• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Review of 112sp 8800GTS 640 mb

Page 3 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
man hang in there like two more weeks 'til we know a lot more about the 128SP 65nm GTS. It'll be worth the wait to not spend $360 now!
 
lmao, i dont get it, is there a 96SP GTS Vs 112SP GTS ? i want to see the huge diffrence you people are so excited about.
 
Originally posted by: Azn
SSC whatever does it really matter? GTS can easily overclock to those levels and beyond.... Better performance in higher resolutions with AA than 8800gt even if it's older core.

Again games doesn't always scale with how many more stream processors. It's the combination. 16 more PS is not going to do anything in most situations. What matters is higher clock speed and memory bandwidth with a card like GTS.


So the new STANDARD Clocked 112sp GTS shouldnt be any faster than the original STANDARD Clocked GTS meaning the STANDARD 8800GT is FASTER.



 
Originally posted by: keysplayr2003
Originally posted by: Azn
SSC whatever does it really matter? GTS can easily overclock to those levels and beyond.... Better performance in higher resolutions with AA than 8800gt even if it's older core.

Again games doesn't always scale with how many more stream processors. It's the combination. 16 more PS is not going to do anything in most situations. What matters is higher clock speed and memory bandwidth with a card like GTS.

How do you figure? If you took away those 16 SP's and reduced the G80 to 80SP's do you think that wouldn't make a difference either? Of course 16SP's would yield better performance than not having them. Look at the GTX. 32SP's over a standard GTS is substantial in EVERY situation.

All things being equal (core/mem clocks) between a GTS640 96sp and a GTS640 112sp, It's pretty much a given that the 112sp card would be more powerful. I could see variations from game to game, and the lead may shorten or lengthen, but the lead will always be there.

If the 8800GT and the 112 GTS640 were equal in price, It's a no brainer to go with the GTS.
Unfortunately, the 112 GTS640 is still brutally expensive to manufacture and it's price is prohibitive relative to it's performance. 8800GT is the winner hands down.

I'm waiting to see how the 128sp G92 GTS compares. I know it will be faster than the 8800GT and will most likely rival or best a GTX in most situations.

It totally depends on the game. Why would anyone upgrade their GTS to the new GTS. Nvidia is replacing the old GTS with the new. Yup it's a no brainer to pickup the new over the old.
 
Originally posted by: SniperDaws
Originally posted by: Azn
SSC whatever does it really matter? GTS can easily overclock to those levels and beyond.... Better performance in higher resolutions with AA than 8800gt even if it's older core.

Again games doesn't always scale with how many more stream processors. It's the combination. 16 more PS is not going to do anything in most situations. What matters is higher clock speed and memory bandwidth with a card like GTS.


So the new STANDARD Clocked 112sp GTS shouldnt be any faster than the original STANDARD Clocked GTS meaning the STANDARD 8800GT is FASTER.

Sure average frame rate wise without AA but when you add in AA and run them in higher resolutions it's a diff. story.
 
Sorry but you make little sense JAG87.

They are different architectures so you can't compare their clock speeds. They are what they are and thats it. If one comes at 600 stock and the other at 500 or whatever, then those are each card's respective stock and if you OC one you must OC the other one to have a fair comparison. That the GTS cant run at 600 stock is not the GT's fault. Each card has a different strenght and in order to utilize the one on the GT you need the extra speed.
The GTS has more memory and a bigger bus so should we remove the extra memory and shrink the bus to make it fair comparison? Of course not because thats what that architecture is based upon.

Sorry but I have to totally disagree with your point of view. OC both cards and then we will see.
 
Originally posted by: SniperDaws
Originally posted by: Azn
SSC whatever does it really matter? GTS can easily overclock to those levels and beyond.... Better performance in higher resolutions with AA than 8800gt even if it's older core.

Again games doesn't always scale with how many more stream processors. It's the combination. 16 more PS is not going to do anything in most situations. What matters is higher clock speed and memory bandwidth with a card like GTS.


So the new STANDARD Clocked 112sp GTS shouldnt be any faster than the original STANDARD Clocked GTS meaning the STANDARD 8800GT is FASTER.

How do you come to that conclusion? You think the gtx gets all its performance just from having more memory than the gts? The g80 architecture is divided into clusters, and each cluster not only has shaders but also texture units. I would be willing to bet money that the 112 shader gts is clock for clock faster than the old gts in any situation that's not limited by the cpu or video memory.
 
Originally posted by: shaolin95
Sorry but you make little sense JAG87.

They are different architectures so you can't compare their clock speeds. They are what they are and thats it. If one comes at 600 stock and the other at 500 or whatever, then those are each card's respective stock and if you OC one you must OC the other one to have a fair comparison. That the GTS cant run at 600 stock is not the GT's fault. Each card has a different strenght and in order to utilize the one on the GT you need the extra speed.
The GTS has more memory and a bigger bus so should we remove the extra memory and shrink the bus to make it fair comparison? Of course not because thats what that architecture is based upon.

Sorry but I have to totally disagree with your point of view. OC both cards and then we will see.

i tend to agree with this logic.
 
Munky I think it really depends on the game how much Shader Operations it needs in different resolutions. I suspect newer engines require more.
 
so is the new 112sp GTS going to be sold at the same price as the original 640mb GTS? because if it is its not worth it, in the UK you can buy (when in stock) the 8800GT for £144 but the original 640mb GTS is still being sold for around £230-240 so an extra 5fps with AA on isnt worth £100.
 
Are the current batch of GTs overclocking better than the G80 based cards? That may even out the differences between these cards.

I would wait and see what the new GTS is like anyway before buying either of these, especially given the inflated prices right now.
 
im gonna stick to my guns and say it out loud-


The GTS has no place now because the GT is kicking the heals of the GTX and because the GTX still beats the GTS 112 edition, you would be very silly to buy this card now.

Nvidia should of just fazed out the GTS until the release of the proper new G92 GTS/GTX.
 
Originally posted by: shaolin95
Sorry but you make little sense JAG87.

They are different architectures so you can't compare their clock speeds. They are what they are and thats it. If one comes at 600 stock and the other at 500 or whatever, then those are each card's respective stock and if you OC one you must OC the other one to have a fair comparison. That the GTS cant run at 600 stock is not the GT's fault. Each card has a different strenght and in order to utilize the one on the GT you need the extra speed.
The GTS has more memory and a bigger bus so should we remove the extra memory and shrink the bus to make it fair comparison? Of course not because thats what that architecture is based upon.

Sorry but I have to totally disagree with your point of view. OC both cards and then we will see.

The architectures aren't so different that you can't draw relevant conclusions from running the cards at the same clock speeds, especially in the case of the G80 GTS where much of those architectural/clock speed differences are artificially imposed in order to differentiate products where the parts are otherwise the same. While NV may have binned/tested different G80 cores, then neutered some cores to make the GTS, there's little doubt that the current differentiation is little more than artifically imposed performance cuts to meet market/product differentiation specifications.

The new G80 112SP GTS is just further proof of this. Same for clock speeds. G80 GTS and GTX have typically had the same amount of OC'ing headroom since launch, with GTS users easily achieving similar OC's as GTX owners, but the actual increase over stock speeds was much greater for the GTS again, because of artificially imposed restrictions. In the end, with max OC's considered, the GTX still held a healthy lead over the GTS due to actual architectural differences (4 ROP, 8 TMUs, +64-bit bus, +128 VRAM, +32 SP), but not as drastic as the benchmarks you see when comparing a 500MHz GTS to a 576 MHz GTX (stock speeds).

Now, if you extend this comparison to the new G92s, running the parts at the same clock speeds help isolate the actual differences between the parts. The GT gains in some places (16 SP, @30 TMU, faster core/shader clocks) but loses in others (4 ROP, 128 VRAM, 64-bit bus/bandwidth). The Tech-Report review, as well as this review, tend to show that the differences tend to be a wash at best, a negative at worst. Comparing max OC's for both is certainly relevant, but from what I've seen both first hand and in reviews, the benefits of OC'ing the core for the GT are much less than that of the GTS/GTX. This to me points to a bottleneck holding the GT back, whether its the 16 ROPs, bandwidth or VRAM. This is why I'm really interested in seeing how the G92 GTS fares, although a lot of that depends on what else is added besides the 16 SPs. Unless they add ROPs to bring it closer to the 24 on the GTX and increase VRAM to 768+, I don't expect it to surpass the GTX even with clock speeds approaching 700MHz.
 
Originally posted by: thejez
Originally posted by: shaolin95
Sorry but you make little sense JAG87.

They are different architectures so you can't compare their clock speeds. They are what they are and thats it. If one comes at 600 stock and the other at 500 or whatever, then those are each card's respective stock and if you OC one you must OC the other one to have a fair comparison. That the GTS cant run at 600 stock is not the GT's fault. Each card has a different strenght and in order to utilize the one on the GT you need the extra speed.
The GTS has more memory and a bigger bus so should we remove the extra memory and shrink the bus to make it fair comparison? Of course not because thats what that architecture is based upon.

Sorry but I have to totally disagree with your point of view. OC both cards and then we will see.

i tend to agree with this logic.


You are both brainwashed my marketing and labels. It doesn't matter if the card is OC, OC2, KO, SCC, what matters is the numbers.

With the exception of ROPs (which the GTS has more of), texture address/filters (which the GT has more of), the fact remains that the GT is clocked much higher than the GTS

So what is makes the GTS better. memory + bandwith make the difference.
 
Originally posted by: chizow
Originally posted by: shaolin95
Sorry but you make little sense JAG87.

They are different architectures so you can't compare their clock speeds. They are what they are and thats it. If one comes at 600 stock and the other at 500 or whatever, then those are each card's respective stock and if you OC one you must OC the other one to have a fair comparison. That the GTS cant run at 600 stock is not the GT's fault. Each card has a different strenght and in order to utilize the one on the GT you need the extra speed.
The GTS has more memory and a bigger bus so should we remove the extra memory and shrink the bus to make it fair comparison? Of course not because thats what that architecture is based upon.

Sorry but I have to totally disagree with your point of view. OC both cards and then we will see.

The architectures aren't so different that you can't draw relevant conclusions from running the cards at the same clock speeds, especially in the case of the G80 GTS where much of those architectural/clock speed differences are artificially imposed in order to differentiate products where the parts are otherwise the same. While NV may have binned/tested different G80 cores, then neutered some cores to make the GTS, there's little doubt that the current differentiation is little more than artifically imposed performance cuts to meet market/product differentiation specifications.

The new G80 112SP GTS is just further proof of this. Same for clock speeds. G80 GTS and GTX have typically had the same amount of OC'ing headroom since launch, with GTS users easily achieving similar OC's as GTX owners, but the actual increase over stock speeds was much greater for the GTS again, because of artificially imposed restrictions. In the end, with max OC's considered, the GTX still held a healthy lead over the GTS due to actual architectural differences (4 ROP, 8 TMUs, +64-bit bus, +128 VRAM, +32 SP), but not as drastic as the benchmarks you see when comparing a 500MHz GTS to a 576 MHz GTX (stock speeds).

Now, if you extend this comparison to the new G92s, running the parts at the same clock speeds help isolate the actual differences between the parts. The GT gains in some places (16 SP, @30 TMU, faster core/shader clocks) but loses in others (4 ROP, 128 VRAM, 64-bit bus/bandwidth). The Tech-Report review, as well as this review, tend to show that the differences tend to be a wash at best, a negative at worst. Comparing max OC's for both is certainly relevant, but from what I've seen both first hand and in reviews, the benefits of OC'ing the core for the GT are much less than that of the GTS/GTX. This to me points to a bottleneck holding the GT back, whether its the 16 ROPs, bandwidth or VRAM. This is why I'm really interested in seeing how the G92 GTS fares, although a lot of that depends on what else is added besides the 16 SPs. Unless they add ROPs to bring it closer to the 24 on the GTX and increase VRAM to 768+, I don't expect it to surpass the GTX even with clock speeds approaching 700MHz.

you seem to be one of the few who understands, kudos to you.
 
Originally posted by: JAG87
Originally posted by: thejez
Originally posted by: shaolin95
Sorry but you make little sense JAG87.

They are different architectures so you can't compare their clock speeds. They are what they are and thats it. If one comes at 600 stock and the other at 500 or whatever, then those are each card's respective stock and if you OC one you must OC the other one to have a fair comparison. That the GTS cant run at 600 stock is not the GT's fault. Each card has a different strenght and in order to utilize the one on the GT you need the extra speed.
The GTS has more memory and a bigger bus so should we remove the extra memory and shrink the bus to make it fair comparison? Of course not because thats what that architecture is based upon.

Sorry but I have to totally disagree with your point of view. OC both cards and then we will see.

i tend to agree with this logic.


You are both brainwashed my marketing and labels. It doesn't matter if the card is OC, OC2, KO, SCC, what matters is the numbers.

With the exception of ROPs (which the GTS has more of), texture address/filters (which the GT has more of), the fact remains that the GT is clocked much higher than the GTS

So what is makes the GTS better. memory + bandwith make the difference.

Which card is better is answered by one question: which one performs faster at its MAX... PERIOD. take both cards and ramp them up as high as they go and the one that has better numbers wins... you can argue ROPS FLOPS MOPS TOPS all you want...but in the end the one that can overclock higher and perform better wins. Thats the point of the comment and the sentiment i agreed with.

So take the 650/1000 GTS 640 and put up against the 800/1000 GT 512 and see what happens. You can't handicap either and call it fair fight!
 
I think the question was answered on previous page and the results speak for it self.

GTS is still faster in higher resolutions with AA. GT is faster in lower resolution and raw frame rates without AA

Knowing basic 3d architecture gives you a better understanding what is good at what and what it's weaknesses are. Sometimes a card is so similar in specs that they do tend to perform very similar and 1 can't just point and say it's faster than x card.
 
that's very interesting since we were all saying that 8800gt was the 2nd coming etc. now we find out that the gt is really just equal to or slightly behind 8800gts (and, therefore, 2900xt) at the higher resoutions favored by enthusiasts. It is looking more and more like we have an 8800gts 640 equivalent card that is priced more like the 320. Hopefully 3870 will drive the price down a little bit more so that those of us who are holding out will be able to truly get a "deal" on a card with strong performance.
 
I can't believe people are claiming the 140 extra megs of memory is significant on the GTS vs. the GT. Most games don't even come close to filling 512MB!

As for the rest, I agree that the highest clockspeed wins. Both cards have 112 shaders right?
 
Originally posted by: CP5670
Are the current batch of GTs overclocking better than the G80 based cards? That may even out the differences between these cards.

I would wait and see what the new GTS is like anyway before buying either of these, especially given the inflated prices right now.

These GTS's are G80 based... I'm waiting it out for the G92 and the new ATI card reviews.
 
Originally posted by: Azn
Originally posted by: SickBeast
I can't believe people are claiming the 140 extra megs of memory is significant on the GTS vs. the GT. Most games don't even come close to filling 512MB!

As for the rest, I agree that the highest clockspeed wins. Both cards have 112 shaders right?

You sure about that? http://www.yougamers.com/artic...ch_do_you_really_need/

thats a very good link... very informative. I'd still be curious to see a head-to-head with both cards maxed out to see if the higher clock on the GT can compensate for less mem... maybe not.

also, i think everybody and their mom is just gonna wait anyway until the G92 GTS comes out in a few weeks... i smell a step-up! 🙂
 
Originally posted by: taltamir

So how will a GT fare in a fair test? also the GT is PCIE2 so it has DOUBLE the bandwidth on a PCIE2 mobo....

This is wrong, the new pci-e slots has double the bandwidth, it does not double the bandwidth of the card itself. And currently not a singal card even approaches the bandwidth of the "old" pci-e slots so doubling the bandwidth will give zero performance increase at the moment.

"The GeForce 8800 GT is also the world?s first PCI Express 2.0 graphics card. PCIe 2.0 offers double the bandwidth of PCIe 1.1; 8.0GB/sec in each direction, providing a total of 16GB/sec of total memory bandwidth. Of course, we still haven?t come close to saturating the bandwidth offered by PCIe 1.1, but presumably this could come in handy for lower-end value cards with less onboard memory."

 
Originally posted by: keysplayr2003
Originally posted by: Azn
SSC whatever does it really matter? GTS can easily overclock to those levels and beyond.... Better performance in higher resolutions with AA than 8800gt even if it's older core.

Again games doesn't always scale with how many more stream processors. It's the combination. 16 more PS is not going to do anything in most situations. What matters is higher clock speed and memory bandwidth with a card like GTS.

How do you figure? If you took away those 16 SP's and reduced the G80 to 80SP's do you think that wouldn't make a difference either? Of course 16SP's would yield better performance than not having them. Look at the GTX. 32SP's over a standard GTS is substantial in EVERY situation.

All things being equal (core/mem clocks) between a GTS640 96sp and a GTS640 112sp, It's pretty much a given that the 112sp card would be more powerful. I could see variations from game to game, and the lead may shorten or lengthen, but the lead will always be there.

If the 8800GT and the 112 GTS640 were equal in price, It's a no brainer to go with the GTS.
Unfortunately, the 112 GTS640 is still brutally expensive to manufacture and it's price is prohibitive relative to it's performance. 8800GT is the winner hands down.

I'm waiting to see how the 128sp G92 GTS compares. I know it will be faster than the 8800GT and will most likely rival or best a GTX in most situations.

QFT, exactly. those extra SP count.
And the new GTS still has lower clocked SP and core compared to the GT. So that further evens the score.
 
Originally posted by: CP5670
Are the current batch of GTs overclocking better than the G80 based cards? That may even out the differences between these cards.

I would wait and see what the new GTS is like anyway before buying either of these, especially given the inflated prices right now.

Yes they are. The G80 gts can overclock from 500mhz stock core to just shy of 580mhz core on the max factory OC.
The G92 OCs from the stock 600mhz core to 700mhz core max factory OC...

"In fact the article was comparing a 580mhz oced new GTS to a stock 600mgz GT. Instead of comparing a stock 500mhz GTS to stock 600mhz GT... or a 580mhz GTS to 700mhz GT...
 
Back
Top