Radeon HD 7000 Revealed: AMD to Mix GCN with VLIW4 & VLIW5 Architectures

[Pantsu]

GK104 is the mid-range part. If the rumors hold true, it's a 768 shader 384-bit monster of a midrange part that should easily fly in comparison to even the GTX580. GK100 is the high-end, and it's supposedly 1024 shaders and a 512bit bus.

It's going to end up a clash of the titans for sure.

Yes, it would be interesting to see Nvidia and AMD on a level playing field if the rumors about GK104 end up true. GK104 according to estimates would be 50% larger than GF110. Tahiti is 65% larger than Cayman according to the OBR architecture slide. All in all they should be fairly even considering GF110 is around 15% faster on avarage compared to Cayman. Especially if AMD improves tessellation performance and Nvidia has more even improvement.

[ichime]

This 7970 will be a bust if it's not at least 50% faster than the GTX 580.

[DilTech]

Yes, it would be interesting to see Nvidia and AMD on a level playing field if the rumors about GK104 end up true. GK104 according to estimates would be 50% larger than GF110. Tahiti is 65% larger than Cayman according to the OBR architecture slide. All in all they should be fairly even considering GF110 is around 15% faster on avarage compared to Cayman. Especially if AMD improves tessellation performance and Nvidia has more even improvement.

In my eyes this is probably the most exciting pre-release build up since the G80 and R600 (before we found out about all the delays, when we expected it on time) because it's the first time both teams have brought new architectures to the table since that generation. As such, even though we know the rumored specifications, we have no idea about each teams architectural efficiency per unit in all honesty. We can make guesses based off that fact--but, until the release of each we can only make a slightly educated hypothesis on the subject.

I really wish the CPU battlefront was as competitive as the gpu side of things.

[Pantsu]

This 7970 will be a bust if it's not at least 50% faster than the GTX 580.

What would you base a 75% performance increase (against 6970) on, if ALU count goes from 1536 to 2048, bandwidth increases 50% and ROPs apparently stay the same? You expect a Christmas miracle perhaps?

[Zoran]

I really wish the CPU battlefront was as competitive as the gpu side of things.

Better not, or we will change 6 sockets a year with intel

[kaktus1907]

GF114 is 30% faster than GT200b.. i would assume that GK104 should end up the same..

[tbone8ty]

What will DX 11.1 bring?

[Pantsu]

What will DX 11.1 bring?

http://msdn.microsoft.com/en-us/libr...=vs.85%29.aspx

Mostly stuff for developers, no new gimmicks like tessellation for end users.

[SKYMTL]

This 7970 will be a bust if it's not at least 50% faster than the GTX 580.

Why?

All AMD has to do is price their next generation cards accordingly if they can't compete against the top bin NVIDIA offerings. As they have proven time and again: marketing success may lie in the high end but the real sales take place in the $150 - $300 price brackets. Even if NVIDIA's next gen cards blow AMD out of the water, there is still plenty of room for competitive products at lower price points.

[DilTech]

Why?

All AMD has to do is price their next generation cards accordingly if they can't compete against the top bin NVIDIA offerings. As they have proven time and again: marketing success may lie in the high end but the real sales take place in the $150 - $300 price brackets. Even if NVIDIA's next gen cards blow AMD out of the water, there is still plenty of room for competitive products at lower price points.

QFT.

Seriously, that's the one thing people don't get. AMD haven't really been trying to compete with NVidia's top bin for awhile now. Why do people keep expecting them to have the fastest performing single gpu out there? The ATi glory days of the x1950xtx and prior are done and over with, AMD have found it to be a better tactic to aim a bit lower with cheaper to produce parts, plain and simple.

[BoredByLife]

Supposed leaked slides show the GK100 as around double the GTX580 (sometimes higher, sometimes lower). That'd put it at over 60% faster than the 7970. It's suppose to be here in Q1 2012. NVidia's midrange GK104 should be about 75-80% of the GK100 in terms of overall power, that's a scary situation.

AMD's answer to NVidia's high end has generally been a dual-gpu card. Anyone heard anything about a 7990 coming? Last few cards we heard about said card while we were hearing about the rest on the way, but this time all is quiet on the front. Is this chip too big to pull off that set-up?

My post was a reaction to the post above me, putting GK100 at Q4 2012, or Q1 2013. If Nvidia comes that late, AMD will be ready for them. Even GK104 is only Q2, how do you think GK100 will be Q1?

[Dimitriman]

QFT.

Seriously, that's the one thing people don't get. AMD haven't really been trying to compete with NVidia's top bin for awhile now. Why do people keep expecting them to have the fastest performing single gpu out there? The ATi glory days of the x1950xtx and prior are done and over with, AMD have found it to be a better tactic to aim a bit lower with cheaper to produce parts, plain and simple.

You make it sound like AMD has been competing with cheap half arsed chips for a long while, it's not really like that at all, it's simply a different strategy of efficiency and most importantly, timely release.
A lot of people prefer single core solutions, but there is a lot of glory to be had by having the fastest single card on the market, whether dual chip or not...

[gurusan]

I don't think AMD's strategy is about targeting low-end and mainstream, but rather providing a scale-able product that covers all price and performance brackets. I think this isn't too far off NVIDIAs tactics, but their monolithic approach makes it more difficult.

[omega1alpha]

Also, the 5870 wasn't really innovative. It was, for all intents and purposes, a 4890 with DX11 tacked on and a higher shader count. That's what makes the 7970 interesting, because it's the first time AMD have REALLY stepped away from the architecture they used in the R600.

Good point. It always seems like people forget that. This is just my opinion, the 5870 wasn't a bad card, but it was overpriced for what you got (compared to 4800 series not Fermi). The 4870 was released at like $300 and was almost as fast as the nvidia single high end GTX 280. The 5870 released at like $370+ or something. The DX 11 felt rushed; especially tessellation performance. The DX9/10 was obviously faster, but not as much as I expected given twice the shaders. I remember the 4890 going for like $150 or something a few months after release while the 5870 was selling for over $300 just about a year ago. Those 5870 cards took forever to drop below $300 and most inventory was gone before you could get one for under $200 new. I wasn't happy with the 5800 series pricing at all. Of course, you have to factor the market overall. I'll sure have fond memories of the 4800 series, but the 5800 series just doesn't cut the mustard in my opinion. AMD has been increasing prices with newer releases. We need more competition. I sure would love to get a 7970 for $300 or lower.......

[tbone8ty]

yah GCN will be interesting for sure. and im sure the CAT driver updates for the 7000 series will bring significant performance upgrades as they mature.

7990 idle power will be <6w muahaha!

[ichime]

What would you base a 75% performance increase (against 6970) on, if ALU count goes from 1536 to 2048, bandwidth increases 50% and ROPs apparently stay the same? You expect a Christmas miracle perhaps?

Well, if it is a 75% performance increase over the 6970, then it should be about 50% over the GTX 580. I don't expect a christmas miracle, but I also don't expect just a modest increase in performance over its predecessor given that bandwidth has increased 50% along with a 30% dieshrink while having a max TDP of around 250-300w...

[SKYMTL]

You make it sound like AMD has been competing with cheap half arsed chips for a long while, it's not really like that at all, it's simply a different strategy of efficiency and most importantly, timely release.
A lot of people prefer single core solutions, but there is a lot of glory to be had by having the fastest single card on the market, whether dual chip or not...

Actually, what he was saying is that AMD is concentrating upon what they are good at: making small, versatile chips that can maximize yields on a per wafer basis.

High yields and more cores per wafer can equate a lower pricing structure through higher volume while still retaining profit margins.

Having the highest performance single core graphics card allows for bragging rights and not much else if a company is struggling to cope with low yields.

Personally, I'd love to see AMD compete one on one with NVIDIA's best. BUT....I would MUCH rather see their GPU division stay afloat and that means avoiding huge and expensive monolithic dies.

[boxleitnerb]

Btw are there numbers what marging/profit AMDs GPU division is making? I read somewhere they barely make any profit in regard to revenue, so their margin is quite low. If true, why is that?

[tajoh111]

Actually, what he was saying is that AMD is concentrating upon what they are good at: making small, versatile chips that can maximize yields on a per wafer basis.

High yields and more cores per wafer can equate a lower pricing structure through higher volume while still retaining profit margins.

Having the highest performance single core graphics card allows for bragging rights and not much else if a company is struggling to cope with low yields.

Personally, I'd love to see AMD compete one on one with NVIDIA's best. BUT....I would MUCH rather see their GPU division stay afloat and that means avoiding huge and expensive monolithic dies.

With AMD, I would rather them spend efficiently on their GPU research and development and stick to smaller efficient chips(not small but not monolithic). AMD research money is better spent on making their CPU's better because they are freakishly behind. AMD has a massive lead over intel in the GPU department and needs to catch up badly with the CPU market. Making their CPU good again, will do wonders for AMD image. I think AMD removing the ATI moniker for their graphic side was a mistake because the general public have a rather negative opinion of AMD chips. The only way for them to do this is to make good CPU's.

AMD seems unable to make much money in their graphic division which I think is why it is not the best to spend it money here. I think there best quarter in the last few years, AMD's graphic division made 33 million net profit, and this was with the 5xxx series competing against not very competitive products. I consider this a best case scenario. A 33 million dollar net profits(especially with the hundreds of millions lost during less competitive years) are not going to bail AMD out.

The only way to make AMD much more profitable into the graphic segment is to make their professional products better. The problem with this is AMD has been weak into marketing to the professional crowd and driver support is key with the pro market. The basis of professional cards and the reason for their high cost is the driver support(and continued at that), AMD doesn't have the manpower to keep this up. The only numbers I have heard put AMD driver team at a third the size of Nvidia's. Their recent problems with recent games such as skyrim and batman attest to this.

With AMD spending more money on low power APU, SoC. Something has to give and I hope AMD does it to their graphic department, rather their desktop chip side. The rumors of AMD conceding the high end desktop market is scary.

[kaktus1907]

4.3b vs 2.64b

Measured a total of 12 games, including a sub-run software, but can not play 3, a mapping error.
Name of the game is not being said, the resolution is 1920x1200 all max
12%
35%
20%
n / a
n / a
-8%
17%
31% (mapping error)
19%
-5%
33%
n / a
As 4.3b vs 3b we can estimate their own

pcinlife

[kaktus1907]

WinZip 16.5 for AMD GPUs and Fusion in support for hardware acceleration

[kaktus1907]

Sometimes you know something for a while and you can not say. Thanks to a leak of semi-accurate, we can finally say that the Radeon HD 7900 -, 7800 - and 7700-series, all three use the all-new Next-Core Graphics architecture!

The HD 7600 -, 7500 -, 7400 - and 7300-series are direct rebrand of the previous generation. The HD 7400 video cards are exact copies of the HD 6400 cards, the HD 7300 GPUs are the same as the HD 6300 graphics cards and the HD 7600 - and HD 7500-series remain unchanged and are based on the Turkish-GPU, we already encountered in the HD 6000-series. Everything from the HD 7600 series is seen down, based on the old architecture (VLIW5) and therefore only available to OEMs.

The release of the HD 7970 and HD 7950 follows on January 9, on 22 December the reviews of the HD 7970 already

bouweenpc.nl

[kaktus1907]

Power Tune parts:
the introduction of a more a more granular power control, based entirely on digital technology (Status Register?), independent of the driver and configuration files.
In the "black" state or PC into idle mode, GPU core state into the 0 watts, the fan will turn off, CFX configuration, when this technology is very attractive 7970 TDP is about 210 watts, 7950 is ... to be determined.
7970 frequency is 925 MHz (2048 Core/32 CU), the memory is 384-bit@5.5 GT / s..
7950 core frequency specifications to be determined, but the core is 1792 to determine the number of / 28 CU, memory bus is still 384-bit, but the rate of reduction for the 5GT / s (AMD's own not fully determined) level.

HD3D parts:
7900 to achieve the single-GPU multi-screen, multi-track orientation of the output stream
will be launched in February next year, a custom resolution (yeah!), pre-configuration management, the task of cross-screen technology

Video part:
UVD 3.0 generation still remains the same.
Focus here, is called VCE 7900 with the hardware video stream encoder and more, to achieve more than 1080p60 h.264 video encoding hardware to support full hardware encoding and fixed-function GPU shader supporting hybrid coding mode.
Compressed 4:2:0 color space encoding
quality to choose from a variety of compression
previously mentioned QSAD to achieve stability of steady video screen technology video acceleration, the new steady video 2.0 also supports interlaced video, and provide contrast mode about new features.

i hope custom resolution is something about downsampling

pcinlife

[kaktus1907]

i feel like doing monologue

ASUS 7970 696€



http://www.salland.eu/product/108271...7970-3gd5.html

[Lightman]

That's a steep price, let's hope it will come down after launch quickly to 350-450 region.
Ideally this card should take GTX580 price spot over and push competing products down, not create new higher price tag.