Precision 7530 & Precision 7730 owner's thread

Eh, 90% of a 1070 probably isn't happening. Even if you raised the TDP to 115W (assuming the cooling can handle it) and added +500 to the core so it runs at ~1750MHz, the memory still needs to be overclocked to 10GHz or more, which GDDR5 simply cannot do.

 

I'm not 100% sure of all of this. It's not my area of expertise. However, I would assume it's because of how switchable graphics are implemented.

I'm not sure it's the same for these precisions, but in my previous laptop (a latitude with AMD graphics) all of the dGPU output was routed through the intel graphics.
it's likely a different architecture now, since we're 4 CPU generations further and talking about a different GPU manufacturer, but nonetheless you can assume the CPU with it's GPU and the dGPU are entangled with eachother.
In a desktop, however, the GPU interfaces with a regular PCIe bus. This is much simpler and generic, a clear seperation between CPU and GPU.

 

Hello everyone, according to your comments it is clear to me that 7530 can not be done undervolting (xtu or throthledstop) neither cpu nor gpu. Does the same thing happen with the 7730?

Thanks!!!

 

As for the CPU, the core i9 is completely unlocked and can be tuned just fine with XTU.

 

And the i7 8750-8850?
my budget does not reach the i9...

 

Exactly the same versions.

 

If you run with graphics switching turned on (the default configuration), it is as you say, all dGPU output goes through the Intel GPU before hitting the display. If you run with graphics switching turned off, the Intel GPU is disabled and the dGPU drives the displays directly, giving you a simpler configuration (that uses more power because the dGPU is always on). This is implemented by switching various components on or off and a multiplixer on the motherboard that is basically an "input switch" for the eDP and display output ports, allowing them to be hooked to either GPU. The option to turn graphics switching off is only available in the 7000-series Precisions and not in the 5000-series or most other Dell systems.

 

Much as I expected. Plenty of reason why things aren't as plug&play as @Ionising_Radiation had hoped they'd be.

 

You mean, undervolting is only achievable with i9 on 7530? I thought undervolting had nothing to do with the CPU being unlocked...

 

OK, let's clear it up one by one.

CPU

• Undervolting: YES
• Under-clocking: YES
• Overclocking with unlocked multiplier: ONLY ON i9-8950HK
• Power limit unlocking: ONLY ON i9-8950HK

Discrete Quadro GPU

• Undervolting: NO
• Under-clocking: AUTOMATIC
• Overclocking: NO
• Power limit unlocking: NO

There has been little to no information about the performance/locked-down-ness of the AMD FirePro WX4100/7100.

Before my Precision, I owned a 13" Clevo machine, with the GTX 860M handily overclocked, such that it achieved easily 5000 Graphics points in 3DMark FireStrike, with a wonderful BIOS/VBIOS/EC mod package by the incredible efforts of @Prema (he probably still has the notebook).

Things aren't plug-and-play not because I expect a utopia, but because nVidia just chose to lock down the VBIOS citing DRM, hence making it a pain to turn up clocks/down voltages as the user saw fit.

AMD sells 8-core chips as 6-core, and even prices them as 6-core, giving buyers a nice surprise (though this is probably inadvertent). AMD allows cross-flashing of V64 VBIOSes onto V56 graphics cards. On the other hand, nVidia sells a GPU for $1200.

Crap, the regret is coming back. Imagine if AMD releases an 8-core Ryzen 3 3800H that smokes any Intel notebook CPU...

I was speaking purely from a theoretical basis—it's why I said 'approaching 90%'.

The GTX 1070 has 1920 (2048 for notebook) cores, whereas the P3200 has 1792 cores—93% (87.5%) of the desktop and notebook GTX 1070 respectively. At equal clocks, i.e. around 1.4 GHz on the core and 8 GHz on the memory (up 1 GHz from the P3200), the margin should certainly approach the region of 85-90%. I don't see why the memory needs to hit 10 GHz for 85-90%. This is likely to mean a power output likely around 100 W. Then we can let GPU Boost handle the thermals. The P3200 already clocks up to 1.544 GHz.

If the VBIOS tweaker is released for Coffee Lake (I sincerely, sincerely hope it is—I don't even mind paying for it), then rest assured that I will try to remove the power limit and see how far I can push the clocks for stable, maximum performance.

Of course, one may claim that practical performance is very different from theoretical performance, but why would that mean the P3200 should sit at a mere 75% of the performance of a 1070?