New 8600 GT/GTS photos + info

[Cookie Monster]

Well, looking at the 8800GTS 320mb vs 640mb, it seems that the performance difference is largely due to the frame buffer size and NOT effected by bandwidth. (Isnt the memory clock the same between the two cards?)

People seem to misunderstand what memory bandwidth is and how much performance gains you get out of it. One clear example is the X1950XTX and the X1900XTX. High bandwidth by using GDDR4 made it ~5% faster yet its bandwidth was around 29% more than the X1900XTX. You DONT need a 256bit memory interface IF the memory clock can sufficently be clocked high enough.

Most importantly however, the architecture is the most important as chizow pointed out.

Comparing a 7600GT to a 6800 ultra, the 6800ultra literally rapes the 7600GT in terms of who has the "bigger numbers". It has a much bigger bandwidth (256bit), more pipelines, vertex shaders, texture units etc. But the 7600GT is 20% faster than the 6800ultra, and much more faster in newer games. Why is that so? how can a 128bit card beat a 256bit card with AA/AF enabled at high res? think about it for a moment. This also goes to the X1650XT (128bit) vs X850XT PE (256bit).

A 8800GTS 320mb is clearly faster than the 7900GTX especially in shader intensive games and newer games. (The 640mb is in the 7950GX2 territory). Just like the 7 series, i expect the 8600GTS to be quite abit faster than the 8600GT ala 7600GT vs 7600GS.

Im thinking IMO that this card will be somewhere around the 7950GT/X1900XT ballpark, and beating it in shader intensive games and newer titles. (oblivion/CoH comes to mind)

 

[StrangerGuy]

Based on the clocks in the article, my predictions on performance:

8600GT (540/1400): Definitely faster than 7600GT, probably will be as good as 7900GS

8600GTS (675/2000): Definitely faster than 7900GS, probably will be as good as 7950GT or maybe even 7900GTX

 

[coldpower27]

Originally posted by: keysplayr2003

Originally posted by: coldpower27

Originally posted by: BernardP

Originally posted by: coldpower27
The 8600 GT is a 7600 GS replacement in terms of price but has the performance of the last generation 7600 GT.

The 8600 GTS is the true 7600 GT replacement in terms of price and will perform greater then the 7900 GT and likely somewhere in the ballpark of the 7950 GT.

Considering the 8600GT has more shaders (confusing because architecture/terminology have changed) than 7600GT, with similar clocks, I will be disappointed if the 8600GT is not significantly faster than the 7600GT. Look at the price range compared to present 7600GT pricing.

I agree with you for 8600GTS probable performance, but its price is near the 7600GT's price when it first came out.

Well, it's not directly comparable exactly to the 12 Pixel Shader and 5 Vertex Shaders of the 7600 GT, but overall I would expect the 8600 GTS to be at least a tad faster then the 7600 GT. As the last generation 7600 GS did outperform the 6600 GT but not significantly.

And yes, I should have clarified, that I am talking about the 7600 GT MSRP compared to the 8600 GT MSRP, and not what the current price has fallen too on the 7600 GT.

All of this is taking sooo damn long to come to pass, but most likely, the 8600GTS will offer performance over a 7600GT similar to what the 8800GTS offers over a 7900GTX.

7900GTX 24 pipes 8 vertex shaders 512MB 256bit 650/1600
8800GTS 96 shaders 640MB 320bit 500/1600

7600GT 12 pipes 5 vertex shaders 256MB 128bit 560/1400
8600GTS 64 shaders 256/512MB 128bit 675/2000

Actually, with these clock speeds, I think the 8600GTS will throttle a 7600GT by a significant margin and maybe even outperform a 7900GT.

Well we were talking about the 8600 GT as opposed to GTS but I agree with your assessment, for the GTS which is similar to mine, I expect performance greater then the 7900 GT, and in the ballpark of the 7950 GT. I mistyped though, sorry for the confusion.

 

[Cookie Monster]

Originally posted by: keysplayr2003

Originally posted by: bigsnyder
Do we officially know that a 8900GS or similiar coming out? No sense in holding out for something that
might not even come out since the 8800 320MB just came out.

Nothing "concrete" yet, but there is quite a lot of buzz. Just google 8900GS.

Nvidia will also show off the new 8900GS which is really one core from the 8950GX2, core clock will be set at 550Mhz and will run on the slightly slower GDDR3 memory clocked at 1600Mhz. Expect to see a 256MB and a 512MB version of these cards priced from $200-$250 range, Making it the cheapest card of the 8900 series.

The 8900 GS will be the cheapest G80-based card. The 80 nanometre based beast is clocked at 550MHz core and 1600MHZ memory. The card has the 256-bit memory controller and comes with 256 or 512MB of GDDR3 memory. It also has 96 Shaders. The 256MB version will cost $200 while 512MB incarnation will end up at around the $250 price mark.

Word around Taiwan is that the cards are basically ready and nVidia is just waiting for AMD to take the covers off their range of upcoming R600 graphics cards.

ONCE and FOR all. THAT ROADMAP IS FAKE. I still dont know why people think its true? The same chinese site had the road map for R6x0 and that was obviously fake as most of the other rumours was not inline with that roadmap/figures. Im just surprised how many people are buzzing all over something fake. Most of b3d claimed it was fake, but i guess 8900GS's specs do look nice.

Plus 8950GX2 sounds impossible with the current power consumption figures and heat. (including driver stability and overall cost)

 

[ronnn]

These cards look very interesting. I do need to get a card soon for a multimedia/light gaming machine. As of right now the small ram gts is the leading candidate.

 

[CrystalBay]

I agree with SICKBEAST... Price perfomance all the way...

 

[Zap]

Originally posted by: Aberforth
I only meant instead of spending money on these cards you can save your money and buy 8800 GTS/X in future. These are mid-range cards as you say... you won't be able to play games like Alan Wake, Unreal, Crysis at maximum settings and you will be surely dissapointed, Alan Wake for example requires a Quad Core CPU (for optimal performance) and a powerful GPU for heavy texture processing, the 8600 won't even stand a chance at max. I'm not against 8600, its still a good deal if you want to buy a new card every 6 months considering the rate at which the new games are coming.

Obviously you have a different budget* and more stringent requirements than many other people. About the only game in that list which I'm interested in is UT2007 and I'll probably be running it in 1280x1024 (monitor max) at lowish settings and be happy. Heck, I still have Quake 3 set to "simple items" even though my system gets nearly 700FPS in it. Then again, I care more about fragging people in multiplayer than I care about standing around admiring the pretty graphics.

Doesn't matter how pretty those games are, there will still be millions more people playing WoW which doesn't require anything near an 8800 series card and quadcore. Also, not even ads on Steam can get CS 1.6 players to stop playing it even though they grumble about it all the time. "Good" graphics doesn't mean "better" or "more fun" game, plus it's very subjective.

*Budget doesn't mean saving up another month or two. It may be that people simply do not want to spend more because they don't feel the extra performance needed or justified.

Personally, I would rather not spend more than around $1000 for a whole gaming rig (minus OS, monitor and peripherals), and I would rather not have the system suck up 1kW from the wall outlet either.

 

[BernardP]

Originally posted by: Zap
*Budget doesn't mean saving up another month or two. It may be that people simply do not want to spend more because they don't feel the extra performance needed or justified.

Exactly. When configuring a system for myself or others, I try to select components that are at the sweet spot in terms of price/performance. This is just before the point where diminishing returns start. That is why I have a 3700+ in my system and not a 4000+ which at the time costed 40% more for 9% higher clock. (not talking about overclocking here)

But I admit that performance tends to scale more linearly with price in graphics cards, except at the very bottom and very top of the range.

 

[crazydingo]

http://www.vr-zone.com/?i=4775

Thats less than what a 7900GTX scores. Factor in AA at 12x10 and it'll be trailing even more. In one word: disappointing.

Going with 256-bit would have given them three advantages
a) Cooling. Dont need 1GHz memory to get 30+GB per sec.
b) Price. They could use slower and cheaper memory to achieve higher memory bandwidth. (600mhz 256-bit over 1ghz 128-bit)
c) Performance. The card wouldnt be bandwidth limited especially under AA situations even at 12x10.

 

[coldpower27]

Originally posted by: crazydingo
http://www.vr-zone.com/?i=4775

Thats less than what a 7900GTX scores. Factor in AA at 12x10 and it'll be trailing even more. In one word: disappointing.

Going with 256-bit would have given them three advantages
a) Cooling. Dont need 1GHz memory to get 30+GB per sec.
b) Price. They could use slower and cheaper memory to achieve higher memory bandwidth. (600mhz 256-bit over 1ghz 128-bit)
c) Performance. The card wouldnt be bandwidth limited especially under AA situations even at 12x10.

Depends on the type of memory if cooling is an issue, they could switch to GDDR4, though the cooler looks to be sufficient this time around.

The 256Bit Interface nullifies, any cost advantage using lower speed memory would bring, Wider Interface always cost more and stay costing more, while bins of memory speed fall overtime as newer optical nodes show up.

There isn't enough data to suggest the card would be memory bandwidth limited at 12x10 + 4XAA with AF, in DX9 games, and this card isn't really designed to provide more then likely 10x7 with AA +AF in entry level DirectX10 games.

 

[lopri]

Originally posted by: keysplayr2003

Originally posted by: bigsnyder
Do we officially know that a 8900GS or similiar coming out? No sense in holding out for something that
might not even come out since the 8800 320MB just came out.

Nothing "concrete" yet, but there is quite a lot of buzz. Just google 8900GS.

Nvidia will also show off the new 8900GS which is really one core from the 8950GX2, core clock will be set at 550Mhz and will run on the slightly slower GDDR3 memory clocked at 1600Mhz. Expect to see a 256MB and a 512MB version of these cards priced from $200-$250 range, Making it the cheapest card of the 8900 series.

The 8900 GS will be the cheapest G80-based card. The 80 nanometre based beast is clocked at 550MHz core and 1600MHZ memory. The card has the 256-bit memory controller and comes with 256 or 512MB of GDDR3 memory. It also has 96 Shaders. The 256MB version will cost $200 while 512MB incarnation will end up at around the $250 price mark.

Word around Taiwan is that the cards are basically ready and nVidia is just waiting for AMD to take the covers off their range of upcoming R600 graphics cards.

This info is just in line with what I've been hearing. Basically without the competition from AMD, NV doesn't feel the need for any serious upgrade to current G80 line up. As you can see, they're merely cutting the cost thanks to the smaller node and improved yields. (But at least they give some back to the customers, but still too expensive)

I'm more interested in how much power consumption they'll be able to cut. With some tweaks on core/PCB (and of course 80nm), they should be able to reduce the 6-pin requirement from two to one on GTX model. (can't imagine GX2 with FOUR 6-pin connectors hanging off. :laugh: )

 

[BernardP]

Quote from the VR-Zone article:

On a Core 2 Extreme X6800 processor, GeForce 8600 GTS scored a pretty impressive 5,7xx pts in 3DMark06 which is slightly faster than the 7950GT while 8600 GT churned out 4,7xx pts in 3DMark06 overshadowing a 7900GS.

Considering these are midrange cards, this looks quite acceptable: The new top-of-midrange is around the previous-gen top-of-the-line. The low-power, smallish, 8600GT is in the same league as the previous-gen upper-midrange 256-bit 7900GS.

I am not disappointed and will probably add a 8600GT to my system at some point this year, when I get a new monitor.

But it might well be that a better bang for the buck will be available with the future 8900GS, especially for those who don't want a 128-bit memory bus.

 

[R3MF]

Originally posted by: coldpower27

Originally posted by: crazydingo
http://www.vr-zone.com/?i=4775

Thats less than what a 7900GTX scores. Factor in AA at 12x10 and it'll be trailing even more. In one word: disappointing.

Going with 256-bit would have given them three advantages
a) Cooling. Dont need 1GHz memory to get 30+GB per sec.
b) Price. They could use slower and cheaper memory to achieve higher memory bandwidth. (600mhz 256-bit over 1ghz 128-bit)
c) Performance. The card wouldnt be bandwidth limited especially under AA situations even at 12x10.

Depends on the type of memory if cooling is an issue, they could switch to GDDR4, though the cooler looks to be sufficient this time around.

The 256Bit Interface nullifies, any cost advantage using lower speed memory would bring, Wider Interface always cost more and stay costing more, while bins of memory speed fall overtime as newer optical nodes show up.

There isn't enough data to suggest the card would be memory bandwidth limited at 12x10 + 4XAA with AF, in DX9 games, and this card isn't really designed to provide more then likely 10x7 with AA +AF in entry level DirectX10 games.

still would have been a better choice.

 

[GTaudiophile]

ATI's midrange will have new AVIVO features, including HDMI/HDCP capabilities built into the GPU.

If I am looking for a mid-range card, I really only need really good 2D and video features. I don't see nVidia competing very well since they will lack native HDMI/HDCP support.

 

[coldpower27]

Originally posted by: R3MF

Originally posted by: coldpower27

Originally posted by: crazydingo
http://www.vr-zone.com/?i=4775

Thats less than what a 7900GTX scores. Factor in AA at 12x10 and it'll be trailing even more. In one word: disappointing.

Going with 256-bit would have given them three advantages
a) Cooling. Dont need 1GHz memory to get 30+GB per sec.
b) Price. They could use slower and cheaper memory to achieve higher memory bandwidth. (600mhz 256-bit over 1ghz 128-bit)
c) Performance. The card wouldnt be bandwidth limited especially under AA situations even at 12x10.

Depends on the type of memory if cooling is an issue, they could switch to GDDR4, though the cooler looks to be sufficient this time around.

The 256Bit Interface nullifies, any cost advantage using lower speed memory would bring, Wider Interface always cost more and stay costing more, while bins of memory speed fall overtime as newer optical nodes show up.

There isn't enough data to suggest the card would be memory bandwidth limited at 12x10 + 4XAA with AF, in DX9 games, and this card isn't really designed to provide more then likely 10x7 with AA +AF in entry level DirectX10 games.

still would have been a better choice.

IYO perhaps, but not from a financial perspective.

 

[gramboh]

Is there a timeline on the 8900GT/GTX? I'm assuming they are waiting for AMD to drop R600. I just hope it's within 90 days of when I buy an 8800GTS 640MB from eVGA.

 

[Cookie Monster]

Originally posted by: coldpower27

Originally posted by: R3MF

Originally posted by: coldpower27

Originally posted by: crazydingo
http://www.vr-zone.com/?i=4775

Thats less than what a 7900GTX scores. Factor in AA at 12x10 and it'll be trailing even more. In one word: disappointing.

Going with 256-bit would have given them three advantages
a) Cooling. Dont need 1GHz memory to get 30+GB per sec.
b) Price. They could use slower and cheaper memory to achieve higher memory bandwidth. (600mhz 256-bit over 1ghz 128-bit)
c) Performance. The card wouldnt be bandwidth limited especially under AA situations even at 12x10.

Depends on the type of memory if cooling is an issue, they could switch to GDDR4, though the cooler looks to be sufficient this time around.

The 256Bit Interface nullifies, any cost advantage using lower speed memory would bring, Wider Interface always cost more and stay costing more, while bins of memory speed fall overtime as newer optical nodes show up.

There isn't enough data to suggest the card would be memory bandwidth limited at 12x10 + 4XAA with AF, in DX9 games, and this card isn't really designed to provide more then likely 10x7 with AA +AF in entry level DirectX10 games.

still would have been a better choice.

IYO perhaps, but not from a financial perspective.

Nor from a performance prespective. The gains of going 256bit doesnt yield enough performance increases to fully justify using such more expensive memory interface. This is seen by several examples, like this one.

7600GTvs X1800GTO, they performed similarly but which do you think was cheaper to make?

 

[coldpower27]

Originally posted by: Cookie Monster

Originally posted by: coldpower27

Originally posted by: R3MF

Originally posted by: coldpower27

Originally posted by: crazydingo
http://www.vr-zone.com/?i=4775

Thats less than what a 7900GTX scores. Factor in AA at 12x10 and it'll be trailing even more. In one word: disappointing.

Going with 256-bit would have given them three advantages
a) Cooling. Dont need 1GHz memory to get 30+GB per sec.
b) Price. They could use slower and cheaper memory to achieve higher memory bandwidth. (600mhz 256-bit over 1ghz 128-bit)
c) Performance. The card wouldnt be bandwidth limited especially under AA situations even at 12x10.

Depends on the type of memory if cooling is an issue, they could switch to GDDR4, though the cooler looks to be sufficient this time around.

The 256Bit Interface nullifies, any cost advantage using lower speed memory would bring, Wider Interface always cost more and stay costing more, while bins of memory speed fall overtime as newer optical nodes show up.

There isn't enough data to suggest the card would be memory bandwidth limited at 12x10 + 4XAA with AF, in DX9 games, and this card isn't really designed to provide more then likely 10x7 with AA +AF in entry level DirectX10 games.

still would have been a better choice.

IYO perhaps, but not from a financial perspective.

Nor from a performance prespective. The gains of going 256bit doesnt yield enough performance increases to fully justify using such more expensive memory interface. This is seen by several examples, like this one.

7600GTvs X1800GTO, they performed similarly but which do you think was cheaper to make?

As long as you mean from a performance to cost ratio then yes I agree, from a purely performance perspective then adding a 256Bit Interface would indeed produce gains as long as the overall bandwidth is increased which is sustained if the memory clock rates is
greater then half the one on the 128Bit Card.

 

[Cookie Monster]

Performance to cost ratio of course.

 

[bigsnyder]

What kind of time frame on seeing some legitimate benchmarks? Has a hard (or even soft) launch date been set?

 

[Sniper82]

Originally posted by: bigsnyder
What kind of time frame on seeing some legitimate benchmarks? Has a hard (or even soft) launch date been set?

Last I heard April 17th(give or take a day or two). So if true I would look for benchs around that time.

 

[chizow]

Originally posted by: Sniper82

Originally posted by: bigsnyder
What kind of time frame on seeing some legitimate benchmarks? Has a hard (or even soft) launch date been set?

Last I heard April 17th(give or take a day or two). So if true I would look for benchs around that time.

Ya that's the launch date (hard) I heard as well with Intel's cuts/launch on the 22nd. I'd expect demo models to be around CeBit this weekend with review samples sent out around April 1st and the NDA lifted with hard launch on April 17th.

 

[BernardP]

Hardware.fr has just released some benchmarks for the 8600 / 8500 series:

Look on this page

Original French:

Pour 3DMark 06, sur une même plateforme, dans des conditions où une 8800 GTX obtiendrait 11900 points, une GeForce 8600 GTS obtiendrait 7000 points soit le même score qu'une Radeon X1950 XTX. La GeForce 8600 GT avec 5800 points serait légèrement plus performante que la Radeon X1950 Pro et que la GeForce 7950 GT alors que la GeForce 8500 GT (2800 points) serait moins performante qu'une Radeon X1650 Pro.

3DMark 06 summary for the same system:

8800 GTX: 11900
8600 GTS: 7000 = X1950 XTX
8600 GT: 5800 > X1950 Pro and 7950 GT
8500 GT: 2800 < X1650 Pro

 

[Gstanfor]

GeForce 8800 Ultra to Come Out on April 17

GeForce 8800 Ultra to Come Out on April 17.
Nvidia?s Response to ATI R600 Almost Ready

Category: Video

by Anna Filatova

[ 03/16/2007 | 10:12 PM ]

The new ForceWare drivers have already disclosed Nvidia?s plans regarding the launch of the new product on G80 chipset that is known as GeForce 8800 Ultra. While the forum members were guessing if it would be a slower card than the GeForce 8800 GTS or an overclocked version of the GeForce 8800 GTX, NGOHQ site reported that according to some Nvidia?s partners GeForce 8800 Ultra would be launched together with the GeForce 8600, GeForce 8500, GeForce 8400 and GeForce 8300. In other words, on April 17, 2007.

GeForce 8800 Ultra will really be Nvidia?s response to ATI R600, however the specifications of the new cards haven?t been revealed yet. Most sources, however, believe that GeForce 8800 Ultra will run at higher clock speed than GeForce 8800 GTX and may feature GDDR-4 instead of GDDR-3 memory onboard with the corresponding frequency increase. However, even today Foxconn Company is already offering an overclocked GeForce 8800 GTX with GDDR-3 memory working at 2100MHz frequency. We dare suppose that they will select fast memory like GDDR-3 for the new GeForce 8800 Ultra as well.

Some Nvidia?s partners are also claimed to have promised to introduce an AGP version of the GeForce 7900 GS and GeForce 7950 GT solutions. In other words, XFX will no longer be the only Nvidia partner to offer graphics cards like that.

Sounds great, I'll take two of them please.