Precision 7530 & Precision 7730 owner's thread

The P5200 takes a hit from 7.2GHz GDDR5 instead of 10GHz GDDR5X, but this is just getting comical.

BTW these are overclocked 1070 Max-Q scores, not 1070N scores. A 1070N graphics scores 17K+.



After repaste:


Since we already know you can't do any voltage or frequency manipulation on these Quadros due to vBIOS block, the only way to improve performance is to make it run as cool as possible.

 

@Dell-Mano_G, there is a serious performance deficit in these Quadros. If users will not be allowed to boost performance (and let's face it, these scores will also affect compute), then at the very least, Dell should release a firmware update that increases power limits across the board, for the P3200, P5200 (and likely the P4200 as well).

Honestly, a P5200 that performs worse than a GTX 1070 (but has the same number of cores as a GTX 1080) is quite nonsensical.

 

This is ridiculous. I too would expect 1080-tier performance out of the P5200, though with a notable deficit due to the memory difference as you mentioned, but this? Even a typical 1080 Max-Q hits ~18k in Fire Strike. This is a 20% difference after repaste. I would expect 17k out of a Max-Q P5200 at a minimum.

Could this be a limitation of DGFF?

This makes me want to repaste my own machine to see if it yields better performance. Probably not, it was very clearly power throttled.

 

In this case I do think it is the memory that is hurting the P5200 Max-Q the most. It has the same 110W TDP as the 1080 Max-Q and similar boost clock, but almost 30% less mem BW because of 7.2GHz GDDR5 (230GB/s) versus 10GHz GDDR5X (320GB/s).

You can see here, 1080 Max-Q and P5200 Max-Q boost to about the same:

Spoiler

Even the full P5200, with 150W TDP that allows it to boost to 1700MHz instead of 1500MHz, would still be bottlenecked quite badly by its memory and perform between 1070 and 1070 Ti instead of near 1080, I would imagine.

Lower temps directly translate to higher boost clocks and better performance due to Pascal's hardcoded temperature compensation algorithm, which will drop clocks as the card heats up even if it's not power or thermal limited.

 

Ah, I guess I hadn't considered the massive memory bandwidth difference, nor was I aware how largely that would impact performance. I always figured compute cores and clock rate were the leading indicators.

Any recommendations on thermal paste? I usually just use NT-H1 for my PC stuff, but if there's anything better I'd be willing to give it a shot and post my results.

 

Yeah GP104 is actually pretty bandwidth starved, that's why GDDR5X was tapped for the 1080, and why the 1070 Ti (same mem as 1070) performs closer to the 1070 than 1080 even though it has the nearly the same compute power as the 1080.

I've heard good things about Cooler Master MasterGel Maker Nano on laptops.

 

Hi,

I've been lurking these forums for the past few months. I ordered a 7730 yesterday so decided to register on here.

I configured one with a P4200. I'll report here with the appropriate benchmarks as soon as I have it (Dell claims that will be the first of August).

 

Also consider Thermal Grizzly Kryonaut.

 

Got the FullHD with IR camera and Windows Hello is working fine for signing into Windows 10

 

Wondering why the IR camera is not an option if you order with the 4K screen. Anyway, Windows Hello also works with the fingerprint reader.