Hardware Intel Core i7-5930K and Core i7-5820K Review

Has there been a snafu with the game benchmarks, they're surprisingly uniform?

 

It was the same story with 5960X review; GPU limited even on lower settings with a 780 apparently.

 

What's going on with the idle power consumption graph? You've got two entries for both CPUs being tested.

 

You might need to go for some games that are a bit more CPU reliant.

 

Second to last sentence should be i7-4790K not 4970K.

 

Gah I'm still frustrated by the whole 5960X being the 8-core deal. I don't see much of a point in upgrading from one 6-core to another (coming from a 3930k). The extra cores would definitely be handy for my usage, but whether it's worth not only the cost of the CPU, but the mobo and RAM is another point entirely. Something just feels odd about upgrading the RAM, but not increasing the capacity. At current prices, anything more than 32GB would simply be monstrously expensive.

Also, seems Mod of the Month is lagging behind again Please don't let the same thing happen as last year where we ended up missing months.

 

It's the first Intel 6 core system that's not stupidly expensive. Shame it doesn't o/c particularly well so ipc isn't as good as the cheap 4 cores we are all using.

 

I should probably read the entire article, but I generally don't read the overclocking section and I wouldn't expect the pretty graphs to be explained there. Maybe add an NB by the graphs? Got more questions

 

Good performance for a single CPU, but too expensive and requires dedicated gpu. for the price I still prefear to build an ITX render farm. I keep thinking that the future of computing is in offloaded computations.

 

Same issue here. I've just moved from a Q6600 with a geforce 9300 IGP to an i5-4570 with GTX770. While it is a lovely machine, especially with a small case like the lian li PC-V353 ... I still think it's too big.

I remember learning 3D at school, 10 years ago, on a single core Athlon 64 with 1GB of RAM and a geforce 4 Ti 64MB. Learned to model / animate using low poly and then using subdiv for preview / rendering. My only mater was that rendering my 3 minutes animation for my diploma took a full month 24/7 to render, 100% CPU. I was a student back then and I had to live without the computer for a month.

The idea of having rendering nodes is primarily to offload the main rig, thus you can still work on something while the farm renders a scene (even if it renders slower than the main rig). Plus, it is expandable.

I've been looking at FPGA based "caustic" card to help with previewing on the main rig. You could thus have a fast cpu (modelers hardly use more than 1 core when modelling) with the IGP and use the FPGA for ray traced previews. Only problem actually is the price: $1500 for the dual FPGA / $800 for the single FPGA

http://santyhammer.blogspot.fr/2012/12/imaginationcaustic-graphics-r2500r2100.html

 

I've simply decided that GPU based rendering is the way forward for my work. Usually I can compress it all down to within 6GB, which means I can use a pair of Titans. They absolutely spank my CPU in terms of speed, a 30 minute render may only take a couple minutes using them, which in my book is phenomenal. The handy thing is that GPU memory is continuing to rise at a rapid rate. I wouldn't be surprised if the Maxwell Titan equivalent features 12GB much like the K6000 does now. When GPU VRAM stacks are approaching system memory quantities, you know things are changing. I for one hope this trend continues.

 

I have to question the usefulness of the gaming benchmarks. At least the older reviews used CPU bound games like skyrim and total war. Why bother test something in a manner that doesn't show the differences between the CPUs? Look at Tech Reports coverage of the 5960x for gaming benchmarks that actually highlight the differences between CPUs.

 

Awesome. So the way I understand it, running the ram at 2666MHz auto overclocks the CPU by pushing the bclk up to 125MHz so you guys had to lower the bclk back down to 100MHz? This pushes the power consumption up at idle, but not load, and there's no change in voltage?

 

16/8/4 is not the only option for multi GPU setups with the 5820K. 8/8/8/4 is also achievable, but so far when looking at the fine print of motherboard manuals it seems Gigabyte are the only ones offering 8/8/8/4 lane splitting. Neither Asrock or MSI allow it, not sure about Asus yet.

To be clear, the motherboard must support 8/8/8 with the 5820K by including additional clock gens. The only mobo's I can say for sure support this are GA-X99-UD4 and GA-X99-UD3.

 

I thought Luxrender now supports GPU acceleration? Since you're using blender, I'm surprised that you haven't looked into using Cycles. It's now incredibly proficient, featuring GPU accelerated particles etc, although SSS is still CPU I believe (not for long I imagine). To me, the boost isn't just in the actual render time, but in my workflow. Being able to alter materials in real time in the viewer is insanely efficient, it must have cut a lot of my work in half, time wise. In the same way, it works brilliantly for posing, setting up scenes and lighting as you can instantly see a decent quality representation of the final image. The best part about cycles though is how it supports simply so many GPUs. I use Titans because they have stacks of VRAM, but it works with plenty of other cards too, they even have OpenCL support making a come back. Obviously it depends on individual usage, but Cycles is fast become a serious rendering option.