ATI M98: The Fastest Mobile Graphics Processor in the Making

[btarunr]

PCPop has access to an upcoming mobile (notebook) graphics processor (mGPU) from ATI, codenamed the M98. This graphics board makes use of the RV770 core and will be accompanied by 512 MB of GDDR3 memory, up to 1 GB in some models. It could be branded under the Radeon HD 4000 series. It can be dubbed the fastest till date if it's pitted against NVIDIA's current high performance mGPU, the GeForce 9800M GTX which makes use of the G92 GPU with a similar configuration to that of the desktop 8800 GT / 9800 GT cards (112 shader units). It then becomes comparable to GeForce 8800 GT versus Radeon HD4850 where the latter obtains 15% to 20% performance gains.

Pictures show the M98 Axiom Package featuring the the RV770 GPU next to four GDDR3 memory chips made by Qimonda. The GPU is marked "M98 XT", it is an engineering sample.



View at TechPowerUp Main Site

 

[kaneda]

Pfft, I'm still waiting for a sub 500gbp gaming laptop.

 

[robspierre6]

15-20% performance gains?

The 4850 is 15-20% faster than the 9800gtx not the 8800gt.
http://www.tomshardware.com/reviews/radeon-hd-4870,1964-10.html
http://www.techreport.com/articles.x/14990/13
http://www.techreport.com/articles.x/14990/11
http://www.techreport.com/articles.x/14990/9
I hope you recheck your numbers before posting them.

 

[btarunr]

The 4850 is 15-20% faster than the 9800gtx not the 8800gt.
http://www.tomshardware.com/reviews/radeon-hd-4870,1964-10.html
http://www.techreport.com/articles.x/14990/13
http://www.techreport.com/articles.x/14990/11
http://www.techreport.com/articles.x/14990/9
I hope you recheck your numbers before posting them.

I'll use TPU reviews as my point of reference. Since I have those numbers off hand, I'll not make changes.

 

[KainXS]

The 4850 is 15-20% faster than the 9800gtx not the 8800gt.
http://www.tomshardware.com/reviews/radeon-hd-4870,1964-10.html
http://www.techreport.com/articles.x/14990/13
http://www.techreport.com/articles.x/14990/11
http://www.techreport.com/articles.x/14990/9
I hope you recheck your numbers before posting them.

btarunr means that the 9800M GTX dosen't have 128 sp's or the memory bandwidth of the desktop variant, so when you take take into account that it has 112 sp's and memory bandwidth near the same as the 8800GT then its obvious that it will perform similar to an 8800GT.

 

[PCpraiser100]

I should've stopped my friend from buying that Gateway FX Laptop lol.

 

[Wile E]

The 4850 is 15-20% faster than the 9800gtx not the 8800gt.
http://www.tomshardware.com/reviews/radeon-hd-4870,1964-10.html
http://www.techreport.com/articles.x/14990/13
http://www.techreport.com/articles.x/14990/11
http://www.techreport.com/articles.x/14990/9
I hope you recheck your numbers before posting them.

Please stop trolling the news posts. You keep posting only the reviews that show the 4850 in the best light, yet dismiss all the numerous reviews that say otherwise.

Besides, that has nothing to do with mobile gpus. Mobile gpus have far different requirements than desktop models, one of which is reduced power consumption, and the need to run cool with much less heatsink and airflow. That's usually achieved via reducing voltage and downclocking, and sometimes reducing the number of sp's. There's no guarantee these will be 800sp parts.

 

[WarEagleAU]

why, arent they good?

 

[robspierre6]

Please stop trolling the news posts. You keep posting only the reviews that show the 4850 in the best light, yet dismiss all the numerous reviews that say otherwise.

Besides, that has nothing to do with mobile gpus. Mobile gpus have far different requirements than desktop models, one of which is reduced power consumption, and the need to run cool with much less heatsink and airflow. That's usually achieved via reducing voltage and downclocking, and sometimes reducing the number of sp's. There's no guarantee these will be 800sp parts.

Well, the poster of the article said that the 8800gt is 15-20% slower than the 4850.
Secondly,I put 2 links to two reliable and well known sites,so don't tell me where to get my infos from.

 

[xfire]

My brother need's a new laptop and all most all the laptops have the GMAx3100. Very few models have AMD proccy's and the Puma platfrom hasn't come to India
@robspierre6 TPU reviews are far more trustable than most other websites.

 

[Wile E]

Well, the poster of the article said that the 8800gt is 15-20% slower than the 4850.
Secondly,I put 2 links to two reliable and well known sites,so don't tell me where to get my infos from.

The second part of my post wasn't directed at only you. It was directed at bta as well. You can't use desktop cards as a basis of comparison for mobile gpus.

And I didn't tell you where to get your info. I said you conveniently dismiss any info that doesn't agree with yours. Big difference. I find Tom's reviews to be inaccurate in many cases, thru none other than my own experience. My numbers and findings always differ from theirs. Never really looked into techreport reviews. I have found tpu to be one of the more reliable sources of benchmark information, and tpu's benchmarks put the 4850 only slightly ahead of the 9800GTX. I'll take that over Tom's any day.

Anyway, I'm not going to drag this any further. We have no need to clog the thread any further with this.

 

[btarunr]

It's relative. Sure MGPU's get a brutal dose of underclocking, but the GPU configurations stay the same. G92 has come in a variety of configs: 96 SP + 192bit mem bus , 112 SP + 256bit, 128 SP + 256bit. And these various configs went into making the GeForce 9800M series (barring the 128 SP config). The highest offering from NVIDIA uses a the 112 SP + 256bit mem bus config. These configs exist in their desktop flavours as well. However, RV770 has not seen config variations so far. So we can lay our best word on the 800 SP + 256bit config since there's not been a desktop variant so far. Mobile variants follow desktop parts (as was the case with G92 based NVIDIA mGPUs.)

@xfire: HCL does sell a Puma based notebook, don't you see those countless Vishwanathan Anand ads?

 

[Wile E]

It's relative. Sure MGPU's get a brutal dose of underclocking, but the GPU configurations stay the same. G92 has come in a variety of configs: 96 SP + 192bit mem bus , 112 SP + 256bit, 128 SP + 256bit. And these various configs went into making the GeForce 9800M series (barring the 128 SP config). These configs exist in their desktop flavours as well. However, RV770 has not seen config variations so far. So we can lay our best word on the 800 SP + 256bit config since there's not been a desktop variant so far. Mobile variants follow desktop parts (as was the case with G92 based NVIDIA mGPUs.)

But it's still irrelevant to mGPU. mGPU uses different clocks and voltages than their desktop cousins. What if they have to undervolt and underclock this due to heat reasons, until it's slower than the 9800m GTX? As it stands, it's an irrelevant comparison, at least until we get more info set in stone.

 

[btarunr]

They'd rather use a RV670 than run RV770 slower than a 9800M. I'm not the only one calling these "the fastest in the making", many base it in the fact that RV770 does manage to beat 8800 GT at stock parameters, though assuming despite the underclock/undervolt it still manages to beat the underclocked/undervolted G92 in the 9800M GTX. Some used exorbitant values such as "20 ~ 40 % moar!1!", I stuck to the more realistic-looking TPU scores of 15~20%.

 

[Wile E]

They'd rather use a RV670 than run RV770 slower than a 9800M. I'm not the only one calling these "the fastest in the making", many base it in the fact that RV770 does manage to beat 8800 GT at stock parameters, though assuming despite the underclock/undervolt it still manages to beat the underclocked/undervolted G92 in the 9800M GTX. Some used exorbitant values such as "20 ~ 40 % moar!1!", I stuck to the more realistic-looking TPU scores of 15~20%.

Well, it's just my opinion on the matter. I'll personally wait to see some numbers before I make any judgment calls.

 

[xfire]

@xfire: HCL does sell a Puma based notebook, don't you see those countless Vishwanathan Anand ads?

not with much attention.

At IGP level ATI always had the best chips, if not performance wise they had a lot of features and lower power consumtion. The 7xxx series of IGP from Nvidia didn't even support dual channel memory:shadedshu
edit-This one is supposed to come out with a newer platform
http://en.wikipedia.org/wiki/AMD_mobile_platform#Shrike_platform

 

[Nkd]

I have read through comments, and It is really sad that people think that hd 4850 in reality is really only 14% faster than 8800gt, I have a gtx 280, and I am really not biased but have you guys seen the results of hd 4850 with AA on, does it not kick the living crap out of 8800gt and 9800gtx, if someone wants I can put reviews from several sources, like anandtech, firingsquad, techreport, and so on, and does it not leave the 9800gtx eating dust with 8xAA on and in some games catches up to the gtx 260,(firingsquad.com), I buy what I like but I do not stand for such comments where the card is not given the credit it deserves, it is gets faster and faster under demanding scenarios than the 9800gtx and 8800gt, enough said.

well the tpu review depends on the games they test, but if I look at those numbers they tell me that gtx 280 is 38% faster than 9800gtx, hmm how realistic is that really? I get almost 70-80% increase in peformance in almost all games at 1920x1200 4xAA on average and some games even give me twice the performance. they test games that never show the true potential of the new cards, it is those games that add to the average of cards like 9800gtx and reduce the performance difference between these cards.

 

[GPUCafe]

Wil E, you say that the 4850 is bound to be underclocked but ignore the fact that the 9800M GTX is underclocked too.

9800M GTX
Core: 500MHz
Shader: 1250MHz
Memory: 800MHz

8800 GT
Core: 600MHz
Shader: 1500MHz
Memory: 900MHz

Also the lead for 4850 depends on where you look, for example: http://www.computerbase.de/artikel/...00_gt_9800_gt/23/#abschnitt_performancerating

1280x1024: 4850 is 14% faster than 8800GT
1600x1200 8xAA/16xAF: 4850 is 97% faster than 8800GT

 

[Wile E]

Wil E, you say that the 4850 is bound to be underclocked but ignore the fact that the 9800M GTX is underclocked too.

9800M GTX
Core: 500MHz
Shader: 1250MHz
Memory: 800MHz

8800 GT
Core: 600MHz
Shader: 1500MHz
Memory: 900MHz

Also the lead for 4850 depends on where you look, for example: http://www.computerbase.de/artikel/...00_gt_9800_gt/23/#abschnitt_performancerating

1280x1024: 4850 is 14% faster than 8800GT
1600x1200 8xAA/16xAF: 4850 is 97% faster than 8800GT

I took the nVidia card being underclocked into account. The fact remains that we can't make a comparison based on the desktop cards, until we know more about this mGPU. There just isn't enough info.


And the lead in that chart at 16x12 8xAA cannot be right. The 4850 gets more frames at 16x12 8xAA than it does at 16x12 4xAA. That's a bugged score if I ever saw one.

On average, the 4850 is around 20-25% faster. It's up to 50% faster at some resolutions. Unfortunately, those are at resolutions not commonly found in notebooks, so it's kind of a moot point.

 

[GPUCafe]

I took the nVidia card being underclocked into account. The fact remains that we can't make a comparison based on the desktop cards, until we know more about this mGPU. There just isn't enough info.

Agreed.

And the lead in that chart at 16x12 8xAA cannot be right. The 4850 gets more frames at 16x12 8xAA than it does at 16x12 4xAA. That's a bugged score if I ever saw one..\

Thats not a bugged score, you are reading it wrong. Those arent frame-rates but percentages.

On average, the 4850 is around 20-25% faster. It's up to 50% faster at some resolutions. Unfortunately, those are at resolutions not commonly found in notebooks, so it's kind of a moot point.

1920x1200 is available in 15.4", 17" notebooks where this GPU will most likely offered. Hardly a moot point.

 

[Wile E]

Even still that 16x12 8xAA is bugged. The 9600GT scores the same as the 8800GT, when we both know the 8800 is clearly faster.

And there aren't any 1920x1200 results in that page. There's 1600x1200 and 2560x1600. Going in between the resolutions it's impossible to figure out where it would fall. AT 16x12 8xAA, they claim a 97% advantage from the 8800GT to the 4850, yet at 2560x1600 4xAA, there is only a 58% advantage. Those results are just inconsistent.

So yeah, those charts remain a moot point.

 

[GPUCafe]

Even still that 16x12 8xAA is bugged. The 9600GT scores the same as the 8800GT, when we both know the 8800 is clearly faster.

This might come as a shock to you but the 9600GT is very close to 8800GT. Look at the TPU chart btarunr posted above. There are some scenarios where both of them have virtually the same performance, you cant dismiss charts if they show reality.

And there aren't any 1920x1200 results in that page. There's 1600x1200 and 2560x1600. Going in between the resolutions it's impossible to figure out where it would fall. AT 16x12 8xAA, they claim a 97% advantage from the 8800GT to the 4850, yet at 2560x1600 4xAA, there is only a 58% advantage. Those results are just inconsistent.

Sigh, at higher resolutions the frame buffer or memory bandwidth might be holding the 4850 back. Leads dont always increase, you can look up 10 such examples and yet you would be saying that the charts are "bugged".

So yeah, those charts remain a moot point.

Sorry I was preaching to the choir.

 

[Darkrealms]

This is good for the market. Granted its newer tech than Nvidia's current top end but that means ATI is continuing to stay competetive, WOOT!

robspierre6, honestly please stop trolling. I'm an admited Nvidia and AMD (CPU) fan and I can still be objective in my posts. For the sake of all of us that are here to learn, comment, and help each other. Please stop.

 

[robspierre6]

This is good for the market. Granted its newer tech than Nvidia's current top end but that means ATI is continuing to stay competetive, WOOT!

robspierre6, honestly please stop trolling. I'm an admited Nvidia and AMD (CPU) fan and I can still be objective in my posts. For the sake of all of us that are here to learn, comment, and help each other. Please stop.

What's your problem darkrealms?
I don't care who's fan you are. the poster said that a 8800gt is only about 15-20 5 slower than a 4850.
And it's realy sad how people until now can't post some responsible numbers when they are against nvidia.
trolling for whom?
I think that you, darkrealms, are a nvidia fanboi as you admitted in your comment and you don't like the comments against nvidia.

 

[Wile E]

This might come as a shock to you but the 9600GT is very close to 8800GT. Look at the TPU chart btarunr posted above. There are some scenarios where both of them have virtually the same performance, you cant dismiss charts if they show reality.

Key word = close, not identical.

Sigh, at higher resolutions the frame buffer or memory bandwidth might be holding the 4850 back. Leads dont always increase, you can look up 10 such examples and yet you would be saying that the charts are "bugged".

The 8800 has the same amount of frame buffer. It stands to reason that it would incur the same performance hits, if framebuffer were the issue. But either way, considering how different the results are between those 2 resolutions, we still can't extrapolate 1920x1200 performance numbers for the cards

Sorry I was preaching to the choir.

No need to get sarcastic, just find something else that supports your theories. I am not dismissing all evidence, just that particular chart. I'll gladly change my views if somebody comes up with a different source that shows 1920x1200 results. That way there are no questions about the differences and how they pertain to notebooks.