AMD Rolls Out new HD 7950 BIOS, Expects Existing Customers to Look the Other Way

[SKYMTL]

The marketing factor of the boost edition is exactly the reason why Sky made this thread.

Indeed, we really have come full circle, haven't we?

[raghu78]

Indeed, we really have come full circle, haven't we?

AMD need not have applied such a high voltage of 1.25v. Users have been overclocking manually the HD 7950 to 1 Ghz at stock volts of 1.093V and even higher with slight voltage tweaking. With 1.175v users of cards like Sapphire HD 7950 950 Mhz edition are easily hitting 1150 Mhz game stable. But what this voltage of 1.25v will do is help AMD CCC overclocking when you couple it with a good cooler like Sapphire HD 7950 Vapor-X . AMD CCC limits have been increased with this new Boost edition. I would like to see you dedicate a review to overclocked performance of GTX 670, HD 7950 and GTX 660 Ti with all being custom designs. Throw in voltage tweaking and really prove that the GTX 660 Ti can compete with the big boys HD 7950 and GTX 670. lets see if you can do that

[kromosto]

Maybe not 100% related with topic but i am still looking for a decent nvidia card to replace my 5870. not that i dont like amd/ati just want a nvidia for cuda development with good top gaming capabilities but despite i agree with SKYMTL about the marketing discomfort on this every time i look for a new card i see a better offer from ati/amd. if this goes like this i think i will pass nvidia again and start to use amp from microsoft rather then cuda.

[raghu78]

The HD 7950 is a good replacement for the HD 5870. With its superior compute architecture and bandwidth HD 7950 is just the better choice for professional applications.

http://vr-zone.com/articles/amd-rade...s/16362-8.html

You can see AMD HD 7970 / HD 7970 Ghz dominating specviewperf which uses lightwave , maya and pro-e.

[SKYMTL]

AI would like to see you dedicate a review to overclocked performance of GTX 670, HD 7950 and GTX 660 Ti with all being custom designs. Throw in voltage tweaking and really prove that the GTX 660 Ti can compete with the big boys HD 7950 and GTX 670. lets see if you can do that

Unfortunately I won't. Now, before you start yelling "bias!!" from the rooftops, please hear me out.

Reference performance is a known quantity and the sample size doesn't have to be all that large to yield accurate, repeatable results. Basically, a reference version, with reference clocks should -in theory- perform similarly from one sample to another with a few percentage being taken into account for NVIDIA Boost feature.

On the other hand, doing an overclocking review introduces far too many variables into the equation. As I am sure you will agree, there is always a sample to sample clock speed variance which can become extreme in some circumstances. In order to overcome this, a large sample size has to be used in order assure that an accurate conclusion is reached. For example, I am sure you wouldn't be happy if we had put the OCing results from the PowerColor HD 7950 Boost up against that of a typical GTX 660 Ti. In order to eliminate oddities such as cards that clock higher than normal or lower than average, at least six identical samples of each card have to be acquired. With that, an average can be attained.

Remember, not all custom versions are equal either so that adds another wrinkle into the fabric.

I would love to say yes to an overclocking review but due to the above mentioned points, an accurate conclusion just can't be reached without a massive amount of samples (which I don't have). Nor would I be willing to post a review without doing the necessary legwork.

[SKYMTL]

The HD 7950 is a good replacement for the HD 5870. With its superior compute architecture and bandwidth HD 7950 is just the better choice for professional applications.

http://vr-zone.com/articles/amd-rade...s/16362-8.html

You can see AMD HD 7970 / HD 7970 Ghz dominating specviewperf which uses lightwave , maya and pro-e.

Remember, synthetic applications very rarely tell the whole story when it comes to professional applications. I'd actually call those numbers next to pointless, particularly specviewperf which throws a bunch of numbers out without any context.

API and back-end driver support is critical in OpenGL / GPGPU situations and there aren't ANY mass market products or drivers that can boast the necessary support. Anyone that has tried to use advanced features within AutoCAD or Lightwave will tell you that. Rather, FireGL, Quadro and Tesla are the only games in town right now.

[raghu78]

Unfortunately I won't. Now, before you start yelling "bias!!" from the rooftops, please hear me out.

Reference performance is a known quantity and the sample size doesn't have to be all that large to yield accurate, repeatable results. Basically, a reference version, with reference clocks should -in theory- perform similarly from one sample to another with a few percentage being taken into account for NVIDIA Boost feature.

On the other hand, doing an overclocking review introduces far too many variables into the equation. As I am sure you will agree, there is always a sample to sample clock speed variance which can become extreme in some circumstances. In order to overcome this, a large sample size has to be used in order assure that an accurate conclusion is reached. For example, I am sure you wouldn't be happy if we had put the OCing results from the PowerColor HD 7950 Boost up against that of a typical GTX 660 Ti. In order to eliminate oddities such as cards that clock higher than normal or lower than average, at least six identical samples of each card have to be acquired. With that, an average can be attained.

Remember, not all custom versions are equal either so that adds another wrinkle into the fabric.

I would love to say yes to an overclocking review but due to the above mentioned points, an accurate conclusion just can't be reached without a massive amount of samples (which I don't have). Nor would I be willing to post a review without doing the necessary legwork.

Even if you don't do a direct comparison of cards its possible to take a custom HD 7950 like Sapphire HD 7950 Vapor-X , review it and do stock and overclocked performance with voltage tweaking. But overclocked performance needs to be shown across the board in all of your tests, not just a couple of games. Its something which is definitely possible and very useful for people who value overclocked performance and want to know its OC and perf scaling potential. Obviously they would be making choices based on a number of reviews so they can average the overclocks across reviews from other sites. If this too isn't possible I guess there is nothing more to expect.

[DilTech]

Even if you don't do a direct comparison of cards its possible to take a custom HD 7950 like Sapphire HD 7950 Vapor-X , review it and do stock and overclocked performance with voltage tweaking. But overclocked performance needs to be shown across the board in all of your tests, not just a couple of games. Its something which is definitely possible and very useful for people who value overclocked performance and want to know its OC and perf scaling potential. Obviously they would be making choices based on a number of reviews so they can average the overclocks across reviews from other sites. If this too isn't possible I guess there is nothing more to expect.

If he doesn't already have one in the office, it's liable he'd have to physically go out and buy one.

Then comes his entire point; what if he gets a horrible OC'er (it happens, we ALL know it does)? The AMD crowd will be chanting to burn him at the stake for his blasphemous crime against team Red. Meanwhile, if he gets a golden child of a chip, it instead misleads people and he gets caught in a reverse cross-fire of hell-fire and brimstone by people saying most 7950's don't get anywhere near that...

THIS is why he needs 6'ish of the same exact card, so he can find the average using an identical model, and THEN test. Fact is though, you're talking about making the guy spend $1800 in video cards (NO company is going to send you SIX of the same card free), spend literally days testing, and likely then spending months getting yelled at by people in the community for it anyway. For what...an article that MAY pull in $12 in ad revenue total? It's just not logical, or profitable, to do that article. If it was, we'd see review sites all over the net doing them every time a new card came out!

I see how much you want said article to take place--but, can you now at least see the logic in why he's unable to do it presently?

[raghu78]

The new HD 7950 prices have started to reflect. In fact starting from USD 300 for the HIS Iceq HD 7950. Even the Sapphire HD 7950 Vapor-X which is out of stock is listed at USD 330. The GTX 660 Ti is a poor choice at its current prices. It needs atleast a price correction of 20 bucks.

http://www.newegg.com/Product/Produc...=1&srchInDesc=

[DilTech]

The new HD 7950 prices have started to reflect. In fact starting from USD 300 for the HIS Iceq HD 7950. Even the Sapphire HD 7950 Vapor-X which is out of stock is listed at USD 330. The GTX 660 Ti is a poor choice at its current prices. It needs atleast a price correction of 20 bucks.

http://www.newegg.com/Product/Produc...=1&srchInDesc=

Personally, I want to see the GTX 660ti at $249, with a non-ti version at $199. Then put the 670 at $329 and the 680 at $399. That's how the pricing SHOULD be this round.

[BababooeyHTJ]

Unfortunately I won't. Now, before you start yelling "bias!!" from the rooftops, please hear me out.

Reference performance is a known quantity and the sample size doesn't have to be all that large to yield accurate, repeatable results. Basically, a reference version, with reference clocks should -in theory- perform similarly from one sample to another with a few percentage being taken into account for NVIDIA Boost feature.

On the other hand, doing an overclocking review introduces far too many variables into the equation. As I am sure you will agree, there is always a sample to sample clock speed variance which can become extreme in some circumstances. In order to overcome this, a large sample size has to be used in order assure that an accurate conclusion is reached. For example, I am sure you wouldn't be happy if we had put the OCing results from the PowerColor HD 7950 Boost up against that of a typical GTX 660 Ti. In order to eliminate oddities such as cards that clock higher than normal or lower than average, at least six identical samples of each card have to be acquired. With that, an average can be attained.

Remember, not all custom versions are equal either so that adds another wrinkle into the fabric.

I would love to say yes to an overclocking review but due to the above mentioned points, an accurate conclusion just can't be reached without a massive amount of samples (which I don't have). Nor would I be willing to post a review without doing the necessary legwork.

You could just read the forums and guestimate what an average overclock is for each card. Thats what Hardocp did. To be on the safe side you could stay a little conservative on both cards. Nobody is gong to complain about 100mhz or so.

I don't see how you can debate that Tahiti has more overclocking headroom thanks to voltage control.

I see how much you want said article to take place--but, can you now at least see the logic in why he's unable to do it presently?

Hardocp has already done this review anyways. You can also look at a 7970ghz edition and take a good guess at overclocked performance. 7950 isn't much slower than 7970 clock for clock. It would take a pretty ty 7950 to not be able to hit 1100mhz. You can cry variables all that you want but at the end of the day Tahiti has voltage control and uses a low voltage to keep power consumption down. This is xtremesystems after all. I didn't think that I was on stock gaming systems.org. Unlike Kepler Tahiti also benefits from water cooling which some of us use or any aftermarket cooling for that matter. I run a 50% overclock pretty much all of the time on my two cards.

The new HD 7950 prices have started to reflect. In fact starting from USD 300 for the HIS Iceq HD 7950. Even the Sapphire HD 7950 Vapor-X which is out of stock is listed at USD 330. The GTX 660 Ti is a poor choice at its current prices. It needs atleast a price correction of 20 bucks.

http://www.newegg.com/Product/Produc...=1&srchInDesc=

Huh, clearly prices did just drop on 7950 unlike someone else's claims.

[SKYMTL]

You could just read the forums and guestimate what an average overclock is for each card. Thats what Hardocp did. To be on the safe side you could stay a little conservative on both cards. Nobody is gong to complain about 100mhz or so. .

Absolutely under no circumstance am I going to guesstimate. No. No. No. Just the thought of it is preposterous and tosses verifiable aspects directly into the wind. I'd rather not do a review at all rather than half-a$$ it just for the traffic it would generate. It would be a complete disservice to our readers and everything HWC stands for.

Without direct verification of clock speeds, how can we be assured that the overclock were stable? How do we know for certain additional voltage wasn't applied? Were all forum tests done at the same ambient temperature? The list goes on and on......

[W1zzard]

Absolutely under no circumstance am I going to guesstimate. No. No. No. Just the thought of it is preposterous and tosses verifiable aspects directly into the wind. I'd rather not do a review at all rather than half-a$$ it just for the traffic it would generate. It would be a complete disservice to our readers and everything HWC stands for.

Without direct verification of clock speeds, how can we be assured that the overclock were stable? How do we know for certain additional voltage wasn't applied? Were all forum tests done at the same ambient temperature? The list goes on and on......

+1. rely only on your own data

[SKYMTL]

Huh, clearly prices did just drop on 7950 unlike someone else's claims.

Actually, EXACTLY like some of us predicted.

The HD 7950 Boost now sits at $349 and it looks increasingly like the standard HD 7950 will receive a $30 price cut to clear out remaining inventory. Meanwhile, the HD 7870 will likely lay low at the $249 mark which leaves NVIDIA’s newest card in sole possession of a given –albeit narrow- market segment. In essence, AMD opened a gap in their product stack by pushing one part into End of Life status while unveiling an evolved SKU and let the GTX 660 Ti slide right in, neutralizing what could have been a huge threat.

http://www.hardwarecanucks.com/forum...review-19.html

[BababooeyHTJ]

Absolutely under no circumstance am I going to guesstimate. No. No. No. Just the thought of it is preposterous and tosses verifiable aspects directly into the wind. I'd rather not do a review at all rather than half-a$$ it just for the traffic it would generate. It would be a complete disservice to our readers and everything HWC stands for.

Without direct verification of clock speeds, how can we be assured that the overclock were stable? How do we know for certain additional voltage wasn't applied? Were all forum tests done at the same ambient temperature? The list goes on and on......

Yeah, its so preposterous to assume that a 7950 can hit 1050 mhz and a GTX670 hits a 1200mhz boost clock. I don't see how benching a couple of video cards while overclocked is at all preposterous. If you feel that way fine but don't look down on other sites who choose to show the results that they have been able to obtain.

My point was that the info is out there. The fact of the matter is that one card has voltage control and runs at a lower voltage than its bigger brother. Spin that how ever you want. Your bias is showing.

Actually, EXACTLY like some of us predicted.



http://www.hardwarecanucks.com/forum...review-19.html

Thats not at all what you predicted. If I have to dig up quotes I will. Your predictions were that 7970 would stay at the same price, these price cuts happened weeks ago, and that the 7950 boost edition would sit at a higher cost.

[raghu78]

Forget it SKYMTL won't accept that the HD 7950 is a good card and competes with the GTX 670. this review should give you an indication of the truth. A 1.2 Ghz HD 7950 beating 1.3 Ghz GTX 670 in majority of the tests. And the best part is there are HD 7950 users on OCN are running at 1.25 - 1.3 Ghz with watercooling and custom air cooling like Accelero Xtreme 7970 .

http://www.hardocp.com/article/2012/...vs_670_hd_7950

Cards like the Sapphire HD 7950 950 Mhz edition, Sapphire HD 7950 Vapor-X , MSI HD 7950 Twin Frozr all are carrying HD 7970 PCB designs at newegg. With voltage tweaking and 1.175v almost every card is hitting 1150 Mhz. SKYMTL wants us to believe that HD 7950 is a poor performer based on performance at 800 Mhz. Just pick the right SKUs and overclock them and they really maul the GTX 660 Ti.

[BababooeyHTJ]

Well, that sort of test does raise some good points that Skymtl did bring up. Most 7950s won't hit 1200mhz with the stock cooler. Neither of my cards did and that was with the MSI twin frozor. They'll do more than that now that I'm on water but still. Then again my none of my Kepler cards could hit 1300mhz.

Then there is the fact that a 670 can hit 1200mhz+ on the stock cooler and stay quiet, at least with a decent non-reference model. A 7950 needs really good cooling to hit 1200mhz with acceptable noise. I think that 1100mhz is a more realistic number for a stock cooler, maybe even 1050mhz.

I know that my 670 at 1238mhz performed very similarly to my 7950 at 1100mhz.

[raghu78]

Well, that sort of test does raise some good points that Skymtl did bring up. Most 7950s won't hit 1200mhz with the stock cooler. Neither of my cards did and that was with the MSI twin frozor. They'll do more than that now that I'm on water but still. Then again my none of my Kepler cards could hit 1300mhz.

Then there is the fact that a 670 can hit 1200mhz+ on the stock cooler and stay quiet, at least with a decent non-reference model. A 7950 needs really good cooling to hit 1200mhz with acceptable noise. I think that 1100mhz is a more realistic number for a stock cooler, maybe even 1050mhz.

I know that my 670 at 1238mhz performed very similarly to my 7950 at 1100mhz.

Any body who is serious about overclocking needs to only look at the best custom cooler designs like the sapphire dual x, sapphire vapor-x, his iceq, gigabyte windforce, asus direct cu ii. twin frozr is good but not as good as the others mentioned here. the reference HD 7950 cooler is good upto 1 Ghz at 1.1v. Its not capable of anything more than that. thats why with 1.25v on the HD 7950 Boost the powercolor card at HWC struggled to hit 1 Ghz. mostly you need 1.175v - 1.2v to hit 1150 Mhz and a good custom cooler.

[SKYMTL]

Thats not at all what you predicted. If I have to dig up quotes I will. Your predictions were that 7970 would stay at the same price, these price cuts happened weeks ago, and that the 7950 boost edition would sit at a higher cost.

The cuts did happen weeks ago and I provided links. Price then mysteriously went up again right before this launch. Please go back about 20 or so posts. I provided clear charts indicating that.

From my understanding, the HD 7970 hasn't received a price cut for the last little while.

If you are referring to this article, we were proven right.....

http://www.hardwarecanucks.com/news/...t-count-on-it/

....since AMD only dropped their price weeks later:

http://www.hardwarecanucks.com/news/...ee-game-promo/


The HD 7950 Boost IS at a higher cost. Yeah, we were $10 off. That's a heck of a lot closer than anyone else predicted.

$319 - $329
http://www.newegg.com/Product/Produc...82E16814131478
http://www.newegg.com/Product/Produc...82E16814202003

vs

$299
http://www.newegg.com/Product/Produc...82E16814129205
http://www.newegg.com/Product/Produc...82E16814161407

Actually, we said that the whole "Boost will be the same price as the original" argument was fud. We were ridiculed. People called us bias. And so far that has proven to be true.

[BababooeyHTJ]

Yes, prices fluctuate a bit. If you go on newegg now prices seemed to have dropped across the board from what they were a week ago.

[raghu78]

SKYMTL
the entire HD 7950 Boost episode reflects poorly on you. You started shouting saying the cards are not available so we will not review it. fair enough. then you bought a reference design Powercolor HD 7950 Boost card according to your own statements. Just to prove that the HD 7950 Boost card's cooler is useless. Everybody knew that. but you had to remind again. The HD 7950 Boost in the powercolor review is performing like a HD 7950 at 850 Mhz and not one at 925 Mhz. There could be only one reason- the power option is not set to +20% and so clocks are being throttled because of the voltage. You could have as well just waited for a week to 10 days , looked at the partner versions and then picked the Sapphire HD 7950 Vapor-X and reviewed it.

http://www.sapphiretech.com/presenta...articleID=4717

"On its standard settings, the SAPPHIRE HD 7950 Vapor-X Edition engine clock runs at 850 MHz with 3GB of the latest DDR5 memory clocked at 5000 MHz effective. When the Dual BIOS button is operated, the base clocks are the same but the PowerTune Dynamic Boost is enabled allowing the clocks to rise to 950 MHz or more on the engine. The fan profile and PowerTune limit is also changed to performance settings"


That would not be logical because the HD 7950 Vapor-X would not have any of the heat issues and you couldn't talk bad about the HD 7950 Boost. nice going. the way I see it you have already decided to show AMD products in the poorest manner possible. and you seem to be doing pretty well.

[SKYMTL]

SKYMTL

the entire HD 7950 Boost episode reflects poorly on you. You started shouting saying the cards are not available so we will not review it. fair enough. then you bought a reference design Powercolor HD 7950 Boost card according to your own statements. Just to prove that the HD 7950 Boost card's cooler is useless. Everybody knew that. but you had to remind again. The HD 7950 Boost in the powercolor review is performing like a HD 7950 at 850 Mhz and not one at 925 Mhz. There could be only one reason- the power option is not set to +20% and so clocks are being throttled because of the voltage. You could have as well just waited for a week to 10 days , looked at the partner versions and then picked the Sapphire HD 7950 Vapor-X and reviewed it.

It is quite obvious which side of the fence you sit on. However, your arguments are becoming more and more desperate it seems. Really, I laughed out loud reading this post. With you, it's either excuses, cherry picking cards / results, avoiding points made, shouting bias for no apparent reason or something else.

There's really nothing I can say to it....absolutely nothing.

Other than this proof that the card was indeed running at 925MHz as much as PowerTune allowed it to:

[raghu78]

It is quite obvious which side of the fence you sit on. However, your arguments are becoming more and more desperate it seems. Really, I laughed out loud reading this post. With you, it's either excuses, cherry picking cards / results, avoiding points made, shouting bias for no apparent reason or something else.

There's really nothing I can say to it....absolutely nothing.

Other than this proof that the card was indeed running at 925MHz as much as PowerTune allowed it to:

you still did not mention whether power option was at +20% for the entire review. You did not have a single argument for not picking sapphire HD 7950 vapor-x . also about waiting to see for a while whether other partner designs with custom coolers were out. all you had to say was bias, cherry picking cards. At newegg there are 4 reference design HD 7950 cards out of 20. the rest are custom coolers.

http://www.newegg.com/Product/Produc...=1&srchInDesc=

So you want to say to your HD 7950 Boost review is an accurate reflection of the HD 7950 products out in the market. The major brands like Sapphire, ASUS, MSI, GIGABYTE, XFX have never had a reference HD 7950 product at all. They came out from launch day with their custom cooler designs. Only Powercolor, HIS and Visiontek have reference design HD 7950. And HIS with Iceq and Powercolor with PCS+ have their own custom cooler designs. And you still want to keep picking on the ref HD 7950 card. You very well knew the outcome of the Powercolor HD 7950 Boost review. If you had picked a Sapphire HD 7950 Vapor-X or Sapphire HD 7950 950 Mhz edition you wouldn't have anything to write negatively. sorry but lots more on the web believe that you are biased or you have some grudge with AMD .

[SKYMTL]

Honestly, now you are just trolling

you still did not mention whether power option was at +20% for the entire review.

I have no idea what you are talking about. BOTH the NVIDIA and AMD cards were left at their preset Power values. Setting it to anything else would be disingenuous and would not reflect the true state of the card. Having said that, by default, within the Boost BIOS, the AMD PowerTune limits are raised in order to take into account for the additional P-states. Without that, it wouldn't actually Boost to 925MHz. So the boost BIOS was used, AMD's stated speeds were achieved and the performance increased by the 5-8% every other site has achieved. I really don't know where you are pulling this "20%" number from.

You did not have a single argument for not picking sapphire HD 7950 vapor-x . also about waiting to see for a while whether other partner designs with custom coolers were out. all you had to say was bias, cherry picking cards. At newegg there are 4 reference design HD 7950 cards out of 20. the rest are custom coolers.

In that same vein, I should have reviewed an overclocked, custom cooled GTX 660 Ti at launch. I didn't and quite conveniently, you seem to forget that. I reviewed (and BOUGHT) a card that was available at retail. I can't pick and choose what's available, nor will I intentionally stack the deck as you are stating. Sapphire's card may end up getting reviewed, maybe not but I won't be buying another, that's for sure.

Remember, you were perfectly happy with us testing the HD 7950 Boost....until small parts of the review didn't line up with your slanted view of reality.

So you want to say to your HD 7950 Boost review is an accurate reflection of the HD 7950 products out in the market.

No, I am saying that our review reflects a REFERENCE card. The only true way to get an accurate reflection of the market is to review every card available and that's not possible. However, if I would have reviewed the Sapphire card in the state it is currently in the conclusion would have been simple: great temps, (hopefully) better OCing, good value for high resolution, but ultimately overpriced for simple 1080P gaming. *waits for fanboy explosion*

The major brands like Sapphire, ASUS, MSI, GIGABYTE, XFX have never had a reference HD 7950 product at all. They came out from launch day with their custom cooler designs.

:O

http://www.asus.com/Graphics_Cards/A...es/HD79503GD5/
http://www.sapphiretech.com/presenta...n=&lid=1&leg=0
http://www.sapphiretech.com/presenta...n=&lid=1&leg=0
http://us.msi.com/product/vga/R7950-2PMD3GD5-OC.html

I'm sure there are more....

[[XC] Oj101]

This guy's getting destroyed but he keeps coming back for more