[fud] AMD Radeon 7770/7750 coming in February on 28nm process - specs/prices

[SirPauly]

Well that's the difference between you and I

There are a lot of differences between you an I but do agree with this:

I expect and can see why they will both price the best video card on the market, if they have it, at a premium price.

Now, that they're both pricing like predators and aggressors, one may imagine.

 

[Grooveriding]

http://translate.google.ca/translat...0-1-1.html&hl=en&biw=1747&bih=1217&prmd=imvns

7770 3DMark11 Performance setting

 

[tviceman]

I agree with this. When the 680 launches in May or June next year, expect a nice wallet rape to the tune of $600+. They don't play price war, they play squeeze you for every cent they can.

Unless the 8790 or whatever the refresh of the 7970 is going to be called is ready to release at the same time, they'll be raping everyone hard for their 20% performance advantage.

There is no doubt that if/when Nvidia trounces the hd7970 in performance they will charge a premium over it. But "raping hard" might be extreme, especially given within the context of your comment. We don't know the price, but you put a price tag of $600 on the highest end Kepler GPU, so I must ask: if the gtx680 or whatever they are calling it is 25-30% faster than the hd7970 and is $599, is that a "raping hard" price? I'm genuinely curious as to your opinion. Do you think the hd7970's price tag of $549 is "raping hard"?

Besides their halo product, everything (for the most part) Nvidia came out with below it was priced competitively, or in some instances superior to comparable AMD offerings. I really truly do not care who has the best single GPU card, because I will never pay $500 for a video card. But I do care about good competition as it will bring the bread and butter products down to reasonable prices.

 

[Skurge]

There is no doubt that if/when Nvidia trounces the hd7970 in performance they will charge a premium over it. But "raping hard" might be extreme, especially given within the context of your comment. We don't know the price, but you put a price tag of $600 on the highest end Kepler GPU, so I must ask: if the gtx680 or whatever they are calling it is 25-30% faster than the hd7970 and is $599, is that a "raping hard" price? I'm genuinely curious as to your opinion. Do you think the hd7970's price tag of $549 is "raping hard"?

Besides their halo product, everything (for the most part) Nvidia came out with below it was priced competitively, or in some instances superior to comparable AMD offerings. I really truly do not care who has the best single GPU card, because I will never pay $500 for a video card. But I do care about good competition as it will bring the bread and butter products down to reasonable prices.

Not singling you out, as groov is guilty of this too, but this is off topic and can be discussed in other 7970 thread. This is how the 1st 7000 series thread got locked.

So does anyone know how much a 6850 and a 6870 can in 3Dmark11 at those settings?

 

[RussianSensation]

So does anyone know how much a 6850 and a 6870 can in 3Dmark11 at those settings?

You have the 6870. You should throw up your score! I don't like to correlate real world gaming performance with 3dMark scores, comparing 3dMark11 scores shows that HD7770 is a dog.

Isn't it a given that they will phase out the 40nm cards? When the 6850 and 6870 showed up their initial MSRP didn't make them a compelling choice over the 5830/5850. There are some great deals to be had on the 6800s atm,

Ya, you are right. That's why anyone who wants a ~$150 GPU should just grab the 6870 now.

 

[Skurge]

You have the 6870. You should throw up your score!

lol,

Forgot I had one. I'm home for the festive season so I don't have access to my PC, I can give you 6950 scores from my brother's PC, but that would be pointless.

If they put up actual games, I'll be able to compare as I remember what I got with my 6870.

 

[notty22]

The total score is a little misleading.
Here is a forum post, I found with a slower cpu. Notice the gpu score is 3400, that's about 10% faster than the gpu score posted above.
http://www.thinkdigit.com/forum/ove...t-your-3dmark-results-here-3.html#post1359648

Picture courtesy of tom's hardware

and 3DMark 11 Benchmark by me

This image has been resized. Click this bar to view the full image. The original image is sized 1280x1024.

 

[Grooveriding]

So it looks comparable to a 6850 give or take. Is a 5770/6770 significantly slower than a 6850 or are they rather close ?

 

[notty22]

We need some of those users posting their data !

 

[Vesku]

It won't magically make up for it's 128bit bus in games that utilize high bandwidth. The core might be 28nm but I don't believe the GDDR5 chips will be drastically better OCers than the ones on a 6850.

 

[mkmitch]

I agree with this. When the 680 launches in May or June next year, expect a nice wallet rape to the tune of $600+. They don't play price war, they play squeeze you for every cent they can.

Unless the 8790 or whatever the refresh of the 7970 is going to be called is ready to release at the same time, they'll be raping everyone hard for their 20% performance advantage.

Wallet raping, rape is something you are forced to do. If you think they are too expensive don't buy one, but drop the silly analogies. Rape...good grief.

 

[edplayer]

There are some great deals to be had on the 6800s atm, I picked up a 6850 from NCIX for $120 before $20 MIR. The deals forum has a 6870 at $140ish before rebate. Heck there was even a Newegg deal on a 560 vanilla that sold out in hours.

Ya, you are right. That's why anyone who wants a ~$150 GPU should just grab the 6870 now.

Prices for videocards have been sucking for six months + (back when you could get GTX 460 for $60AR!) but out of all the choices, AMD has been having the best deals recently.


As already posted, 6870 sale prices are usually $140AR (sometimes bundled with games) and 6850 around $120AR. The lowest I have seen a 6850 for was $105AR and today there is a 6870 for $130AR:

http://www.newegg.com/Product/Produc...82E16814161389

 

[Dark Shroud]

5770/6770 is locked/gimped by it's 128bit bus. Some of the models can OC as high as 1Ghz but that can only do so much because of the damn 128bit wall that AMD used.

There was a fair sized performance gap between the 5850 and the 5770 that the 5830 really did not fill. A 5790 would have been a great mid range card.

 

[pw38]

Wallet raping, rape is something you are forced to do. If you think they are too expensive don't buy one, but drop the silly analogies. Rape...good grief.

You better hate rap then otherwise enough of the pedantic ramblings.

 

[LoneNinja]

I really don't care if these are significantly faster than the 5770/6770, they don't need to out perform the 6870 either as long as power consumption is equal to or less than a 5770. My 5770 still meets my needs, but I've got a 4670 and 5670 that really NEED replacement.

 

[Concillian]

This is what? 10-15% more SPs than a 5770 and 10-15% more bandwidth? With the same ROPs? This thing better OC to the moon or not require a 6-pin power connector to be worth buying.

7770 will still be VLIW4, right? So the ROPs won't be the much more efficient ones in the 79xx cards, right? Though the memory bandwidth is probably the biggest bottleneck.

I was pretty optimistic about 28nm a couple months ago. I think my mind has been changed.

 

[Vesku]

This is what? 10-15% more SPs than a 5770 and 10-15% more bandwidth? With the same ROPs? This thing better OC to the moon or not require a 6-pin power connector to be worth buying.

I was pretty optimistic about 28nm, this is not looking good.

Agreed on the 6-pin, hopefully the specs of the 7700 series are partly due to keeping them under the PCIe slot power ceiling.

 

[JAG87]

Can't wait for yet another wave of "where's the value" posts regarding the current fastest GPU (780), oh joy.

It depends how much faster the fastest is. If the 7970 was 70% faster than a 580, you wouldn't hear anything about value. Even if it was priced at $649, you wouldn't hear anything about value. It would just be a kick ass product.

But hmm, a card that's on average 23% faster than it's predecessor and a mere 19% faster than an almost 2 year old competitor, yea we're gonna talk about value.

You see the chart? If the GTX 280 is comparable to the HD6790 (46%). The GTX 285 is comparable to the GTX 460 768MB (50%). The 480 at 81% is an 80% (81/46) improvement over the GTX 280 and the 580 at 92% is an 84% (92/50) improvement over GTX 285. The last decent architecture from AMD was the 5870, bringing around a 63% improvement over a 4870 (which is comparable to a 5770, 41/67). After that we saw 10% from 5870 to 6970, and 23% from 6970 to 7970. Fastest or not this card sucks balls, and if you can't see that then just go put on your red shirt. The only thing that it has going for it is it's low power consumption, but apart from that you'd have to be dumb to buy one. Two 6870s perform better for a good $150 less, which is where this card should have been priced, and even then it wouldn't have been less disappointing performance-wise.

 

[Lonyo]

This is what? 10-15% more SPs than a 5770 and 10-15% more bandwidth? With the same ROPs? This thing better OC to the moon or not require a 6-pin power connector to be worth buying.

7770 will still be VLIW4, right? So the ROPs won't be the much more efficient ones in the 79xx cards, right? Though the memory bandwidth is probably the biggest bottleneck.

I was pretty optimistic about 28nm a couple months ago. I think my mind has been changed.

HD6970 ROPs:
[q]Tessellation also plays a factor in AMD’s other major gaming-related improvement: ROP performance. As tessellation produces many mini triangles, these triangles begin to choke the ROPs when performing MSAA. Although tessellation isn’t the only reason, it certainly plays a factor in AMD’s reasoning for improving their ROPs to improve MSAA performance.

The 32 ROPs (the same as Cypress) have been tweaked to speed up processing of certain types of values. In the case of both signed and unsigned normalized INT16s, these operations are now 2x faster. Meanwhile FP32 operations are now 2x to 4x faster depending on the scenario. Finally, similar to shader read ops for compute purposes, ROP write ops for graphics purposes can be coalesced, improving performance by requiring fewer operations.[/q]

So if it is VLIW4 and not GCN, it will still be somewhat improved over the ROPs in the 5770/6770, although not significantly.
It will give it a boost in some areas, but not all.

But then the 5770 wasn't exactly that great compared to the 4870 it replaced either.

 

[Vesku]

You are all over the place with this post. 7970's predecessor was the 6970 and indeed the die sizes are similar. Just as the GTX 480's predecessor was the GTX 280 (285*) and not the Radeon 5870. SLI and Crossfire bring extra baggage so comparing value linearly with a single GPU product is subjective.

Now in terms of pricing, AMD is pricing against the 580. NVIDIA will be pricing the 780 against the 7970. The price of the 780 will in large part depend on when the 780 shows up and how NVIDIA responds to the 7970 with 580 pricing between Jan. 9 and 780 launch (whether they force 7970 price down).

In relation to the topic, I think the 7700 series will take a bit of time for prices to drop. I expect this generation NVIDIA and AMD will slug it out in the 7800 <-> 760 category.

It depends how much faster the fastest is. If the 7970 was 70&#37; faster than a 580, you wouldn't hear anything about value. Even if it was priced at $649, you wouldn't hear anything about value. It would just be a kick ass product.

*image removed to save space*

But hmm, a card that's on average 23% faster than it's predecessor and a mere 19% faster than an almost 2 year old competitor, yea we're gonna talk about value.

You see the chart? If the GTX 280 is comparable to the HD6790 (46%). The GTX 285 is comparable to the GTX 460 768MB (50%). The 480 at 81% is an 80% (81/46) improvement over the GTX 280 and the 580 at 92% is an 84% (92/50) improvement over GTX 285. The last decent architecture from AMD was the 5870, bringing around a 63% improvement over a 4870 (which is comparable to a 5770, 41/67). After that we saw 10% from 5870 to 6970, and 23% from 6970 to 7970. Fastest or not this card sucks balls, and if you can't see that then just go put on your red shirt. The only thing that it has going for it is it's low power consumption, but apart from that you'd have to be dumb to buy one. Two 6850s perform better for a good $150 less, which is where this card should have been priced, and even then it wouldn't have been less disappointing performance-wise.

 

[Grooveriding]

You see the chart? If the GTX 280 is comparable to the HD6790 (46&#37. The GTX 285 is comparable to the GTX 460 768MB (50%). The 480 at 81% is an 80% (81/46) improvement over the GTX 280 and the 580 at 92% is an 84% (92/50) improvement over GTX 285. The last decent architecture from AMD was the 5870, bringing around a 63% improvement over a 4870 (which is comparable to a 5770, 41/67). After that we saw 10% from 5870 to 6970, and 23% from 6970 to 7970. Fastest or not this card sucks balls, and if you can't see that then just go put on your red shirt. The only thing that it has going for it is it's low power consumption, but apart from that you'd have to be dumb to buy one. Two 6870s perform better for a good $150 less, which is where this card should have been priced, and even then it wouldn't have been less disappointing performance-wise.

Your math is off. The 7970 is 30% faster than the 6970 across the board per techpowerup. The baseline is 100 for the 7970, the 6970 is 77, so you take the improvement of 23 and find what percent that is of the 77.

As far as the the 285 vs 480 you were discussing. 39% not 80%.

None of this has anything to do with the 7770 though.

 

[MrTeal]

But hmm, a card that's on average 23% faster than it's predecessor and a mere 19% faster than an almost 2 year old competitor, yea we're gonna talk about value.

You're reading the graph wrong. If the HD6970 is at 77% vs the HD7970 @ 100%, the HD7970 is 30% faster than the HD6970. The lead is larger at more common resolutions like 1080 and 1200 though.

 

[Arkadrel]

@JAG87


You forgot to do one like this:

Tada! 3D mark 11 performance ~15,063 score with a single card

925mhz ---> 1700mhz GPU core speed.


For those curious how a 5.6ghz Sandy-E + 7970 do, once overclocked alot.

(this gives a P54,725 3DMark Vantage result)


random 590 google + 3dmark11 score:
http://www.tomshardware.co.uk/forum/324778-15-3dmark-score

I just ran the 3DMark 11 test on my gtx 590 (performance preset) and 3d mark told me my score was low for systems with a similar configuration. However when i googled the gtx 590 benchmark it seemed pretty much in line with what the found, im running the latest nvidia drivers, here is my scores:
3DMark Score: P8775
.
.
.
.

other system detials: i7-2600k oced to 3.9 Ghz and 8gbs of ram

Yes above is on sub-zero cooling, lets all wait and see what it can do on water first.... and for some driver improvements, to speed up card performance abit.

 

[notty22]

@JAG87


You forgot to do one like this:

Tada! 3D mark 11 performance ~15,063 score with a single card

925mhz ---> 1700mhz GPU core speed.


For those curious how a 5.6ghz Sandy-E + 7970 do, once overclocked alot.

(this gives a P54,725 3DMark Vantage result)


random 590 google + 3dmark11 score:
http://www.tomshardware.co.uk/forum/324778-15-3dmark-score

What I take from this, is the architecture is scaling well with clock speed. Something Cayman did not do very well.
The world record is held by 4 - gtx 580's clocked at 1400mhz (they do scale already). Eventually we will see some 4-way , below freezing, overclock runs. It seems, they may be poised for the record, when that happens.

 

[wahdangun]

so what is the score for 4 GTX 580 clocked at 1,4 ghz ?