Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
GTX 1070 or 1080 for 5.1
- Thread starter pethenis
- Start date
- Sort by reaction score
It depends on your usage case. Are you going to play games, or use it for professional video work, or something else? For games, you'll probably be CPU limited at anything less than 4K, but if you're going to be using it to run DaVinci Resolve you'll want the biggest GPU you can find.
Yeah, its all about usage case. Its funny I'll see posts on here on how the cMP can't fully use these GPUs or is bottlenecked cause of other factors. But I render all the time and utilize all 12 cores and full GPU VRAM on my 4K video work. When you work with hi-res footage with multiple effects, color grades, etc., you can use all the power you can get. Even with one effect like NeatVideo, you can maximize all system resources, since it allows you allocate as many cores and GPUs with all VRAM to the effect itself.
But yes, if you aren't doing anything in that realm with your machine, then you have to ask yourself whether an upgrade like this is even needed.
I edit video (Some 4K) in Premiere, and colorgrading in Davinci Resolve on a 4.1 -> 5.1 12 core 3.46 with 48 Gig ram and SSD's.
We have very similar systems. I currently use the Titan X (Maxwell) to great effect. A 1080 should serve you well, assuming the drivers are fine.
Thanks for helping to combat the simple FUD that a cMP can't fully utilize a high end GPU.
While there certainly are applications where the CPU is a bottleneck, and perhaps many games are in that category, the softwares using CUDA GPGPU acceleration often are able to saturate the GPUs without putting great load on the CPUs.
Even PCIe bus issues with the older systems can be unimportant for GPGPU apps that spend most of their time crunching on stuff in VRAM, and not so much time moving data between system RAM and VRAM. (And Pascal's unified memory helps with that issue.)
Last edited:
Answered your question in another thread where you asked the same question:
This is simply not true. There are plenty of "pro" workloads like DaVinci Resolve where the GPU will be crunching away on the data with no interaction with the CPU (or PCIe bus). For cases like that, you want the biggest GPU you can get your hands on.
This is simply not true. There are plenty of "pro" workloads like ML training where the GPU will be crunching away on the data with no interaction with the CPU (or PCIe bus). For cases like that, you want the biggest GPU with the most VRAM that you can get your hands on.
ML training workstation:
24 CPU cores, 512 GiB CPU RAM, 14,336 CUDA cores, 48 GiB VRAM.
GPUs pegged, CPUs mostly idle.
Last edited:
The OP posted his workload. Will he not be CPU bottlenecked?
He will be. Simple as it can be.
Why would I talk about a situation, that is out of context of the thread? In his context HE WILL BE BOTTLENECKED REGARDLESS OF WORKLOAD HE WILL PUT ON THE GPU.
If the OP wants, and can afford spending 150$ more, for 10% more performance vs GTX 1070, its up to him to decide. If it would be my money, I would go for either GTX 980 Ti, or GTX 1070. More likely GTX 1070, because it will be much easier to set up in MP 5.1, because there are models that have only 6 pin connector(I think Founders Edition has 6 pin connector, but I can be wrong on this).
On the interwebs, however, few people take the time to read the entire thread so that a simple declarative statement is often taken at face value, rather than being understood in the context of all that went before. It is useful for everyone to repeat a bit of the context before a simple, abosolute statement.
And note that in post #7 the OP says that he uses DaVinci Resolve, and in post #31 it was stated that DaVinci Resolve is GPU bound.
So, before accusing others of not following the context, you should pay attention to the context.
OP said about specific Workload in DaVinci resolve. Will he be in that workload GPU bound 100%?
No. So in his case, he will sometimes rely on CPU performance, and therefore - bottlenecked by it.
Please read this test, I believe you won't be bottlenecked too much. Those readings are very optimistic.
https://forums.macrumors.com/threads/webdriver-for-gtx-1080-1070.1979778/page-10#post-24472142
But if sometimes she is not bound by CPU, then the 1080 could be a better choice.
So, the answer depends on what "sometimes" means. If she's constrained by CPU 90% of the time, then a cheaper GPU might be more cost effective. If 90% of the time the GPU is her bottlenect, then she should get the fastest GPU that she can find.
Your Nvidia hate blinds you to looking at the true situation and the need to balance the pros and cons.
Right, because I am saying that in his/her particular scenario GTX 1080 will be bottlenecked It shows my hate for Nvidia. Aiden. I guess it depends on what YOU want to see in my posts.
I know all of what you are writing. But think about a little about the particular scenario. Will it be bottlenecked? Yes. 90% of the time. Is it worth to pay 150$ more for those situations? No. End of the story. I do not know, how the f*** you can come up with idea that I am hating Nvidia this way. Maybe its not me who is blind?
4K video editing and DaVinci Resolve are definitely GPU-limited tasks, even on a cMP. If he was planning on just playing games, I'd definitely agree with you that he would be CPU limited for the majority of cases. I'm not sure why you think 4K video editing with Adobe products that use CUDA would be CPU limited?
Hi Frankdesign,
How does the 1080 seem to be running? Do you have any tests to share? Also, I am just curious, how did you power the card on your 5,1?
Thanks!
Rob