Official HD 5000 specs

[Maziar]

There you go:
http://www.notebookcheck.net/Newsentry.153+M5898b1db710.0.html
* ATI Mobility Radeon HD 5870

800 Shader, 128 Bit memory bus, (G)DDR3, GDDR5, 700 / 2000 MHz Core / Memory
* ATI Mobility Radeon HD 5850

800 Shader, 128 Bit memory bus, (G)DDR3, GDDR5, 625 / 2000 MHz Core / Memory

I was expecting more really

 

[lostandwandering]

I'm not impressed with either NVidia or ATI this time around.

 

[Maziar]

Me neither

 

[lostandwandering]

It's especially less impressive that over 2 or 3 generations, ATI has yet to surpass the performance of that darned mobile G92b chip that is still being used in the GTX 280M and 260M. That's 2 generations of stagnant performance and IMO it's unacceptable. Kudos to NVidia for being able to use a chip that is now 3 years old and hang onto the performance crown in laptops. But shame on both of them for not doing better.

 

[Maziar]

Yes we are still having re-badged cards

 

[TheGreatGrapeApe]

They aren't re-badged, but neither has launched a full high-end desktop card since the X800, and even that ended up in like 2 laptops because they couldn't make it fit the power envelope. Even the old MR9700 was just a mobilized R9600, and the following GF6800 wasn't a desktop version, and the GF7 and X1K series were not the desktop models, and so on (never saw the G80 or R600 make it to mobile), just the G92 and HD4770 essentially.

Almost all of the high end have been restricted refreshes or low end cards, and definitely all the recent ones.

Yes, the MHD5870 is just a mobilized version of the HD5770, but it's still going to be the fastest kid on the block and it's going to be nice for anyone wanting something approaching desktop power.

As for the G92b being faster, depends on what you put it to the task of doing.
The GTX280 is not that much faster than the GF9800M GTX, and it's well back of the MHD4870 in 3DMarkVantage;
http://service.futuremark.com/hardware/graphics_cards/nvidia_geforce_gtx_280m?chipsetId=531&popularChipsetIds=462&popularChipsetIds=425&popularChipsetIds=525&popularChipsetIds=531

It's really the older titles that favour the G92, not the newer ones.

As for doing better, it's kinda hard to fit all this in a mobile platform, and compared to what they used to offer us during the early GeforceGo and mid-year Mobile Radeons, this is actually pretty good really. Sure I'd love Cypress or Fermi in a laptop, but it ain't gonna happen, and never has, I just thank them for not sticking me with just the Redwood/5670 or GT216 as the only options.

 

[megamanx00]

The mobile 5870 is pretty much just a 5770, which is still generally faster than any G92b card and is a DX11 part. For nVidia though, yeah they just don't seem to want to spend money developing a new part so they'll keep sticking with the G92b so long as people buying it even if it's simply under another name.

 

[frozenlead]

I'm kind of with TGGA. It's a bit sad that these really aren't the high parts that the desktops get (even though those are more refreshes now) - but they are throwing the mobile sector some bones. Especially when you consider most (even here) think gaming notebooks are a waste. The entire gaming notebook market is pretty much the most difficult one to make components for - the limitations are much more strict than they are for desktop parts, which seem to be getting bigger and bigger all the time.

Also, the manufacturers know that notebook parts are hardly interchangable and replaceable - anyone who buys off of them is pretty much stuck with what they get. There aren't any alternatives...so there's not much incentive for them to make some lucrative products.

I don't know if I'm trying to say we're lucky we get what we get or we should be angry that we don't get more...but since the hardware is dead-end when it comes out, it really doesn't matter either way.

 

[Maziar]

They aren't re-badged, but neither has launched a full high-end desktop card since the X800, and even that ended up in like 2 laptops because they couldn't make it fit the power envelope. Even the old MR9700 was just a mobilized R9600, and the following GF6800 wasn't a desktop version, and the GF7 and X1K series were not the desktop models, and so on (never saw the G80 or R600 make it to mobile), just the G92 and HD4770 essentially.

Almost all of the high end have been restricted refreshes or low end cards, and definitely all the recent ones.

Yes, the MHD5870 is just a mobilized version of the HD5770, but it's still going to be the fastest kid on the block and it's going to be nice for anyone wanting something approaching desktop power.

As for the G92b being faster, depends on what you put it to the task of doing.
The GTX280 is not that much faster than the GF9800M GTX, and it's well back of the MHD4870 in 3DMarkVantage;
http://service.futuremark.com/hardware/graphics_cards/nvidia_geforce_gtx_280m?chipsetId=531&popularChipsetIds=462&popularChipsetIds=425&popularChipsetIds=525&popularChipsetIds=531

It's really the older titles that favour the G92, not the newer ones.

As for doing better, it's kinda hard to fit all this in a mobile platform, and compared to what they used to offer us during the early GeforceGo and mid-year Mobile Radeons, this is actually pretty good really. Sure I'd love Cypress or Fermi in a laptop, but it ain't gonna happen, and never has, I just thank them for not sticking me with just the Redwood/5670 or GT216 as the only options.

Yes you are right,re-badged isn't a word to explain 5870M,but i was really expecting more from it

 

[rodney_ws]

Just think how much longer the 5870 is than previous gen cards... one of the main things pushing that extra length is the heat generated by the card and the cooling system required to keep that in check. There was just no way anything remotely similar to the 5870 desktop part was going to fit in a laptop. Am I disappointed in the specs of the 5870 mobile part? Sure. Am I surprised? No, not really. Sure, they could have put something faster in the 18-20 inch laptops out there... those have the size (and potential cooling capacity) to handle something like that, but those are such low volume systems... I'm sure AMD/ATI knows what it's doing. Just be thankful this is NOT a 4000 series re-badge and that it has DX11 support.

 

[rodney_ws]

I do have a question though... is there a thermal penalty by having a wider memory bus? Was the decision to go with the 128 bit bus purely for cost reasons or were there thermal considerations as well that went along with that?

 

[TheGreatGrapeApe]

I think the two 64 bit controllers built on die for the HD5770 limited the bus more than thermals, although it does use a bit more power (but maybe 5-10W). It would be nice to have 256bit, but even 128bit w/ GDDR5 is still pretty nice compared to the old 256bit GDDR3, and should draw less power even while providing more bandwidth.

 

[ubernoobie]

Why not just make the mobile 5870 1600 shaders and 600 core while the mobile 5850 gets 1440 shaders @ 500 core, both with 256bit memory, I'm sure they could of done that if they have refined the processing and make the space beside each transistor almost empty

 

[fgocards]

Reading up on this, and I have to be doing something wrong. Calculating FLOPS, or processing power, but I don't think I'm doing it right because I'm getting this:

2(ATIs work per cycle)x800(Their stream processors)x700(the shader clock)=1.12 TeraFLOPS of power, that sounds right...

But then I try to put it up against the 280m...

3(Nvidias work per cycle)x128(Their stream processors)x1463(the shader clock)=.562 TeraFLOPS.. that doesn't sound right, that would mean it takes 2 280m's in SLI to equal 1 5870! I wish, but theres no way thats true.

Regarding the 5870 though, I'm satisfied. I mean yeah, I expected more, but 1.12 TeraFLOPS is 1/2 of a desktop 5870, which means whatever it can do, this laptop can do half. I'm very happy with the 40nm design (Great for OCing) and DX11 so ATI has satisfied me; but by all means not made me happy.

Already pre-ordered a G73JH

EDIT: BTW, I was in an email convo with Josh from XoticPC and he said Asus is making their 5870's with 256bit memory bus, not 128 like the specs say. I even showed him the link of the specs for the 5870 and he said he was aware, but the G73JH is made with GDDR5 on a 256 bit bus.

 

[TheGreatGrapeApe]

Could be, like the HD4870 mobile which was based on the HD4770.

Although it does mean at the chip level past the memory interface it's still running into two 64bit memory controllers unless they changed the design somewhat;

http://techreport.com/articles.x/17747

Potential for direct access from the RBEs? So not global access but access for specific things like AA and texture filtering functions.

 

[TheGreatGrapeApe]

Looks like ATi says it's 128bit so Josh might want to re-check his facts, especially if he's telling customers it's 256bit memory.

I'm just a little concerned about one thing, the HD5770 has no DoublePrecision support, so I'm assuming without any major change to the GPU that will be the case with the mobile version too, which is a little annoying. While I no longer work in that department I'd still like to tweak with GPGPU tools to stay fresh and this seems like a limiting factor.

 

[Maziar]

Reading up on this, and I have to be doing something wrong. Calculating FLOPS, or processing power, but I don't think I'm doing it right because I'm getting this:

2(ATIs work per cycle)x800(Their stream processors)x700(the shader clock)=1.12 TeraFLOPS of power, that sounds right...

But then I try to put it up against the 280m...

3(Nvidias work per cycle)x128(Their stream processors)x1463(the shader clock)=.562 TeraFLOPS.. that doesn't sound right, that would mean it takes 2 280m's in SLI to equal 1 5870! I wish, but theres no way thats true.

Regarding the 5870 though, I'm satisfied. I mean yeah, I expected more, but 1.12 TeraFLOPS is 1/2 of a desktop 5870, which means whatever it can do, this laptop can do half. I'm very happy with the 40nm design (Great for OCing) and DX11 so ATI has satisfied me; but by all means not made me happy.

Already pre-ordered a G73JH

EDIT: BTW, I was in an email convo with Josh from XoticPC and he said Asus is making their 5870's with 256bit memory bus, not 128 like the specs say. I even showed him the link of the specs for the 5870 and he said he was aware, but the G73JH is made with GDDR5 on a 256 bit bus.

Don't forget to write a review when it arrives

 

[lostandwandering]

I'm just a little concerned about one thing, the HD5770 has no DoublePrecision support, so I'm assuming without any major change to the GPU that will be the case with the mobile version too, which is a little annoying. While I no longer work in that department I'd still like to tweak with GPGPU tools to stay fresh and this seems like a limiting factor.

Yes, everything that I have heard points to no Double Precision. And yes, I agree, Justin really needs to double check before he gets too many people excited about the Asus G73.

 

[fgocards]

Don't forget to write a review when it arrives

Oh for sure. I'll be writing for days on it. Lol, now the anticipation is going to kill me waiting for it.

 

[TheGreatGrapeApe]

So an idea of how the new solutions match-up from 3Dcentre including desktop reference points;



I'm thinking the MHD5770 will likely be the solution that matches my needs, and the MHD5850/5870 being the ones I would 'settle for' if I had to. Not sure about the MHD5830 though if I'd want the restricted GDDR3 with the still high power consumption. Either I'd want the power/heat savings of the MHD57xx series, or else go whole hog with the MHD5850 or above. I can't see myself going the GDDR3 route for any solution anymore.

Just hope they get some nice packages for about the time I come back from Cuba in Feb, even though I usually don't spend money during ski season.