XFX Suxkxs

[soccerballtux]

Factory refurbed 4890 that scores 8200 in Vantage on default settings (my overclocked 4850 scores the exact same).

They said no fault is found.
So now I have a worthless 4890 that's 20% slower than what I paid for.
Spread the word please. I'm never buying from XFX again. This was my first purchase from them.

 

[happy medium]

How does it game?

 

[sonoma1993]

what steps did you take to try to solve this issue? did you clean out the old drivers? install the latest drivers from ati, clean install of windows? system specs?

 

[Quiksilver]

.....whine more about a benchmark score as if those ever really mattered in the real world performance.

Unless your 4850 was performing better than your 4890 in real gaming tests then you can complain but who gives a damn about synthetics.

 

[Magusigne]

Join the club. My RMA to them went terrible.

3 Weeks and they say no fault found.

I suppose it could be driver based but I really doubt it as my GTX 260 performed fine..anyways...

 

[Magusigne]

.....whine more about a benchmark score as if those ever really mattered in the real world performance.

Unless your 4850 was performing better than your 4890 in real gaming tests then you can complain but who gives a damn about synthetics.

Seriously?

Generally there is a very good correlation to bechmark performance and gaming performance.

Why the f' do you even think they make benchmarks.

(Granted I'll give you various benchmark have chipset bia's)

 

[konakona]

.....whine more about a benchmark score as if those ever really mattered in the real world performance.

Unless your 4850 was performing better than your 4890 in real gaming tests then you can complain but who gives a damn about synthetics.

I don't think you are very well informed about the horror stories of the infamous ZDFC XFX 4890s of later revisions. They are louder, oc a lot worse, bench worse in 3dmark AND perform worse in some of the actual games including FO3. I had bought a 4890 from TD only to find out they didn't specify it was the crappier version.

Google for ZDFC and you will find a few threads on this.

 

[SSChevy2001]

Edit

There does look to be issues with some of the XFX 4890 cards. Still it's strange that no one has found out why the card is not performing right?

 

[DaveSimmons]

> (my overclocked 4850 scores the exact same).

Overclocked how much?

An intel E8400 OC to 3.6 GHz will be faster than an E8500 at stock speed. That doesn't mean the E8500 is defective.

 

[Sylvanas]

Wow is this thread serious? You are basing an RMA off a 3rd party, synthetic benchmark application and don't specify anything else you've done to determine/solve the problem.

Perhaps it's clocking down in 3D, perhaps it's a driver issue, perhaps it's a software issue, perhaps its an OS issue, perhaps it has a BIOS that has disabled cores, perhaps it has a power supply issue....the list goes on. What did you do to eliminate things like this and attempt to diagnose the problem yourself?

 

[SunnyD]

> (my overclocked 4850 scores the exact same).

Overclocked how much?

An intel E8400 OC to 3.6 GHz will be faster than an E8500 at stock speed. That doesn't mean the E8500 is defective.

This.

Unless you had physical issues with the card (Lock-ups, crashing, BSODs, overheating, artifacting, etc.) I'm just going to take the time to laugh at you now.

 

[ArchAngel777]

I am confused here. Some people have mentioned these 'crappy' 4890s. If that is the case, these must have left the referrence design and would no longer carry the 4890 spec, thus they would not be 4890s. Any links on AMD allowing a third party to deviate from the referrence design (and thus performance) while allowing this partner to market such a peiece of junk? I'll entertain the thought, but I am skeptical over these 'crappy' 4890s existing in the first place... I need evidence.

 

[Patrick Wolf]

Factory refurbed 4890 that scores 8200 in Vantage on default settings (my overclocked 4850 scores the exact same).

Uh, didn't you compare the score of the stock 4850 to the stock 4890?

 

[fffblackmage]

My RMA experience was very good overall. Returned my ZDFC v5.4 and I got another ZDFC, v1.0. No, it's not what you think. It turned out to be a reference design 4890 that had 6-pin and 8-pin PCI-E power connectors and the digital voltage regulator controller.

More details in this thread, along with several benchmarks that showed the defective 4890 performing the same as my old uber OC'ed 4830.

 

[konakona]

I am confused here. Some people have mentioned these 'crappy' 4890s. If that is the case, these must have left the referrence design and would no longer carry the 4890 spec, thus they would not be 4890s. Any links on AMD allowing a third party to deviate from the referrence design (and thus performance) while allowing this partner to market such a peiece of junk? I'll entertain the thought, but I am skeptical over these 'crappy' 4890s existing in the first place... I need evidence.

We are just debating semantics here, but is not using the digital VRMs found on the reference card and using a crappy loud cooler count as non-reference? If so, these ZDFC cards are not reference 4890s by any stretch of imagination. Speaking from first hand experience.

 

[aigomorla]

im sorry im lost in why u RMA'd a working card.

 

[ArchAngel777]

We are just debating semantics here, but is not using the digital VRMs found on the reference card and using a crappy loud cooler count as non-reference? If so, these ZDFC cards are not reference 4890s by any stretch of imagination. Speaking from first hand experience.

Fair enough on the VRM's, but how would this translate into such a huge degredation on performance?

 

[fffblackmage]

im sorry im lost in why u RMA'd a working card.

Fair enough on the VRM's, but how would this translate into such a huge degredation on performance?

I had a working non-reference design XFX 4890 too, but it performed significantly worse than a reference design 4890.

Seriously, should a 4890 on stock clocks perform nearly identical to an OC'ed 4830?

The 4890 have a few advantages over my OC'ed 4830 (800/1000):
-Higher core clocks, 850MHz on the 4890 compare to my 4830's 800MHz.
-More SPs: 800 SPs versus 640 SPs.
-Nearly double the memory bandwidth: 1GHz GDDR3 (2GHz effective) versus 975MHz GDDR5 (3.9GHz effective).

However, these advantages did not translate to a corresponding boost in framerates in the various games I used in my suite of benchmarks. Rather, the 4890 produced framerates that were sometimes identical to my overclocked 4830. Please, don't tell me there's nothing wrong with that... I don't know why the non-reference design was underperforming significantly, but the benchmarks just don't lie.

---------------------------------------------------------------------------
And since no one seemed to have bothered to look... I'll repost my results:

I did some extensive benchmarking using Resident Evil 5, Street Fighter 4, Devil May Cry 4, Lost Planet, STALKER COP, DIRT2, and Far Cry 2.

The results: the XFX 4890 (ZDFC, stock 850/975) equals my overclock 4830 1GB (800core/1000vram).

I'm not even kidding or exaggerating; the numbers are so close, it's a practical match in most cases. Let me also remind some of you that my 4830 is also a refurb.

I also ran the benchmarks on stock clocks for my 4830. Interestingly, with the ~37% bump in gpu clock, and ~11% bump on vram, I get roughly a 20% performance boost in framerates. I don't even see that kind of performance boost going from my OC'ed 4830 to the 4890...

--------------
All numbers are average framerates, settings set on max
using comp in sig: unlocked 550BE, stock cpu clock
Using Catalyst 10.2 for both cards.

(OC'ed 4830/stock 4890)

RE5 DX9 8xAA Variable bench: 50.7 / 50.0
RE5 DX9 8xAA Fixed banch: 43.3 / 43.3

SF4 0xAA: 91.76 / 95.6
SF4 8xAA: 59.22 / 61.73

DMC4 DX9 0xAA:
part1: 87 / 90
part2: 70 / 71
part3: 111 / 105
part4: 70 / 72

DMC4 DX9 8xAA:
part1: 72 / 72
part2: 55 / 56
part3: 86 / 82
part4: 53 / 55

LP DX9 0xAA:
Part1: 32.4 / 32.4
Part2: 37.3 / 36.9

LP DX9 8xAA:
Part1: 25.1 / 24.9
Part2: 37.3 / 36.9

STALKER:COP DX9 0xAA:
part1: 55.8 / 58.3
part2: 61.8 / 68.2
part3: 67.7 / 75.2
part4: 21.5 / 23.5

STALKER:COP DX10 0xAA:
part1: 52.2 / 53.9
part2: 43.1 / 45.1
part3: 45.6 / 47.6
part4: 20.9 / 23.5

DIRT2 0xAA: 45.9 / 46.6
DIRT2 8xAA: 40.0 / 39.1

FC2 DX9 0xAA: 30.09 / 31.10
FC2 DX9 8xAA: 18.69 / 18.42

(Accidentally set "very high" preset instead of the "ultra" preset for FC2 DX10, I think)
FC2 DX10 0xAA: 45.21 / 45.62

----------------------------------------------------------------------------

And here's the results from a few benchmarks with my replacement 4890 (reference design, but with 6-pin and 8-pin PCI-E power connectors):

Same comp as before, running Catalyst 10.2 for both video cards.

(stock 4830 / OC'ed 4830 / replacement stock 4890)

FC2 DX9 0xAA: 24.13 / 30.09 / 36.80
FC2 DX9 8xAA: 15.66 / 18.69 / 24.44

SF4 8xAA: 47.9 / 59.22 / 76.33

While the OC'ed 4830 was roughly 20-30% faster than stock, the replacement 4890 is roughly 50-60% faster than the stock 4830. That's more like what I had originally expected jumping from the 4830 to the 4890.

 

[ArchAngel777]

However, these advantages did not translate to a corresponding boost in framerates in the various games I used in my suite of benchmarks. Rather, the 4890 produced framerates that were sometimes identical to my overclocked 4830. Please, don't tell me there's nothing wrong with that... I don't know why the non-reference design was underperforming significantly, but the benchmarks just don't lie.

I wound't dream of telling you that there is nothing wrong with that. I am just trying to figure out what is really wrong with the cards. It would be interested to know of during these tests you were able to graph the actual clock speed. I am curious if these things were fluctuating with their clock speed. Having 20% more shaders and having them clocked higher, along with a faster memory bus does strike me as very odd and I, too, would be dissapointed and upset.

 

[fffblackmage]

I wound't dream of telling you that there is nothing wrong with that. I am just trying to figure out what is really wrong with the cards. It would be interested to know of during these tests you were able to graph the actual clock speed. I am curious if these things were fluctuating with their clock speed. Having 20% more shaders and having them clocked higher, along with a faster memory bus does strike me as very odd and I, too, would be dissapointed and upset.

That's an interesting point. I'm working off my memory here, but I don't recall seeing fluctuations in clock speeds when doing any testing. Of course, I didn't pay attention to the clock speeds all the time, but I just don't remember seeing anything unusual, otherwise, I'd certainly would have made note of it.

I might remember the GPU utilization jumping up and down a lot; I mean oddly jumping up and down excessively. I could have assumed that was normal since load changes throughout the benchmarks, unlike clock speeds. Though now that I think about it, that was probably abnormal...

 

[soccerballtux]

I don't think you are very well informed about the horror stories of the infamous ZDFC XFX 4890s of later revisions. They are louder, oc a lot worse, bench worse in 3dmark AND perform worse in some of the actual games including FO3. I had bought a 4890 from TD only to find out they didn't specify it was the crappier version.

Google for ZDFC and you will find a few threads on this.

I think I'm going to file with the BBB and see what happens. This is a lot of money that I don't want to be out.

 

[soccerballtux]

I wound't dream of telling you that there is nothing wrong with that. I am just trying to figure out what is really wrong with the cards. It would be interested to know of during these tests you were able to graph the actual clock speed. I am curious if these things were fluctuating with their clock speed. Having 20% more shaders and having them clocked higher, along with a faster memory bus does strike me as very odd and I, too, would be dissapointed and upset.

That's an interesting point. I'm working off my memory here, but I don't recall seeing fluctuations in clock speeds when doing any testing. Of course, I didn't pay attention to the clock speeds all the time, but I just don't remember seeing anything unusual, otherwise, I'd certainly would have made note of it.

I might remember the GPU utilization jumping up and down a lot; I mean oddly jumping up and down excessively. I could have assumed that was normal since load changes throughout the benchmarks, unlike clock speeds. Though now that I think about it, that was probably abnormal...

they don't fluctuate, I've had gpu-z enabled on my second monitor and watched it, it's always at 850mhz. Load changes, clocks don't, I'm at a loss as to why it sucks.

 

[soccerballtux]

im sorry im lost in why u RMA'd a working card.

that's like asking why I would complain to my dealer if I ordered a Ferrari and they sent me a Hyundai Tiburon. Lol. Aigo you know better than that.

 

[AzN]

that's like asking why I would complain to my dealer if I ordered a Ferrari and they sent me a Hyundai Tiburon. Lol. Aigo you know better than that.

So what you are saying is you didn't get a 4890? What is wrong with Tiburon? You do realize Aigo is Korean.

 

[konakona]

That's an interesting point. I'm working off my memory here, but I don't recall seeing fluctuations in clock speeds when doing any testing. Of course, I didn't pay attention to the clock speeds all the time, but I just don't remember seeing anything unusual, otherwise, I'd certainly would have made note of it.

I might remember the GPU utilization jumping up and down a lot; I mean oddly jumping up and down excessively. I could have assumed that was normal since load changes throughout the benchmarks, unlike clock speeds. Though now that I think about it, that was probably abnormal...

yeah, tis really weird.. the vast majority of people who got these particular cards had lowered performance across the board, yet no one came up with a convincing explanation as to why this is happening. strange indeed.