Kaby Lake Precision pre-release discussion (5520 / 7520 / 7720)

I have a question about the difference between the Xeon E3-1505M v6 compared to the i7-7820HQ. Is the E3-1505M mostly identical to the 7820HQ except for slightly higher clock speed and support for ECC RAM? Suppose money was not a problem, would I miss out anything by chosing the Xeon?

I ask this because a quick search shows some benchmarks where the 6820HQ beats the E3-1505M v5 by some non-trivial margin, which was confusing (scroll down to "real world tests...") http://cpuboss.com/cpus/Intel-Xeon-E3-1505M-v5-vs-Intel-Core-i7-6820HQ

Notebookcheck also says the E3-1505M v6 is behind the 7820HQ http://www.notebookcheck.net/Intel-Xeon-E3-1505M-v6-Notebook-Processor.190356.0.html

Also, will there be any notable difference in performance for some kind of applications, or power consumption?

 

Full turbo speed is usually only achieved if only one core is loaded. The maximum speed is pregressively reduced as more cores are loaded.

Intel has a very useful CPU comparison tool. Rows highlighted show differences (eg support for VPro). Add/remove CPUs to suit your interest. In terms of CPU performance, the notebookcheck results provide bettery insight when there are numerous results. If there are few results then the difference may be caused by the notebook chassis.

John

 

One word: High compute calculations. Remember, these cards are designed for professional application where accuracy and consistence are far more important than raw speed. An errant pixel while playing a game is insignificant and inconsequential. Errors in computational data applications can yield false results. This precision is the reason behind ECC at double the memory.

 

I mean whether it runs at x32 (or whatever kaby lake is clocked at for 4 cores) or throttles to below turbo when under heavy loads.

 

I quote Intel's definitions:
Max turbo frequency is the maximum single core frequency at which the processor is capable of operating using Intel® Turbo Boost Technology. Frequency is measured in gigahertz (GHz), or billion cycles per second.

Processor Base Frequency describes the rate at which the processor's transistors open and close. The processor base frequency is the operating point where TDP is defined. Frequency is measured in gigahertz (GHz), or billion cycles per second.

In reality, under CPU-only loading conditions, the CPU can sustain about 3 clock steps (ie 300MHz) below the maximum speed when fully loaded but might drop to the Base Frequency under combined CPU+GPU loading. So the i7-7820HQ (rated at 3.6GHz max) should sustain 3.2 or 3.3GHz under full CPU load. If you have a dGPU then it's likely to relieve the CPU package of any heavy graphical work and avoid the CPU slowing toards the Base Frequency. (My observations are based on the 6820HQ).

John

 

I'm trying to choose between the two myself, and will most likely go with the i7. Being able to overclock and/or undervolt with throttlestop or XTU will potentially allow the i7 to clock higher than the Xeon, while consuming less power.

 

Access to > 64 gb ram?

The ability to OC.

Same as with GPU. Professional vs consumer. The perfomace may be identical but the Xeon will go much further. The differnce is feature base, not necessarily about perfomance.

Like the pro version of anything from SSD, tools, vehicles, etc., they are designed for stability and reliability. The Xeon may offer many more features that you may not need or care about with your use. ECC is one example. Is it critical to your application?

 

If you do any sort of iterative computing, simulations or large data crunching, you absolutely cannot skip on ECC. Everything else is pretty much negotiable these days.

 

Sorry, but I need to disagree here. The 1080, 1070 (desktop and mobile), P5000 and P4000 all use exactly the same chip, the GP104. They just differ by firmware and probably board electronics.

There is no technical reason here that keeps nVidia from offering Pascal-based Quadro with full shader count like the 1080 has. In fact nVidia did prove so themselves when they offered the M5500M, which has all the shader core enabled like the 980 (Notebook) does. This renders your argument void.

I suspect that if AMD offers a VEGA-based FirePro with performance superior to the P5000 we will see a P5500 with full shaders enabled.

To be clear, I don't say that I blame nVidia here. The reason we don't see Pascal-based Quadro with full shaders enabled might just be that business notebook makers don't want a GPU solution with more than 100W TPD as they would need to rework their cooling solutions. Still it's disappointing that you can buy gaming notebooks with 1080 level graphics, but no business notebooks with the same gpu power.

 

Once the power and thermal aspects start to get very big then I would be thinking of desktop / server based GPUs.Alternatively, Thunderbolt connected dGPU boxes discussed in this forum here and also people are doing this with the XPS15 9550 are another option which won't adversely affect the portability of the basic workstation.

John