X1800xt Crossfire - 14200 in 3Dmark05

[Topweasel]

Originally posted by: Corporate Thug

Originally posted by: Elfear

Originally posted by: Topweasel

No its about Value, the type of feature list included on that board doesn't exceed that of a board made a year and a half ago, why pay pay an extra 60-100 dollars for it. People also complaine about the Price difference between picking up a 7800GTX 512MB compared to a X1800XT, yet this extra cost brings the price on 7800GTX 512MB to only $50 more then the 512MB version. Their is a difference between paying more even if the value isn't there for faster hardware, then paying more for more or less the same product.

Well, actually more like $200, but who's counting right.

X1800XT
512MB 7800GTX

Those were the lowest prices on Pricewatch. The 512MB 7800GTX isn't even in stock at that price.

how bout $499?


so lets see what we have: 2 512 GTXs (750 per)+ 150 mobo = $1650?
2 x1800xt ($499 for regular, $550 for master)+200 Mobo=1250

hmm...

My point wasn't about over all cost I even said it, I am sorry I went off about prices and I hadn't looked at them recently. Last I checked the XT was selling for 599 and the GTX 512 was 750. Along this side comment I got my SLI board recently for $134 shipped.

My Point was the Value in buying a $200 board when you can get the same features and performance on a board for, in my case, $70 cheaper. Also a more feature rich (and if you like this kind of thing, more future proof) motherboard is available for the same price. Plus you get the a bonus of being able to pair up the best performing card instead of having a upper limit of a two way/Slightly better tie for 2nd place (which back to money the other card can be purchased for about $70-$100 less).

I am not accusing the chipset of being a bad chipset, I just don't see how it can make sense to purchase it. The only value I see for it would be soley If you intend to run two x1800XT 512MBs, since they can be used on a Nforce4 with a single card otherwise, at that amount I personally would, even with the extra $200-$300, go all out with the GTX 512MBs, but thats just me. I don't understand why you would spend more on one product just because you can spend less on others and still get less features and less performance.

 

[Cookie Monster]

I think everyone missed my question. Im talking about the price of the X1800XT CF cards. Are they going to be around 550? what is the known MSRP of them, so i can get an idea of how much a Cfire setup is going to be. Crossfire looks as good as SLi, i wouldnt mind Crossfire X1800XL.

btw I wasnt even talking about 7800GTX 512mb .

 

[Corporate Thug]

Originally posted by: Topweasel

Originally posted by: Corporate Thug

Originally posted by: Elfear

Originally posted by: Topweasel

No its about Value, the type of feature list included on that board doesn't exceed that of a board made a year and a half ago, why pay pay an extra 60-100 dollars for it. People also complaine about the Price difference between picking up a 7800GTX 512MB compared to a X1800XT, yet this extra cost brings the price on 7800GTX 512MB to only $50 more then the 512MB version. Their is a difference between paying more even if the value isn't there for faster hardware, then paying more for more or less the same product.

Well, actually more like $200, but who's counting right.

X1800XT
512MB 7800GTX

Those were the lowest prices on Pricewatch. The 512MB 7800GTX isn't even in stock at that price.

how bout $499?


so lets see what we have: 2 512 GTXs (750 per)+ 150 mobo = $1650?
2 x1800xt ($499 for regular, $550 for master)+200 Mobo=1250

hmm...

My point wasn't about over all cost I even said it, I am sorry I went off about prices and I hadn't looked at them recently. Last I checked the XT was selling for 599 and the GTX 512 was 750. Along this side comment I got my SLI board recently for $134 shipped.

My Point was the Value in buying a $200 board when you can get the same features and performance on a board for, in my case, $70 cheaper. Also a more feature rich (and if you like this kind of thing, more future proof) motherboard is available for the same price. Plus you get the a bonus of being able to pair up the best performing card instead of having a upper limit of a two way/Slightly better tie for 2nd place (which back to money the other card can be purchased for about $70-$100 less).

I am not accusing the chipset of being a bad chipset, I just don't see how it can make sense to purchase it. The only value I see for it would be soley If you intend to run two x1800XT 512MBs, since they can be used on a Nforce4 with a single card otherwise, at that amount I personally would, even with the extra $200-$300, go all out with the GTX 512MBs, but thats just me. I don't understand why you would spend more on one product just because you can spend less on others and still get less features and less performance.

i hear ya,

but CF mobos are new. gie it a chance for more to hit the market and they will be a lot cheaper. BTW, the asus board that AT Weasley reviewed with high acclaim can soon be had for a very reasonable price

http://www.excaliberpc.com/ASUS_MOTHERB..._RD480/A8R-MVP/partinfo-id-561912.html

 

[Keysplayr]

Originally posted by: Sc4freak


The number of pipes in a core is irrelevant; Since when?

1 pipe at a million ghz will draw more power than 100 pipes at 1mhz. Thats right, because 1 pipe at a million GHz is running 999,999,900 Mhz faster than 100 pipes at 1 MHz. Bad example. There is only a 75MHz difference in core speed between a X1800XT and a 7800GTX512. How much more power do you think the X1800XT would use if it had 8 more pipes at its present speed? I think it would go up considerably don't you? So, pipes are not so irrelevant.


I also don't see how the low-k process comes in here...
I posted a link in one of my posts above regarding exactly what Low-k has to do with power consumption. I would say, maybe you should take a look at it. Link

we're discussing the power draw of the card as a whole, regardless of the processes and power draw optimisations.
I was not speaking about certain portions of each card Of course I am talking about the entire card. That would be what the power comsumption testing would represent.

True, there is a performance delta between the two (the X1800XT 512mb sits between the 7800GTX 256mb and 7800GTX 512mb in terms of performance), but what you're suggesting is that we need to increase the clockspeeds, increase the number of pipes, and remove TSMC's low-k in order for the X1800XT to become eligible to be compared to the 7800GTX 512mb in terms of power consumption...
I didn't suggest any such thing. The X1800XT is technically one gen (manufacturing process) ahead of the GTX/512. Its 90nm (less power is needed to run smaller circuits in smaller processes) utilizes low-k dielectric process (greatly reduces power leakage and crosstalk between said circuits, sort of an insulator. Check my link to TSMC's site like I mentioned) and has 8 less pipes to run, and it STILL uses a large amount of power. More than a 7800GTX and less than a 512.

In plain English, what I am trying to say is: Take a 7800GTX512 as it exists right now. Shrink the die to 90nm. Lets be conservative and reduce power consumption by 5% and that alone should knock down 15W of power consumption. Now apply low-k. TSMC states power savings of 20% using their low-k process. Lets make it 5% again to be conservative. That would knock off a total of 30W of power needed to run it.
My conclusion: A 90nm low-k GTX512 would use less power than the current X1800XT using the same 90nm low-k design. G71 is exactly what this is supposed to be. A die shrunk, low-k 7800GTX512 with cranked up clocks. So it may very well use the same amount of power with the clocks cranked.

I really do not understand what you meant as "the cards as a whole" because I meant nothing other than that.

To sort of bring this back OnTopic, 2x G71's SLI'd would use less power than Crossfired X1800XT's.

 

[Munky]

Originally posted by: keysplayr2003

Originally posted by: Sc4freak


The number of pipes in a core is irrelevant; Since when?

1 pipe at a million ghz will draw more power than 100 pipes at 1mhz. Thats right, because 1 pipe at a million GHz is running 999,999,900 Mhz faster than 100 pipes at 1 MHz. Bad example. There is only a 75MHz difference in core speed between a X1800XT and a 7800GTX512. How much more power do you think the X1800XT would use if it had 8 more pipes at its present speed? I think it would go up considerably don't you? So, pipes are not so irrelevant.


I also don't see how the low-k process comes in here...
I posted a link in one of my posts above regarding exactly what Low-k has to do with power consumption. I would say, maybe you should take a look at it. Link

we're discussing the power draw of the card as a whole, regardless of the processes and power draw optimisations.
I was not speaking about certain portions of each card Of course I am talking about the entire card. That would be what the power comsumption testing would represent.

True, there is a performance delta between the two (the X1800XT 512mb sits between the 7800GTX 256mb and 7800GTX 512mb in terms of performance), but what you're suggesting is that we need to increase the clockspeeds, increase the number of pipes, and remove TSMC's low-k in order for the X1800XT to become eligible to be compared to the 7800GTX 512mb in terms of power consumption...
I didn't suggest any such thing. The X1800XT is technically one gen (manufacturing process) ahead of the GTX/512. Its 90nm (less power is needed to run smaller circuits in smaller processes) utilizes low-k dielectric process (greatly reduces power leakage and crosstalk between said circuits, sort of an insulator. Check my link to TSMC's site like I mentioned) and has 8 less pipes to run, and it STILL uses a large amount of power. More than a 7800GTX and less than a 512.

In plain English, what I am trying to say is: Take a 7800GTX512 as it exists right now. Shrink the die to 90nm. Lets be conservative and reduce power consumption by 5% and that alone should knock down 15W of power consumption. Now apply low-k. TSMC states power savings of 20% using their low-k process. Lets make it 5% again to be conservative. That would knock off a total of 30W of power needed to run it.
My conclusion: A 90nm low-k GTX512 would use less power than the current X1800XT using the same 90nm low-k design. G71 is exactly what this is supposed to be. A die shrunk, low-k 7800GTX512 with cranked up clocks. So it may very well use the same amount of power with the clocks cranked.

I really do not understand what you meant as "the cards as a whole" because I meant nothing other than that.

To sort of bring this back OnTopic, 2x G71's SLI'd would use less power than Crossfired X1800XT's.

I see what you're trying to get at, but instead of counting the card on a per-pipe basis, why dont you concider the total amount of transistors on each card? Both the gtx and the xt have around 310-325 million transistors, and each one of those draws power, whether it's used in the pixel pipes or not. So while th eactual "pipes" on the xt may take up less transistor space, it has a complex 512-bit memory bus/controller on that gpu, which would also take up a huge number of transistors, and thus it's no surprise that the xt sits right between the 512 and 256 gtx's in terms of power draw.

 

[aatf510]

Originally posted by: keysplayr2003
In plain English, what I am trying to say is: Take a 7800GTX512 as it exists right now. Shrink the die to 90nm. Lets be conservative and reduce power consumption by 5% and that alone should knock down 15W of power consumption. Now apply low-k. TSMC states power savings of 20% using their low-k process. Lets make it 5% again to be conservative. That would knock off a total of 30W of power needed to run it.
My conclusion: A 90nm low-k GTX512 would use less power than the current X1800XT using the same 90nm low-k design. G71 is exactly what this is supposed to be. A die shrunk, low-k 7800GTX512 with cranked up clocks. So it may very well use the same amount of power with the clocks cranked.

I really do not understand what you meant as "the cards as a whole" because I meant nothing other than that.

To sort of bring this back OnTopic, 2x G71's SLI'd would use less power than Crossfired X1800XT's.

yup, it's entirely possible, and who knows if 2x R580 would use less power than SLIed G71.

 

[Malak]

Originally posted by: keysplayr2003

Originally posted by: Dman877
If their psu can handle 2 1800XT's (which use more power then 7800's iirc) it should handle a couple gtx's...

Unless there is something physically wrong with one of the GTX512's, I'm gonna go with user error on this one.

Many benchmarkers have had issues with nvidia's SLI being unstable.

Truthfully, I still think both SLI and crossfire are a waste of money.

 

[Paratus]

Keys its not the pipes its #of transistors and Freqs that determines power output. The XT has more of both but uses less power....

 

[Keysplayr]

Originally posted by: Paratus
Keys its not the pipes its #of transistors and Freqs that determines power output. The XT has more of both but uses less power....

Ok, lets go with that then. Increase the die size to 110nm and remove low-k. Whats going to happen? As per munky, the huge transistor increase on the X1800XT could be due to that very cool memory controller. For what the X1800XT is, it does use a good amount of power. Heck, they all do. R600/G80 will probably be even more insane and need 3 molex connectors on each PLUS a power brick

 

[imported_michaelpatrick33]

Originally posted by: Rollo

Originally posted by: Elfear
Not sure if I agree on their results for SLI'd 256MB 7800GTX's. I don't see many many at all that hit 13,500, let alone at 2.8GHz and stock clocks on the cards. An SLI'd 512MB 7800GTX rig at stock clocks with an FX-57 scores about 13,300. With some average overclocking they get about 14,000-14,500.

I ran 3Dmark05 the other day and got `11456 with my 7800GTXs/4600+, all stock. The FX57 is obviously a faster cpu for things that aren't multithreaded, but those are impressive 3dMark scores for the Crossfire.

AFAIK this is due out January, so review sites having them doesn't shock me.

Also, I have to retract on my "Crossfire for 256MB Only" post, with apologies. Apparently the B3d thread was wrong, and we just got to see a peak of the 256MB version first, for reasons unknown.

My bolded text for emphasis. The bold is Rollo's comment
:shocked: :shocked:
:laugh:

Rollo I take back every positive thing I have said about you. I want FUD and controversy :laugh: (and yes I am joking)

I just hope NVIDIA releases a mixed card SLI mode so my 7800GTX can have a 7800GT buddy. It does seem like ATI is beginning to right the ship slightly here.

 

[Keysplayr]

Originally posted by: michaelpatrick33

Originally posted by: Rollo

Originally posted by: Elfear
Not sure if I agree on their results for SLI'd 256MB 7800GTX's. I don't see many many at all that hit 13,500, let alone at 2.8GHz and stock clocks on the cards. An SLI'd 512MB 7800GTX rig at stock clocks with an FX-57 scores about 13,300. With some average overclocking they get about 14,000-14,500.

I ran 3Dmark05 the other day and got `11456 with my 7800GTXs/4600+, all stock. The FX57 is obviously a faster cpu for things that aren't multithreaded, but those are impressive 3dMark scores for the Crossfire.

AFAIK this is due out January, so review sites having them doesn't shock me.

Also, I have to retract on my "Crossfire for 256MB Only" post, with apologies. Apparently the B3d thread was wrong, and we just got to see a peak of the 256MB version first, for reasons unknown.

My bolded text for emphasis. The bold is Rollo's comment
:shocked: :shocked:
:laugh:

Rollo I take back every positive thing I have said about you. I want FUD and controversy :laugh: (and yes I am joking)

I just hope NVIDIA releases a mixed card SLI mode so my 7800GTX can have a 7800GT buddy. It does seem like ATI is beginning to right the ship slightly here.

Maybe ATI is getting their bilge pumps working finally.

 

[Keysplayr]

Originally posted by: RobertR1

Originally posted by: Cookie Monster

Originally posted by: Corporate Thug

Originally posted by: Elfear

Originally posted by: Topweasel

No its about Value, the type of feature list included on that board doesn't exceed that of a board made a year and a half ago, why pay pay an extra 60-100 dollars for it. People also complaine about the Price difference between picking up a 7800GTX 512MB compared to a X1800XT, yet this extra cost brings the price on 7800GTX 512MB to only $50 more then the 512MB version. Their is a difference between paying more even if the value isn't there for faster hardware, then paying more for more or less the same product.

Well, actually more like $200, but who's counting right.

X1800XT
512MB 7800GTX

Those were the lowest prices on Pricewatch. The 512MB 7800GTX isn't even in stock at that price.

how bout $499?


so lets see what we have: 2 512 GTXs (750 per)+ 150 mobo = $1650?
2 x1800xt ($499 for regular, $550 for master)+200 Mobo=1250

hmm...

Is the X1800XT CF going to cost 550? wasnt the MSRP of the X1800XT 599?

Let me get this straight.....when the 7800GTX 256 was out and the x1800xt was announced, it was all about "street price" but now that the street price on the x1800xt is considerably lower than the street price of the 7800GTX 512, MSRP is important again?

If you want to get things straight, the 499.00 XT is right where it's supposed to be right now HUGELY better than the 649.00 is was going for at "launch". 7800GTX's are about 450.00, and 7800GTX512's are just plain ridiculous.

MSRP is a big deal or was a big deal at the launch of the X1800XT, because it was actually selling OVER MSRP and the 7800GTX was selling UNDER MSRP. It was difficult to justify buying an XT over a GTX because of those price differences. Now, its a lot easier because both are under MSRP and a lot closer to where they belong in the price performance area.

 

[RobertR1]

Originally posted by: keysplayr2003

Originally posted by: RobertR1

Originally posted by: Cookie Monster

Originally posted by: Corporate Thug

Originally posted by: Elfear

Originally posted by: Topweasel

No its about Value, the type of feature list included on that board doesn't exceed that of a board made a year and a half ago, why pay pay an extra 60-100 dollars for it. People also complaine about the Price difference between picking up a 7800GTX 512MB compared to a X1800XT, yet this extra cost brings the price on 7800GTX 512MB to only $50 more then the 512MB version. Their is a difference between paying more even if the value isn't there for faster hardware, then paying more for more or less the same product.

Well, actually more like $200, but who's counting right.

X1800XT
512MB 7800GTX

Those were the lowest prices on Pricewatch. The 512MB 7800GTX isn't even in stock at that price.

how bout $499?


so lets see what we have: 2 512 GTXs (750 per)+ 150 mobo = $1650?
2 x1800xt ($499 for regular, $550 for master)+200 Mobo=1250

hmm...

Is the X1800XT CF going to cost 550? wasnt the MSRP of the X1800XT 599?

Let me get this straight.....when the 7800GTX 256 was out and the x1800xt was announced, it was all about "street price" but now that the street price on the x1800xt is considerably lower than the street price of the 7800GTX 512, MSRP is important again?

If you want to get things straight, the 499.00 XT is right where it's supposed to be right now HUGELY better than the 649.00 is was going for at "launch". 7800GTX's are about 450.00, and 7800GTX512's are just plain ridiculous.

MSRP is a big deal or was a big deal at the launch of the X1800XT, because it was actually selling OVER MSRP and the 7800GTX was selling UNDER MSRP. It was difficult to justify buying an XT over a GTX because of those price differences. Now, its a lot easier because both are under MSRP and a lot closer to where they belong in the price performance area.

Agreed that's why for me, the x1800xt was the best buy out of bunch. I personally couldn't justify an extra $250 for a 7800GTX 512 and the x1800xt was easily worth $50 more for the performance it offered over the 7800GTX 256.