AMD (frankenfire/MCC) mixed card crossfire review! *The TBG*

[wand3r3r]

Guys n gals and GPU gurus of ATF, it's time for a new topic, or at least a topic that hasn't been widely reviewed at all!

Check out the latest niche article in The Tech Buyers Guru's site. :thumbsup:
http://techbuyersguru.com/Frankenfire.php

A few things that are noteworthy, definitely spend a few minutes figuring out what the chart is showing (they are full of data).

There's some pretty massive scaling with mixed cards, it's surprising you could overclock a lesser model and via frankenfire reduce the difference.

Simply based on that article (and hypothesizing) I wouldn't hesitate to buy a 2nd (from the same gpu core at least) lesser gpu if there was a significant cost difference.

I'd also be curious what 13.10 brought. I guess you should still check if your favorite games are targeted by the framepacing drivers (isn't it still a game by game basis?).

 

[NTMBK]

Nice article, good to see some unusual topics getting coverage. If you already have the high end part, dropping in the lower end counterpart and OC'ing it seems like a decent way to get a performance boost.

 

[wand3r3r]

Exactly! The fact you can OC the lower card to nearer the higher card and gain that much in frankenfire is pretty sweet.

If I had e.g. a 7970 I'd definitely try one of those cheap $180 7950's and just OC it.

 

[Erenhardt]

Exactly! The fact you can OC the lower card to nearer the higher card and gain that much in frankenfire is pretty sweet.

If I had e.g. a 7970 I'd definitely try one of those cheap $180 7950's and just OC it.

or with a good 7870 XT

 

[SolMiester]

I dont understand their chart scaling?...0 = x1, 50 = x2...single GPU at 35 yet 2.2 x scaling at 65?

 

[wand3r3r]

or with a good 7870 XT

Oh yeah I forgot about their bad name on that one being it's a 79xx core and not the 78xx one.

I dont understand their chart scaling?...0 = x1, 50 = x2...single GPU at 35 yet 2.2 x scaling at 65?

There's a ton of info in each chart.

The scaling in orange could be completely in it's own chart but they are placed on top of each other to show several aspects about the data in one place.

 

[Face2Face]

Interesting review, thumbs up for Termie :thumbsup:

 

[SolMiester]

Oh yeah I forgot about their bad name on that one being it's a 79xx core and not the 78xx one.



There's a ton of info in each chart.

The scaling in orange could be completely in it's own chart but they are placed on top of each other to show several aspects about the data in one place.

I do understand charts....but that scaling makes no sense at all...

 

[Termie]

Interesting review, thumbs up for Termie :thumbsup:

I do understand charts....but that scaling makes no sense at all...

Read it as two separate graphs - the frames per second bar graphs are read off the left-hand side, the scaling line graphs are read off of the right-hand side.

In your example of a single HD 7850 GPU at 35fps and dual HD 7870 GPUs at 65fps, you in fact correctly read the scaling as 2.2x [EDIT: actually, the correct scaling line is 1.95x]. There is no scaling for the single HD7850 at stock speeds, as it's the slowest card on the chart. And that's the average fps. The minimum fps actually scales higher.

 

[SolMiester]

Read it as two separate graphs - the frames per second bar graphs are read off the left-hand side, the scaling line graphs are read off of the right-hand side.

In your example of a single HD 7850 GPU at 35fps and dual HD 7870 GPUs at 65fps, you in fact correctly read the scaling as 2.2x. There is no scaling for the single HD7850 at stock speeds, as it's the slowest card on the chart. And that's the average fps. The minimum fps actually scales higher.

Eh?....35 to 65 is 2.2 scaling? Tell me more!

 

[BallaTheFeared]

Not much can be gleaned from this without FCAT, sorry.

 

[VulgarDisplay]

Not much can be gleaned from this without FCAT, sorry.

Not all reviewers have are heavily sponsored by Nvidia. Too bad right?


Flame-baiting is not acceptable, keep on topic and stop antagonizing other members.

Administrator Idontcare

 

[BrightCandle]

Not much can be gleaned from this without FCAT, sorry.

Tomshardware showed just how bad frakenfire was with its review a few months ago, so it does seem FCAT results are necessary to confirm that all is well in that department as well. Because while frakenfire has in the past showed scaling when looked at critically with FCAT and fraps frametimes it has had much more significant stuttering problems than normal crossfire.

They openly admit they don't see a problem with crossfire pre frame pacing as they didn't see it at higher frame rates (ie they are in the 20% or so of people who can't tell there is a problem re Toms blind tests). While that is comforting to the minority the rest of us rely on quantitative scientific review methods shown to capture whether stuttering is an issue. So yeah I want FCAT data as well, I would have liked some FRAPS frame times as well, they stutter more when FCAT is bad so we can actually see it from those traces in a lot of cases as well. But we have neither, just an admission that its OK from guys who admit on page one they didn't have a problem with normal crossfire before.

I very much want to see this mixed card approach work and for AMD to get it right, it should bring significant benefits to customers once its all working. But just a few months ago I would have called it a disaster where you were better off not having the second card, even normal crossfire was like that in a lot of cases. FCAT is required when we are talking about AMD crossfire in any form today, especially an area where we haven't seen the data or we believe its changed due to recent driver updates. Crossfire has always done well in the FPS average numbers, its in the traces (fps and fcat) it is bad and its the traces that actually matter to perceived motion to most users.

 

[3DVagabond]

Not all reviewers have are heavily sponsored by Nvidia. Too bad right?

Touche'


Flame-baiting is not acceptable, compounding someone else's flamebait is also not acceptable.

Administrator Idontcare

 

[Jaydip]

Nice article, good to see some unusual topics getting coverage.

[redacted]

PSA - this member's post was in response to a spammer's post...please don't respond to spammers and trolls because all you do is make more work for the mods who then not only have to cleanup the original spammer's posts but must also go through the thread and cleanup all the needless off-topic posts other members (like yourself) felt compelled to add to the thread.

Administrator Idontcare

 

[Skurge]

Tomshardware showed just how bad frakenfire was with its review a few months ago, so it does seem FCAT results are necessary to confirm that all is well in that department as well. Because while frakenfire has in the past showed scaling when looked at critically with FCAT and fraps frametimes it has had much more significant stuttering problems than normal crossfire.

They openly admit they don't see a problem with crossfire pre frame pacing as they didn't see it at higher frame rates (ie they are in the 20% or so of people who can't tell there is a problem re Toms blind tests). While that is comforting to the minority the rest of us rely on quantitative scientific review methods shown to capture whether stuttering is an issue. So yeah I want FCAT data as well, I would have liked some FRAPS frame times as well, they stutter more when FCAT is bad so we can actually see it from those traces in a lot of cases as well. But we have neither, just an admission that its OK from guys who admit on page one they didn't have a problem with normal crossfire before.

I very much want to see this mixed card approach work and for AMD to get it right, it should bring significant benefits to customers once its all working. But just a few months ago I would have called it a disaster where you were better off not having the second card, even normal crossfire was like that in a lot of cases. FCAT is required when we are talking about AMD crossfire in any form today, especially an area where we haven't seen the data or we believe its changed due to recent driver updates. Crossfire has always done well in the FPS average numbers, its in the traces (fps and fcat) it is bad and its the traces that actually matter to perceived motion to most users.

That was hybrid crossfire, not mixed crossfire. The TOMS article was Trinity plus 7750.

 

[Termie]

Eh?....35 to 65 is 2.2 scaling? Tell me more!

If you specify which game and which test conditions you are referring to, I could explain more. But I think you may actually be looking at the average fps bar graph and minimum fps scaling line graph. Those naturally won't match up.

Sorry for the confusion.

 

[AntonioHG]

Not all reviewers have are heavily sponsored by Nvidia. Too bad right?

You sound so very childish. Must be a huge conspiracy to take down AMD, I mean, AMD even wants to take down AMD! Why else would they put out drivers to fix nonexistent problems?

All of this is if you were not being sarcastic, of course. Sorry if you were being sarcastic, I'm sarcasm blind.


Flaming and flame-baiting are not acceptable, compounding someone else's flamebait is also not acceptable.

Administrator Idontcare

 

[Rvenger]

Thanks everyone who is flaming. Why don't we get this thread closed now?

 

[wand3r3r]

Thanks everyone who is flaming. Why don't we get this thread closed now?

Or better still, just remove the bad post(er)s.

 

[Shmee]

I have removed the off topic/inflammatory posts. Continue on topic, and mind the rules please.

 

[KingFatty]

I have a different interpretation of this article.

To me, it just illustrates that it's not the best value to buy top-tier cards, and instead you should just buy their little brother and overclock it.

I mean, why even get a 7870 in the first place, when you can save a bunch of money by going with Geico and getting a 7850? Then when you crossfire, you can save money yet a 2nd time. Same argument for 7970 vs 7950, etc.

 

[wand3r3r]

You could also view it as buy a 7950 (perhaps a year ago) and then crossfire it with a cheap 7970 when the prices have dropped a lot.

There are many viewpoints. Either way I agree, buy a GPU and overclock it.

 

[el etro]

Tomshardware showed just how bad frakenfire was with its review a few months ago, so it does seem FCAT results are necessary to confirm that all is well in that department as well. Because while frakenfire has in the past showed scaling when looked at critically with FCAT and fraps frametimes it has had much more significant stuttering problems than normal crossfire.

They openly admit they don't see a problem with crossfire pre frame pacing as they didn't see it at higher frame rates (ie they are in the 20% or so of people who can't tell there is a problem re Toms blind tests). While that is comforting to the minority the rest of us rely on quantitative scientific review methods shown to capture whether stuttering is an issue. So yeah I want FCAT data as well, I would have liked some FRAPS frame times as well, they stutter more when FCAT is bad so we can actually see it from those traces in a lot of cases as well. But we have neither, just an admission that its OK from guys who admit on page one they didn't have a problem with normal crossfire before.

I very much want to see this mixed card approach work and for AMD to get it right, it should bring significant benefits to customers once its all working. But just a few months ago I would have called it a disaster where you were better off not having the second card, even normal crossfire was like that in a lot of cases. FCAT is required when we are talking about AMD crossfire in any form today, especially an area where we haven't seen the data or we believe its changed due to recent driver updates. Crossfire has always done well in the FPS average numbers, its in the traces (fps and fcat) it is bad and its the traces that actually matter to perceived motion to most users.

This is hybrid technology crossfire.

HybCFX is a thing that i hope someday will work, even if not so good as the dGPU Crossfires.

 

[BrightCandle]

Perhaps I am under the mistake impression that AMD consider this largely the same technology. In toms review he had a GPU on chip and was pairing it with a more powerful card and in this case we have two different performance cards both discrete. Maybe AMD call this something different but I largely consider they basically work the same and rely on the same principle of different performing parts working on alternate frames.

I want to see the FCAT traces, and new area based on AMD crossfire requires them until this whole sorry scandel is done. I really wish AMD had just fixed everything with 12.8 because I am now constantly looking at reviews and wondering every time FCAT missing. I would love to think this boost comes without compromise, but bases on the way AFR works it seems plausibly likely it stutters.