Overclock.net banner

1 - 9 of 9 Posts

·
Old dog, old tricks
Joined
·
9,755 Posts
Discussion Starter #1
Case in point: XFX 280X Double Dissipation Black Edition OC VS. Sapphire 280X Dual-X OC
(Update: Rx480 PowerColor vs AMD/Sapphire reference at end of post)




So I had the chance recently to examine these two cards.

XFX Specs:
1000 Mhz base clock
1080 Mhz turbo clock
1550 Mhz vRAM
6 heatpipe cooler

Sapphire Specs:
870 Mhz default clock
1020 Mhz turbo clock
1500 Mhz vRAM
4 heatpipe cooler

Right off the bat, the XFX looks more premium, feels more premium, and its specs are more premium - every single number is higher. And of course is in the bit more premium pricing range, compared to the Dual-X.
Every regular customer would strongly believe that the XFX is the better product overall.

Here is where they are wrong:
Both cards are 6-phase power design. The Dual-X is cooled passively by a solid aluminum radiator that covers the entire VRM. The XFX is cooled passively by a thin aluminum plate that covers only the MOSFETs. Sapphire's radiator has fins for better dissipation. XFX's does not.




One may argue that XFX's VRM dissipates additionally through the frame, believe me, that is not the case. I have done some experimentation on these MOSFETs, they produce so much thermal output that anything but immediate dissipation through a large finned sink under air flow, or a heatpipe, is useless.

Why does this matter:
6-phase VRM heats up way more than 12-phase (like on the Sapphire Toxic and other more expensive models). That is why VRM cooling is crucial for these cards. XFX lacks terribly in this regard. How does that affect the card?
Let's examine the BIOS:
XFX's stock BIOS TDP limit is 204W.
Sapphire's stock BIOS TDP limit is 247W.

This means that it is much more likely that the XFX will hit its TDP ceiling, than the Sapphire. This effect is doubled when we take under consideration the fact that the VRM's efficiency drops rapidly with the raise of temperature, and XFX's VRM is very poorly cooled. Sapphire's VRM cooling is also not top notch, but is miles better.
The result is that under most load scenarios, the XFX will run at its base clock of 1000Mhz, while the Sapphire will run at its turbo clock of 1020Mhz. These chips can even do load time throttling, which means that the card can drop its usage % while gaming in order to stay within TDP.

And this, folks, is how a card that is superior in every published spec, fails in engineering to the point where it runs slower.

A bonus fact is that XFX uses Elpida RAM while Sapphire uses Hynx, even though Dual-X is its cheapest model.

Conclusion: You should not take specs at face value. Even benchmarks in reputable review sites can lie, if the hardware is not tested realistically (eg. for certain cards, a test bench will produce waaaay better results than a case, and most reviews do exactly that, due to convenience).
Do your best to gather as much information about the products as possible, try to find photos of the bare PCBs and cooling components, try to find BIOS, and post all in an electronics or overclock forum, so people can tell which model is better from an engineering standpoint.
If you are reading this, chances are you are an overclocker, so the engineering quality of the part matters even more.

Peace.

UPDATE:

Here is a fresh example of an expensive custom card being much crappier than a reference card:
 

·
Matcha Soda
Joined
·
1,216 Posts
I got a gigabyte g1 gaming because its the only 970 with a dedicated different section of the cooler designed to cool the VRM. Until ofcourse they roll out the REAL cream, the Xtreme series which IMO should have been the g1. Meh.
 

·
Registered
Joined
·
2,344 Posts
Good to know... also always makes you wonder if when a company has increased clock speeds did they bin the GPUs or did they just figure the large majority will be able to handle an increased base clock. Because unless they bin the GPUs its not worth paying extra. For example sometimes the ASUS TOP cards are binned and usually cost around $30 more than the exact same unbinned card... worth the $30.

When a card just has the clock increased... do your research.
 

·
Old dog, old tricks
Joined
·
9,755 Posts
Discussion Starter #6
Here is a fresh example of an expensive custom card being much crappier than a reference card:
.
.
 

·
Registered
Joined
·
1,095 Posts

·
Old dog, old tricks
Joined
·
9,755 Posts
Discussion Starter #9
Quote:
Originally Posted by Vario View Post

Might as well cover the EVGA graphics cards that have heatpipes which don't even line up with the core and heatpipes that don't even function.

http://www.eteknix.com/evga-respond-possible-design-flaw-gtx-970-acx/
I have posted about this and other cases from Asus in other threads, but you are right, I probably should add it here. This was initially supposed to be a quick one-off post, but since we have gathered so much evidence of shady marketing over the years, why not systemize it here. Thanks for the suggestion, I'll probably go with it.
 
1 - 9 of 9 Posts
Top