HIS Radeon HD 6970 2 GB

[bear jesus]

I think some people think that something has been locked out thus the switch was to swap between 2 different bios chips with different settings and not 2 bios chips that are the same.

its just 2 bios chips that you can switch via switch. what potential are you looking for, other than the obvious?

It's the switch to change it between 1536 SP's and 1920 but only once AMD releases the second bios that you have to flash yourself you know you want to make us a 1920SP bios w1zz, it's not harder than overclocking 2gb of ram into 4gb

Just so my comment is not confused for something serious, that was a joke

 

[N3M3515]

so you are saying "test opencl, even though no actual applications exist other than benchmarks?"

why benchmark something that nobody uses?

does anyone know an actual consumer application that uses opencl or dx compute? (not games)

we had folding in our reviews for a while. nobody said anything about it so i assumed people were not interested in it

What i really don't understand is the 1024x768 result, they don't say anything and messes the performance summary for all resolutions.

Also, if the test are for 300USD cards why bother to put 1024*768 and 1280*1024 ?, who buys a 300USD card to play with a 17" monitor?, imho for midrange and highend those 2 resolutions doesn't mean anything.

 

[Wshlist]

@N3M3515
I guess lower resolution results when compared with high resolution ones can give an indication of the bottlenecks and strengths since at low resolutions each component can probably run its fullest where at higher resolutions you get the RAM speed playing, and one thing waiting for another to finish.
As parts of a puzzle those things can be used to deduce information.

As for the remarks to me from W1zzard about OpenCL, you know every damn company is spending billions on developing support for it, even on handsets, from AMd through IBM to intel for their newest chips, a quarter of the damn transistors on the latest nvida and ATI cards are designed for computing, you seriously think that that's all for naught and they are all wrong?
And as I said it works on OSX and windows and linux, and you think all the software companies are wrong too?
Sure it's early day for applications, but you don't buy a new graphics card every month, at least I don't, so I take OpenCL and also DirectCompute serious as a thing to consider when purchasing a higher end card, and I imagine I'm not all alone in that.
Now that techpowerup focuses on the games is not unreasonable, but surely while doing all those test to throw in some computing test can't harm? Even if it was for just 5% of the consumers that's still millions of people.

 

[H82LUZ73]

its just 2 bios chips that you can switch via switch. what potential are you looking for, other than the obvious?

Because some cards I seen in reviews and manufactures web-page specs read 1536sp but in gpuz on some of the reviews it says 1620 to 1600 and that is after they overclock them.

Go here and the 2 side by side are the Saphire and the one on your right is the XFX and see the SP count is different on the Sappy.

http://www.overclockersclub.com/reviews/amd_hd6970_hd6950_review/6.htm

this leads me to think
1,A bios update
2,A bug in the GPUZ
3,It depends on what chip it is IE binned or not binned

 

[W1zzard]

those people used the wrong build of gpuz, the one that says "2011" for release date has no support for hd 6900 and calculates the shaders wrong. the one with the correct release date works right

 

[W1zzard]

As for the remarks to me from W1zzard about OpenCL, you know every damn company is spending billions on developing support for it, even on handsets, from AMd through IBM to intel for their newest chips, a quarter of the damn transistors on the latest nvida and ATI cards are designed for computing, you seriously think that that's all for naught and they are all wrong?
And as I said it works on OSX and windows and linux, and you think all the software companies are wrong too?
Sure it's early day for applications, but you don't buy a new graphics card every month, at least I don't, so I take OpenCL and also DirectCompute serious as a thing to consider when purchasing a higher end card, and I imagine I'm not all alone in that.
Now that techpowerup focuses on the games is not unreasonable, but surely while doing all those test to throw in some computing test can't harm? Even if it was for just 5% of the consumers that's still millions of people.

i do take it seriously. if there was an application that 5% of users use and that i could test i'd certainly look into it.

 

[N3M3515]

i do take it seriously. if there was an application that 5% of users use and that i could test i'd certainly look into it.

W1zz, any comments on what i wrote?

 

[W1zzard]

W1zz, any comments on what i wrote?

not really. dont look at those results if you dont like them. they can serve as an indicator at which fps rate you see cpu bottlenecks in a particular benchmark

 

[MxPhenom 216]

When I bought my 5870, I bought it with the intention of skipping this generation. This review now allows me to smugly affirm that I was correct to do so.

I don't know whether AMD or forum hype is to blame, but certainly everybody was expecting much more from these cards and I can't help feeling that, given the lead established over Nvidia with the 58xx series, the 6xxx series is, for the most part, a missed opportunity.

Thanks for the review Wizz.

yeah. the hype was far to high.

I kind of chuckled to myself when i checked the reviews during presentations in my Psychology class

 

[Wile E]

What i really don't understand is the 1024x768 result, they don't say anything and messes the performance summary for all resolutions.

Also, if the test are for 300USD cards why bother to put 1024*768 and 1280*1024 ?, who buys a 300USD card to play with a 17" monitor?, imho for midrange and highend those 2 resolutions doesn't mean anything.

I just look at the chart for my res. I have a 1920x1200 screen, so the only chart I actually look at in that section is the 1920x1200 relative performance chart. I completely ignore the overall chart.

That said, these results are disappointing. I was hoping for at least trading blows with the 580.

Well, that settles it. 580 will be my next card when money allows.

 

[Thatguy]

I think some people think that something has been locked out thus the switch was to swap between 2 different bios chips with different settings and not 2 bios chips that are the same.



It's the switch to change it between 1536 SP's and 1920 but only once AMD releases the second bios that you have to flash yourself you know you want to make us a 1920SP bios w1zz, it's not harder than overclocking 2gb of ram into 4gb

Just so my comment is not confused for something serious, that was a joke

putting on the tin foil hat for a moment. without a spec sheet on the hardware, hiding something like that ram size, bus width and number of SP's is actually very doable. If you had a datasheet with a complete ISA and all the registers. You might be able to figure out if they did hide SP's.

but thats from a practical perspective. yes you could in thoery disable in bios 2gb of ram and 400sp's.

in thoery, and no one would be the wiser.

 

[Steevo]

They probably do have a few extras on the die so they hit their target yield while being able to lazor cut bad sections.


If you had the exact widths of the shader sections, insulating areas, number of die traces and insulating area you might be able to figure out if and when there was anything hidden.


However, ATI faces a tough competitor, and a set of awesome cards this time around, so why hold back? I expect anywhere from 2-5% improvement through drivers and that is all.


Who wants to try the BIOS mod first?

 

[blu3flannel]

They definitely got a smack in the face when Nvidia released the 5xx series, but this is still a decent price/performance generation.

 

[Thatguy]

They probably do have a few extras on the die so they hit their target yield while being able to lazor cut bad sections.


If you had the exact widths of the shader sections, insulating areas, number of die traces and insulating area you might be able to figure out if and when there was anything hidden.


However, ATI faces a tough competitor, and a set of awesome cards this time around, so why hold back? I expect anywhere from 2-5% improvement through drivers and that is all.


Who wants to try the BIOS mod first?

one simply need look at the power difference between the 6950 and 6970 and determine one thing.

where is all that power going ? then again tsmc has proven to be rampant shitty at 40nm and it could just be a really leaky chip.

I suspect they have a new bios comming to deal with the 2 way DMA changes that are likely bottlenecking the card.

 

[Wshlist]

those people used the wrong build of gpuz, the one that says "2011" for release date has no support for hd 6900 and calculates the shaders wrong. the one with the correct release date works right

http://www.techpowerup.com/reviews/HIS/Radeon_HD_6970/32.html

Notice how the OpenCL box is not checked? So you say you never installed the OpenCL driver I guess? As I initially thought.
Thanks for clarifying.

It does start to almost seem you have some sort of hate against OpenCL now though, some personal dislike, with the rather fervent insisting it's nothing and ignoring my arguments and specifically not installing it at all and stuff.

I feel like if I spoke a kind word about assange in the presence of hillary

 

[erocker]

http://www.techpowerup.com/reviews/HIS/Radeon_HD_6970/32.html

Notice how the OpenCL box is not checked? So you say you never installed the OpenCL driver I guess? As I initially thought.
Thanks for clarifying.

It does start to almost seem you have some sort of hate against OpenCL now though, some personal dislike, with the rather fervent insisting it's nothing and ignoring my arguments and specifically not installing it at all and stuff.

I feel like if I spoke a kind word about assange in the presence of hillary

I don't get it. What OpenCL application is there that could be benchmarked? As a 5850 owner I have yet to see any application that needs it.

 

[W1zzard]

i love opencl and would love to see applications for it.

i love porn too, but didnt install it and didnt benchmark it.

stop complaining and suggest opencl applications to use

 

[bear jesus]

i love opencl and would love to see applications for it.

i love porn too, but didnt install it and didnt benchmark it.

stop complaining and suggest opencl applications to use

such an awesome example.

I agree with the suggesting opencl app's to use as I'm yet to find a need for it but would happily take a look at something available but i don't know of anything so some suggestions would give me a direction to go in to look for things.

One thing though, i thought the new folding client that would help use the power for ATI/AMD cards was supposed to be using opencl, i am far from sure though, but if it did I'm sure that would be something many ATI/AMD card owners would be interested to see.

 

[W1zzard]

folding doesn't use opencl. it hasn't been updated in ages for ATI cards. NVIDIA got the new gpu3 client, ati didnt. as mentioned before, we had folding, no feedback for it, so i dropped it. also it's not too reliable to benchmark

 

[bear jesus]

folding doesn't use opencl. it hasn't been updated in ages for ATI cards. NVIDIA got the new gpu3 client, ati didnt. as mentioned before, we had folding, no feedback for it, so i dropped it. also it's not too reliable to benchmark

No i meant as in the future client that would be coming as in the one that was said to solve the fail performance of ATI cards I thought it was either opencl or direct compute... i have no idea what I'm talking about

*edit*
Forgot to say that i agree folding is far from a reliable benchmark unless you could use the exact same work unit every single time.

 

[nt300]

Once again nice review W1zzard. just ordered my nice XFX Radeon HD 6970 2GB, can't wait to blow away zombies with it in Left 4 Dead 2 and soon to come 3 Can't beat the Price/Performance, this is priced A LOT better than the HD 5870 when it was released. I also read that AMD is going to release a driver fix that will release a massive performance increase of upto 20% plus in most if not all games. Let's see what plays out.

 

[Magikherbs]

I demand a rebench with the 10.12a hotfix

I have found that, regardless of the version (10.2, 10.4, 10.7-12) most mobo/gpu driver updates do not install properly the first time. This applies to Amd chipset/ Nvidia configs but not as frequent.
Glitchy startup/shutdowns and lower than expected scores are an easy sign of bad sb/gpu driver installs. eg. Performance Test 7.0 - Graphics 3D Simple and Medium tests score 60-70% lower. For some reason the complex and DirectX 10 test results are not affected.
Things get back to norms once I "express uninstall all Ati software"(before the reboot, I uninstall ATI Stream SDK, Microsoft Visual C++ 2010x64 and Catalyst Install manager, if still present), then re install the SB and Gpu drivers followed by DirectX June 2010.

My gut tells me the 6970 has not been given a fair shake.

Peace

Edit
Amd Catalyst 10.12a hotfix
http://support.amd.com/us/kbarticles/Pages/AMDCatalyst1012ahotfix.aspx

 

[Wile E]

I have found that, regardless of the version (10.2, 10.4, 10.7-12) most mobo/gpu driver updates do not install properly the first time. This applies to Amd chipset/ Nvidia configs but not as frequent.
Glitchy startup/shutdowns and lower than expected scores are an easy sign of bad sb/gpu driver installs. eg. Performance Test 7.0 - Graphics 3D Simple and Medium tests score 60-70% lower. For some reason the complex and DirectX 10 test results are not affected.
Things get back to norms once I "express uninstall all Ati software"(before the reboot, I uninstall ATI Stream SDK, Microsoft Visual C++ 2010x64 and Catalyst Install manager, if still present), then re install the SB and Gpu drivers followed by DirectX June 2010.

My gut tells me the 6970 has not been given a fair shake.

Peace

Edit
Amd Catalyst 10.12a hotfix
http://support.amd.com/us/kbarticles/Pages/AMDCatalyst1012ahotfix.aspx

You demand it? Don't you think REQUESTING it is a better way to ask for it? Besides, I'm willing to bet the differences are small.




On an unrelated note, I too, would like to see some OpenCL apps. Especially ones that accelerate video encoding, and don't suck at it like ATI's does.

 

[W1zzard]

i demand that there will be less comments about driver versions in vga reviews

a little bit of patience .. i might have something for you in the future to solve the "omgz teh new driverz enable teh sideport and give +456348569845% performanc0rz!1" discussions

 

[Wile E]

I was beginning to think sideport was a myth.