Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Skyrunner]

Well, I heard that Sapphire is the best AMD producer, and stuff...

[Thief^]

But that wasn't one of the options you gave earlier...

In all seriousness it's all the same chip, if the warranty's good it doesn't really matter which brand is on the box.

[Evaris]

Well, I heard that Sapphire is the best AMD producer, and stuff...

Sapphire is always good, but HIS and Powercolor are reliable and are sometimes better deals.

But that wasn't one of the options you gave earlier...

In all seriousness it's all the same chip, if the warranty's good it doesn't really matter which brand is on the box.

except it does when it comes to custom coolers, the RAM on the board, the power relay, layout efficiency... yeah brand does matter.

[Thief^]

The majority just use the reference design though, so have none of those different.

For those that are:
We can ignore any reliability difference because the warranty covers that
Custom cooler - so what? Doesn't affect performance
RAM - must at least match the spec of the reference design
Power relay - possibly power efficiency difference? But no brand I know advertises what power converters they use
Layout efficiency - I fail to see how this affects performance

Fair enough some brands make an "OC" type model that's got higher clocks than the reference design, but they normally cost more, so I'm assuming the cards listed are all basic designs.

[Evaris]

The majority just use the reference design though, so have none of those different.

For those that are:
We can ignore any reliability difference because the warranty covers that
Custom cooler - so what? Doesn't affect performance
RAM - must at least match the spec of the reference design
Power relay - possibly power efficiency difference? But no brand I know advertises what power converters they use
Layout efficiency - I fail to see how this affects performance

Fair enough some brands make an "OC" type model that's got higher clocks than the reference design, but they normally cost more, so I'm assuming the cards listed are all basic designs.

Powercolor and sapphire both use high end RAM and power converters. 
Layout efficiency effects latency by a small amount.
A custom cooler does effect performance on modern cards with 'boost' functions.  (R9 series and Nvidia cards)
A custom cooler also effects overclocking.
I don't know about Korea but a good amount if not most of the cards from Sapphire, HIS, Powercolor, Asus, Gigabyte, and MSI utilize custom boards here in the USA alongside their custom coolers. 

[Thief^]

If you can find any evidence (e.g. a benchmark) of a "good" brand card having higher performance than a "bad" brand card (ignoring OC models) I'd love to see it!

[Tellemurius]

If you can find any evidence (e.g. a benchmark) of a "good" brand card having higher performance than a "bad" brand card (ignoring OC models) I'd love to see it!

Custom coolers by far are much better than the reference blowers. Check out the 290x at launch, the reference design choked on heat until manufacturers were able to build their double/triple fan coolers on the platform.
Manufacturers have free reign on the designs of the cards, AMD has reference specifications but they have allowed full range of custom work allowed like switching out components for higher grade materials, different PCB bases so you can get more space or fit into a ITX build. AMD will never officially support OCing their products but that doesn't mean the manufacturers cannot market to that group of people.

Performance-wise theres really not much of a difference with fancier parts but instead you have increased reliability. You can safely push OCs at a higher speed without burning out the card or withstand being in a hot environment for a long time. Including that can also be a manufacturer's warranty, ones that trust in their products can even slap on a 5+ year warranty on their products. Its the same with motherboards, Intel and AMD provides the controller chips, everything else is free reign.

For my experience, i equipped an office with a mix of EVGA and ZOTAC nvidia cards. The ZOTACs had a higher rate of failure due to the shitty fans they used so i bought more of the EVGAs.

[Thief^]

You'll notice I already mentioned to look at the warranty. It's far more important than the brand name. In my experience reliability isn't inherent to a brand, they all use dodgy parts occasionally. But a good warranty saves you from that.

[gimlet]

Even in cards that don't change the clock speeds, there can be measurable differences due to hardware things like using GDDR4 vs GDDR5 ram (although I think everything current is using GDDR5, GDDR6 should be out in a few months).  A lot of them *can* be addressed by overclocking/tuning, but these differences will lead to some cards having more OC "headroom" in general, in specific OC areas depending on the card.

It's kind of hard to compare cards unless there's a review that specifically does benchmark them against each other because each review uses different hardware setups.  Or, not quite as good, when one site does different reviews of several brands, and hope they didn't change the test setup too much in between.   The between-brand comparisons are often called a "roundup".

Example, for R9-290's:
http://us.hardware.info/reviews/5379/amd-radeon-r9-290x-roundup-msi-and-xfx-join-the-party
The various R9-290's have different base and memory clockspeeds, but the differences in the various benchmarks don't necessarily match those speeds.

Couldn't easily find one for R9-270's, here's one for R9-270X's that includes benchmarks for a couple R9-270's: http://us.hardware.info/reviews/5003/five-amd-radeon-r9-270x-tested-asus-msi-sapphire-and-xfx

[Thief^]

That comparison is between gaming/OC/etc upclocked models, I specifically said I was ignoring those (because they are more expensive than the normal version, and a waste of money). They normally only have a few tiny percent difference in performance, for example, in this benchmark the "ASUS Radeon R9290X 4GB" is a stock config 290X (it's the only stock one), and in some benchmark's it's so close to the "OC" versions in performance that measurement error actually puts it ahead of some of them! http://us.hardware.info/reviews/5379/8/amd-radeon-r9-290x-roundup-msi-and-xfx-join-the-party-benchmarks---crysis-3

[gimlet]

I included that review, as *I* specifically stated, for the R-290s, several of which look like stock models (I only skimmed it I admit).  It's just difficult to find a comparison of *stock models alone*, you generally have to sift through a sort of all cards.

And your other statement also reaffirms my other point - that the performance differences DO NOT TRACK the OC speed differences.

[Thief^]

Actually, the OC speed differences are typically only a few percent also. It's just that that is too small to measure accurately (and therefore too small to notice).

[gimlet]

For the 290X cards - there were bigger differences in the R290 on various games.

And argh, I posted the wrong link for the R9-270 - I think I meant to link to this roundup: http://www.tomshardware.com/reviews/radeon-r9-270-benchmark-review,3769.html  because the one I posted did not include R9-270s.

[Thief^]

Sure, but again it's a comparison of different "OC" models. My point was that the "stock speed" models all perform identically between manufacturers, because they are the same gpu,and the same spec for everything else, if not the same components. The difference in other components does nothing for the performance of the card.
(Before you mention overclocking: overclocking varies by individual card anyway, it irks me no end when a review tests the overclocking headroom on a single sample and tries to draw conclusions from that...)

Separately I make the case that OC models are pointless, because they have only a few percent difference at best, for needless extra expense.

[gimlet]

I have repeatedly said to look at the (few, admittedly) stock models in the roundups.  I'm done - good luck finding your card.

Separately I make the case that OC models are pointless, because they have only a few percent difference at best, for needless extra expense.

This I agree with, because nowadays it's easy to just OC yourself.  The extra price is possibly worth it for someone who wants the (supposed) performance benefit guaranteed without going through that effort.