x1800--->9700marks(default) at 3dmark05?

[cadaveca]

Again let's clear up the misconceptions and false data.

Never has our chipset had a problem overclocking the HT bus. On the contrary, others have problems even running 1GHz. We have many times demonstrated that over 1 GHz operation is easy. Macci, Fugger, Bigtoe and others will attest to that.

Grayskull... if someone other than working for the industry had a board, and some results, we wouldn't be having this conversation, and you wouldn't be making these posts. The sources you mention are biased, IMHO, as they all get engineering samples from you ATI folk. We all know Macci gets good hardware from you guys too, even though he works/is affiliated with 3dmark. Just because Jason is here pumping nVidia just shows how scared nvidia is.

As for Texas, the problem there was the DDR interface. This was on Bullhead, our OEM reference board that was never designed for overclocking, and despite that ran extremely well except under phase. Texas was the first time that our boards were put under phase so it was a learning experience. We've since figured it all out and have demonstrated that Grouper and Halibut are both very phase friendly.

Is this why the Bullhead is not shown anymore, or is the bullhead the normal crossfire layout? That's some great info. Basically what you are saying is that the halibut and grouper are seriously tweaked versions of the bullhead, which was a decent performer in it's own right. That's VERY good to hear, but we all would really like to see these boards for ourselves.

[Grayskull]

Sometimes I really don't know why I try ... seems like some people just have an anti-ATI slant and it's always the same people, but for everyone else out there who are little more open minded...

Bullhead was never designed for overclocking in mind. I will tell you that with 100% certainty because our team designed it. So for the doubters, why not ask Wesley Fink at Anandtech and/or Tony Leach at OCZ. Both saw the first incarnation of Bullhead and can attest to the fact it had no overclocking features at all.

A second revision of Bullhead was produced with only minor mods, Vdimm to 2.8V, CPU Vcore to 1.6V and that's it. Nothing else, all other voltages were fixed. CPU and memory power delivery were not designed for overclocking neither. Even with these limitations the board did pretty good. This is what Anandtech reported.

Grouper and Halibut were designed ground up with overclocking in mind with too many changes to list. It's overclockability speaks for itself. As for bios, if you would compare the Bullhead bios and the Grouper/Halibut bios you would quickly realize that they don't even compare. Again, for all the doubters, why not ask any reviewer that still has a Bullhead and just got a Halibut for the Crossfire launch? Don't let me tell you, let them.

Bullhead was never designed for overclocking? Why does the bios have all these overclocking features? Why was it used in an overclocking competition then?
Why did Anandtech praise its overclocking capacities? Misconception? False data?

http://www.anandtech.com/cpuchipsets...spx?i=2269&p=3

[Grayskull]

Yes I realize that when these chipsets make it to market everyone will have a go at it themselves and that will remove all doubt.

As for Bullhead not being shown anymore, Bullhead served it's purpose very, very well and has been retired. It helped us gain the lion share of all the AMD IGP business and we've since moved on to new chipsets and new boards.

Last, the Grouper / Halibut designs are not tweaked Bullheads. They are a totally new design and layout. You will get to play with these boards literally and not so literally. Literally because some vendors are bringing the board to market directly. I would keep my eye on the DFI version. Not so literally because others have taken the best of Grouper/Halibut and grafted it into their own designs.

As for the green guys, if I had to lay down a wager, I'd say they're more than a little concerned.

Grayskull... if someone other than working for the industry had a board, and some results, we wouldn't be having this conversation, and you wouldn't be making these posts. The sources you mention are biased, IMHO, as they all get engineering samples from you ATI folk. We all know Macci gets good hardware from you guys too, even though he works/is affiliated with 3dmark. Just because Jason is here pumping nVidia just shows how scared nvidia is.



Is this why the Bullhead is not shown anymore, or is the bullhead the normal crossfire layout? That's some great info. Basically what you are saying is that the halibut and grouper are seriously tweaked versions of the bullhead, which was a decent performer in it's own right. That's VERY good to hear, but we all would really like to see these boards for ourselves.

[cadaveca]

Last, the Grouper / Halibut designs are not tweaked Bullheads. They are a totally new design and layout. You will get to play with these boards literally and not so literally. Literally because some vendors are bringing the board to market directly. I would keep my eye on the DFI version. Not so literally because others have taken the best of Grouper/Halibut and grafted it into their own designs.

Good info. As always.

[AACDIRECT]

I have used both boards. They are night and day compared to each other. The Grouper board I used seemed like a reference board. It ran good but was not very polished. (Heck the board I received had hand done voltmods and rework wires running down the back of it.) It seemed like an engineering sample. Even saying that with a little tweaking it ran very stable and fast. (It made me excited to see what was coming in the form of production product.)

The Halibut on the other had seems like a production board. If board makers copy this board exactly as it sits it will be a fantastic motherboard. It has not fought me on any setup I have tested with it. The only thing that gives any hint that it is not a full retail product is the lack of the 16x to 8x/8x in bios instead it uses a terminator. Even the bios is far more full featured. Silly things like hardware monitoring was not in Grouper it is in Halibut. The Halibut seems to not lack anything.

[Grayskull]

Kyle, you got one of the first batch of Groupers, that's why it was a little rough.

As for the 1x16 to 2x8 switching, it is now fully automatic, no need for any cmos intervention. The terminator card design was done purposely. It's cheap, has the least disruption to onboard routing and is very user friendly. Even if the terminator card is omitted, the board will still boot up and function perfectly fine on one card, but it will just run on 8 lanes instead of 16 lanes. Even uneducated consumers won't have a problem.

The use of electronic switching although convenient, is quite expensive and reduces margin on the bus. Paddle boards also reduce margin, but not as much as the electronic switches. Paddle boards also occupy valuable motherboard real estate unnecessarily.

I have used both boards. They are night and day compared to each other. The Grouper board I used seemed like a reference board. It ran good but was not very polished. (Heck the board I received had hand done voltmods and rework wires running down the back of it.) It seemed like an engineering sample. Even saying that with a little tweaking it ran very stable and fast. (It made me excited to see what was coming in the form of production product.)

The Halibut on the other had seems like a production board. If board makers copy this board exactly as it sits it will be a fantastic motherboard. It has not fought me on any setup I have tested with it. The only thing that gives any hint that it is not a full retail product is the lack of the 16x to 8x/8x in bios instead it uses a terminator. Even the bios is far more full featured. Silly things like hardware monitoring was not in Grouper it is in Halibut. The Halibut seems to not lack anything.

[AACDIRECT]

Kyle, you got one of the first batch of Groupers, that's why it was a little rough.

As for the 1x16 to 2x8 switching, it is now fully automatic, no need for any cmos intervention. The terminator card design was done purposely. It's cheap, has the least disruption to onboard routing and is very user friendly. Even if the terminator card is omitted, the board will still boot up and function perfectly fine on one card, but it will just run on 8 lanes instead of 16 lanes. Even uneducated consumers won't have a problem.

First batch or not it still ran good.

The terminator does work fine. It just happens to be one of the few things that I know most retail boards will not have that the reference do. Really hard to find faults in the motherboard so the terminator is just one of the few things I can mention that I had heard was not making it to retail. (BTW the solder pads between the 16X slots are they for the electronic lane switching?)

The use of electronic switching although convenient, is quite expensive and reduces margin on the bus. Paddle boards also reduce margin, but not as much as the electronic switches. Paddle boards also occupy valuable motherboard real estate unnecessarily.

By expensive do you mean cost? Could you explain what you mean by margin on the bus? Just curious.....

Also what is the difference between the boards that have different color silk screening. Anything? Mine has White silkscreening with all red connectors and red NB heatsink. I assume there is no difference just different runs but again was curious if there was anything else.

[Grayskull]

By expensive I do mean cost. The electronic switches are expensive compared to a terminator card.

By reduced margin, I mean that the current electronic switches out there are passive, not active devices. This means that any signal that passes through them will have some attenuation and that always reduces margin on the interface.

All dual slot reference boards from ATI are black (Halibut). All single slot boards (Grouper) are white. Connector and heatsink colors have varied over the lifetime of the boards.

First batch or not it still ran good.

The terminator does work fine. It just happens to be one of the few things that I know most retail boards will not have that the reference do. Really hard to find faults in the motherboard so the terminator is just one of the few things I can mention that I had heard was not making it to retail. (BTW the solder pads between the 16X slots are they for the electronic lane switching?)



By expensive do you mean cost? Could you explain what you mean by margin on the bus? Just curious.....

Also what is the difference between the boards that have different color silk screening. Anything? Mine has White silkscreening with all red connectors and red NB heatsink. I assume there is no difference just different runs but again was curious if there was anything else.

[AACDIRECT]

By expensive I do mean cost. The electronic switches are expensive compared to a terminator card.

By reduced margin, I mean that the current electronic switches out there are passive, not active devices. This means that any signal that passes through them will have some attenuation and that always reduces margin on the interface.

All dual slot reference boards from ATI are black (Halibut). All single slot boards (Grouper) are white. Connector and heatsink colors have varied over the lifetime of the boards.

Thanks for the information. Mine is the most beautiful one..... MY PRECIOUS.....

On the terminator. Just to make sure I am understanding correctly. The terminator produces less single loss Vs a Electronic switch. So for overclocking/performance the terminator could be consider the best solution? If so is anyone you know planning on offering a board that has the terminator?

Also what does the four LEDs above the CPU socket show? I assume they are not there just to be pretty....AKA Fatal1ty mobos. On mine only two are lit during normal use. (BTW the LED's for SATA and PATA are a nice touch too.)

Again thanks for the info. Very interesting to learn why certain things are done at the design level.

[Grayskull]

DFI is the only vendor that I am aware of that is using the terminator scheme on the CF-BT board. Others may as well, but I'm not aware of it.

There's actually 5 LEDs... or there should be 5 They represent the VID code of the cpu power delivery system. These probably won't make it into production as they are not really needed. But I guess we'll just have to see.

Thanks for the information. Mine is the most beautiful one..... MY PRECIOUS.....

On the terminator. Just to make sure I am understanding correctly. The terminator produces less single loss Vs a Electronic switch. So for overclocking/performance the terminator could be consider the best solution? If so is anyone you know planning on offering a board that has the terminator?

Also what does the four LEDs above the CPU socket show? I assume they are not there just to be pretty....AKA Fatal1ty mobos. On mine only two are lit during normal use. (BTW the LED's for SATA and PATA are a nice touch too.)

Again thanks for the info. Very interesting to learn why certain things are done at the design level.

[AACDIRECT]

DFI is the only vendor that I am aware of that is using the terminator scheme on the CF-BT board. Others may as well, but I'm not aware of it.

There's actually 5 LEDs... or there should be 5 They represent the VID code of the cpu power delivery system. These probably won't make it into production as they are not really needed. But I guess we'll just have to see.

LOL If I could count ..... so what does off,off,on,on,off mean?

Guess we got a little off topic on this post..... For those waiting on the DFI CF-BT there was some info that I had not personally heard yet. Again thanks for the info Grayskull.

[cadaveca]

DFI is the only vendor that I am aware of that is using the terminator scheme on the CF-BT board. Others may as well, but I'm not aware of it.

I'll never understand why people will take a ferrari engine and put it into a Pinto.


The industry is in a sad state of affairs.

[Vapor]

Can you explain what you're saying, cadaveca? Isn't the upcoming CF-BT DFI's top of the line ATi board?

EDIT.....got it

[cadaveca]

edited for clarity, ok?

[AACDIRECT]

I just got off the phone with DFI about the Crossfire boards. (I was told I could pass on the info so here it is.) Both boards would be consider a premium part. It sounds like the CF-BT will be a little cheaper than the CF-DR. The CF-BT is a four layer design while the CF-DR is a six layer adding to the cost. The CF-BT will not be blue as they could not get it to look just right. It will be black like the CF-DR. We should see CF-DR boards very very soon. The CF-BT are still a few weeks out.

It sounds like they spared no expense on both boards. It should be interesting to see which one turns out to be the best. From what I have heard both should be great boards.

BTW In case anyone care X850 Crossfire videocards are available.

So CF-BT Vs CF-DR...FIGHT

[Vapor]

Why would they release both if they're both premium boards? Any other differences that they alluded to?

[cadaveca]

Price. Although i think if BigToe had something to do with the BT, it may be more ram friendly.

[AACDIRECT]

Why would they release both if they're both premium boards? Any other differences that they alluded to?

I think price is part of it. Overall the features seem very much the same (I have not received a final spec on both). BT does = BigToe

As I said it will be interesting to see which is better. Sounds like we will know soon.

DFI did say that they have been working hard on ram compatibility. One interesting tidbit was that the CF-DR will run 4 modules at 400+ Mhz 1T with no issues. I was told they also worked hard on high density memory compatiblity. (AKA 1+ GIG modules)

[Vapor]

Looks like the CF-DR is in my future....I run 4 sticks of memory.

[dinos22]

DFI did say that they have been working hard on ram compatibility. One interesting tidbit was that the CF-DR will run 4 modules at 400+ Mhz 1T with no issues. I was told they also worked hard on high density memory compatiblity. (AKA 1+ GIG modules)

what about Expert board......man it would be really nice to be able to run some big ass capacity RAM in there at 1T.......4GB RAM at 1T at DDR400 or more

[AACDIRECT]

what about Expert board......man it would be really nice to be able to run some big ass capacity RAM in there at 1T.......4GB RAM at 1T at DDR400 or more

We talked mainly about crossfire but the Expert board is supposed to be improved in this department too. I did not ask specificly but he did say it was "much" improved over the current SLI-DR boards. (Kind of scary that they could make such a good board better.) The Expert is to use an electronic switch for SLI, more space between 16X slots, better high VRAM support, etc.

[dinos22]

he did say it was "much" improved over the current SLI-DR boards. (Kind of scary that they could make such a good board better.) The Expert is to use an electronic switch for SLI, more space between 16X slots, better high VRAM support, etc.

i wonder what he means aside from the aforementioned features...........i've got one on order but if the CF is better in that department (4xsticks of RAM at 1T DDR400) than i'd have to look at CF as well

[b0bd0le]

lots of good information here! thanks gray

is there a possible ETA for these motherboards?

[JuanFlaiter]

We talked mainly about crossfire but the Expert board is supposed to be improved in this department too. I did not ask specificly but he did say it was "much" improved over the current SLI-DR boards. (Kind of scary that they could make such a good board better.) The Expert is to use an electronic switch for SLI, more space between 16X slots, better high VRAM support, etc.

You talked with Jacky?

[AACDIRECT]

You talked with Jacky?

Yep. He is nice guy. Not every day that the GM of a company is as excited about the product they sell. (Most care about selling product not the product itself.) I had always talked with Vivian before she left but now have talked with him a few times.

The ETA given to me on the CF-DR is the next week or so. You could say any day now.

The CF-BT was OCT....could be NOV.

I can't remember on the NF4 Expert board....sorry.