you need to game at super high settings right now, that only those cards will support?

I was looking into the gtx 1180 but from what I saw online Nvidia said they were not coming out any time soon.

The problem is I have quite the old video card (HD 7950). And it just isn't cutting it in all games.

If the 56 and 64 are only within $50 of each other, definitely get the 64. If anything, a 64 cooler should be better suited to handle the TDP than a 56 that's being overclocked to match a 64.
Certainly since you have a FreeSync monitor, that's a good reason not to go with a GTX1080. VRR is fantastic.

Gandolf wrote:

I was looking into the gtx 1180 but from what I saw online Nvidia said they were not coming out any time soon.

The problem is I have quite the old video card (HD 7950). And it just isn't cutting it in all games.

Sure, that's too old. But ... the GTX 1080 is just over 2 years old (I thinko'd my last post, sorry) and I don't believe the GTX 1180 will be long now.

Even if you don't think any of the recent stuff amounts to anything and is definitely not NVIDIA PR (lots of rumours, a retailer putting up pricing, a talk scheduled at Hot Chips, apparently a press event scheduled during Gamerscon (sp?), an alleged ES board on reddit) the fact is this is the longest gap NVIDIA has ever had between generations of desktop GPUs.

The gaps in months between its highest-end. non-refresh, non-Titan cards have been 19 23 17 18 11 16 13 17 19 21 23 14 16 20 25* and I reckon that NVIDIA knows it needs to keep things fed.

Most recent rumor I heard for the 1180 was Gamescom (August 21) and even then it's not certain that the launch will be anything more than a paper launch at that point. That means, you'll have retail launch in September, and "price premium" would stretch out till close to the end of 2018, especially if AMD doesn't have a Vega gaming card launch in 2018. So if you're waiting for the 1180 at reasonable prices, you're going to be waiting....at least 5 more months.

Green teams features are too exclusive for me to consider red team basically ever. Freesync is nice (unarguably a better implemented solution than G-sync), but I use the streaming features of GeForce Experience weekly, and AMD doesn't have anything like that.

DPete27 wrote:

So if you're waiting for the 1180 at reasonable prices, you're going to be waiting....at least 5 more months.

That's possible, I guess, if OP couldn't get a launch card at RRP.

However, buying a 2.1 y.o. GTX 1080 now just seems to me to be like buying a GTX 480 in spring 2012.

Yes, OP has an ancient HD 7950 and should upgrade it, but a thing to remember is that NVIDIA cards age way less gracefully than that that 6.5 y.o. Radeon has.

Topinio wrote:

DPete27 wrote:

Yes, OP has an ancient HD 7950 and should upgrade it, but a thing to remember is that NVIDIA cards age way less gracefully than that that 6.5 y.o. Radeon has.

That myth needs to die. When the 7950 was released it was second only to the 7970 and pretty much kicked sand in the face of NVidias offerings. Then they didn't do very much and NVidia accelerated away and AMD found themselves an also-ran. So, in relative terms a new NVidia card sure makes an old NVidia card look sad. I don't expect the performance of my GTX970 to somehow deteriorate over the years. The 1070 might be WAY faster and the 1170 more so again but my 970 will serve well until I care about gaming again (GTA VI).

Having said that if I'd get the Vega 64 if I had a Freesync monitor already. Mad not to.

Edit: it seems more people are gaming with 970s then, well, pretty much all current gen AMD cards put together so I imagine support from game publishers will be pretty solid for a while longer.

Pancake wrote:

Topinio wrote:

NVIDIA cards age way less gracefully than that that 6.5 y.o. Radeon has.

That myth needs to die. When the 7950 was released it was second only to the 7970 and pretty much kicked sand in the face of NVidias offerings. Then they didn't do very much and NVidia accelerated away and AMD found themselves an also-ran.

My comment is not about the relative performance of subsequently-launched products from NVIDIA and AMD.

(Obviously, NVIDIA's cards released over the last 4 years have been better than AMD's, though I think not on price-performance or arguably feature set.)

My comment is about the fact that AMD cards have aged better. 6 years ago, £250-300 or so would have got me that HD 7950 or a GTX 660 Ti. 6.5 years ago, for £350 or so I could have got the HD 7950 at launch or a GTX 580. Which vendor would you rather have paid the money for and been stuck on for the length of time OP has?

Topinio wrote:

6 years ago, £250-300 or so would have got me that HD 7950 or a GTX 660 Ti. 6.5 years ago, for £350 or so I could have got the HD 7950 at launch or a GTX 580. Which vendor would you rather have paid the money for and been stuck on for the length of time OP has?

Neither.You're basically arguing over degrees of hopeless obsolescence.

Topinio wrote:

My comment is about the fact that AMD cards have aged better. 6 years ago, £250-300 or so would have got me that HD 7950 or a GTX 660 Ti. 6.5 years ago, for £350 or so I could have got the HD 7950 at launch or a GTX 580. Which vendor would you rather have paid the money for and been stuck on for the length of time OP has?

6 years ago I would have bought the 7950. It was the demonstrably superior card. And in fact I did buy the previous gen 6950 which served me well until I got a GTX970. But it's not as if some GTX660Ti or GTX580 has suddenly aged and become decrepit unless you're unlucky and your board died. It's still doing what it did at the time. Probably better with driver updates.

So, thought experiment: would you have bought a R9 290 or a GTX 970 a few years ago? I would think if you bought the R9 290 you would be standing there with a sad face when the RX480 came out. The GTX 970 is still quite fast and still quite energy efficient. So, generalisations like "Company A products age worse than Company B" are really a bit silly. You buy a product at a given time with given performance characteristics. You can factor in what you will about the relative size of the development teams and sheer R&D expenditure of NVidia vs AMD, the sheer numbers of units sold and used by gamers and how much support you're likely to get over the years for game optimisation, testing etc.

1180 is likely to be 3584 CUDA cores using Volta architecture on 12nm process with 16GB GDDR6.

If you have been following Volta and TSMC's 12nm process, you will know that Volta doesn't really add any significant gaming-focused architectural advantages over Pascal, and TSMC's 12nm process is really just a power-efficiency tweak of the 16nm process that Pascal uses, so the 1180 will roughly match a 1080Ti, but at maybe 15-20% lower power usage.

In terms of memory bandwidth, the 1180 will likely be 256-bit GDDR6, at ~15Gb/s compared to the 384-bit GDDR5X, at 11GB/s of the 1080Ti, the drop in bus width is countered by the improvement in speed, so again the 1180 looks similar to the 1080Ti.

AMD don't have competition for the 1080Ti, so Nvidia won't be aggressive on pricing. Expect a bare-minimum of $699 but with Nvidia likely having an untouchable halo product at least until Vega 7nm arrives, I wouldn't be remotely surprised to see them tack another hundred or two hundred dollars onto the price tag until all of the old 1080Ti inventory is sold off. They can afford to milk the early-adopters without viable competition and Nvidia have a solid track record going back over 10 years of doing exactly this.

Chrispy_ wrote:

AMD don't have competition for the 1080Ti, so Nvidia won't be aggressive on pricing. Expect a bare-minimum of $699 but with Nvidia likely having an untouchable halo product at least until Vega 7nm arrives, I wouldn't be remotely surprised to see them tack another hundred or two hundred dollars onto the price tag until all of the old 1080Ti inventory is sold off. They can afford to milk the early-adopters without viable competition and Nvidia have a solid track record going back over 10 years of doing exactly this.

Probably correct.

The thing I find odd is there seems to be a perception out there that NVidia is unique in this regard. It's business 101: sell your products at the price point that allows you to maximize profit. Intel, Apple, Sony, Microsoft, they all do the same thing: even AMD has priced their products at a high level on the rare occasions that they have been in a dominant position (such as in the Athlon vs. Netburst era).

If you have the PSU and cooling to support a Vega64 AND plan on sticking with your monitor grab the Vega. I prefer Pascal, but adaptive sync is really nice. Just be aware that Vega could be the new Hawaii and you might not be able to ugprade for several years.

Kretschmer wrote:

Just be aware that Vega could be the new Hawaii and you might not be able to ugprade for several years.

As someone who upgraded a Hawaii XT to a Vega 10 Pro, I this message.

On timelines, it all depends which market segments get fed when, but you can look at the Linux kernel support as foreshadowing, and that shows

Polaris 10:
2016-03-26 Linux driver initial support
2016-06-29 Consumer boards
2016-11-10 Pro boards
2016-12-12 Compute boards

Vega 10:
2017-03-20 Linux driver initial support
2017-06-21 Compute boards
2017-06-27 Pro boards
2017-08-14 Consumer boards

Vega 20:
2018-05-17 Linux driver initial support
??

K-L-Waster wrote:

The thing I find odd is there seems to be a perception out there that NVidia is unique in this regard. It's business 101: sell your products at the price point that allows you to maximize profit.

I don't understand that mentality either. It's free-market capitalism 101; The only reason for a company to NOT set a price point that allows them to maximise profit is if they're taking a hit to increase their market penetration (before then selling at a price point that allows them to maximize profits).

As a marketshare leader, and uncontested at the very high end, Nvidia have no reason to undersell their 1180 and every reason to avoid hurting sales of their continuing Pascal lines.

Topinio wrote:

Kretschmer wrote:

Just be aware that Vega could be the new Hawaii and you might not be able to ugprade for several years.

As someone who upgraded a Hawaii XT to a Vega 10 Pro, I this message.

On timelines, it all depends which market segments get fed when, but you can look at the Linux kernel support as foreshadowing, and that shows

Polaris 10:
2016-03-26 Linux driver initial support
2016-06-29 Consumer boards
2016-11-10 Pro boards
2016-12-12 Compute boards

Vega 10:
2017-03-20 Linux driver initial support
2017-06-21 Compute boards
2017-06-27 Pro boards
2017-08-14 Consumer boards

Vega 20:
2018-05-17 Linux driver initial support
??

We're talking about gaming, not Linux.

I currently have a GTX 980 and a G-sync monitor. With that being said, I'd go with a Vega 64 and Freesync monitor given the cost of G-sync. In 2018, you have to have adaptive sync and Freesync makes the most sense economically.

Kretschmer wrote:

We're talking about gaming, not Linux.

¿por qué no los dos?

Srsly, though, card availability hasn't been that many days after the Linux kernel support, even if YMMV on getting decent gaming performance without booting Windows.

• Radeon RX 580 was 89
• Radeon Instinct MI25 was 87 and Radeon RX Vega 64 was 141
• today is 49 days and counting...

Kretschmer wrote:

If you have the PSU and cooling to support a Vega64 AND plan on sticking with your monitor grab the Vega. I prefer Pascal, but adaptive sync is really nice. Just be aware that Vega could be the new Hawaii and you might not be able to ugprade for several years.

I have a corsair Obsidian 450D case and a Antec high current pro 850 power supply. My cpu is liquid cooled. Is this enough? My bottleneck could possibly be the case being too small?