Laptop equivalent of the 4870?

Is there one? And if so, what would it be? I'm not talking about an exact match, I'm just talking in general performance.

 

5870 would be a close match

 

Technically a mobile 4870 with GDDR5. But they were never made The mobile 4870 and desktop 4870 use the same core, but mobile card makers only ever put GDDR3 on them, so they performed like HD4850s.

Yeah the 5870M is the closest match.

 

What about a 6990m

 

Well obviously it's faster.

 

Overclcoked HD5870m is a close match to the HD4870. Altho technically, it is on par with an HD4850 stock (both cards have the exact same specs overall, tho less clock speed on the HD4850, but same memory bandwidth).

HD6990m is faster.

 

If you assume the mem controllers and shaders are identical between the 4 and 5 series then yes.

But they are not.

5870M should pull ahead of the 4850 consistently.

 

So a laptop 6990m is faster than a desktop 5870?

 

No it's not.

 

The 6990m is close to matching the desktop 6870 which in turn is slower than the desktop 5870 but probably just a little faster than the desktop 5850. AMD made some odd naming choices with the 6000 series card in their desktops... The 5850 was faster than the newer 6850, the 5870 was faster than the 6870. The 6950 was a little faster than the 5870. The 6970 was faster than any single card in the 5000 series.

There is a chart on tom's hardware that gives a pretty good comparison of all modern gpu's, no hard numbers included and sometimes its not specific enough to answer all questions but they update it each month.

Graphics Card Hierarchy Chart : Best Graphics Cards For The Money: August 2011

 

Actually the 4870 is only about 8k Vantage GPU score. So an OC'ed 5870 is much faster.

 

HD4870 with older drivers had anywhere form 8 to 9k vantage scores. the P set uses 1280x1024 which does not benefit from the extra memory bandwidth the HD4870 has, and at stock the GPU generally had 750mhz clock so they perform on par.

An OC'd HD5870m will perform better obviously, but at higher resolutions you will see diminishing returns compared to the HD4870 due to having almost twice the memory bandwidth.

Not to mention a HD4870 can obviously be OC'd too. HD4850 is the one that acts close first but never really catches up.

Actually, they are largely the same. There was not a significat architecture overhaul from 4000 to 5000 series. Sure the 5000 series are more powerful, but not like the change from 5000 to 6000. In DX9 and DX10 the HD4850 performs close to HD5870m but will never overcome it because on stock it is actually lower clocked.

Overclock both to similar speeds and you will see a bit similar performance, but obviously not exactly the same. Hardwarewise they are largely the same, even if the HD5870m is more optimized for better power management, usage, and performance.

I did the testing with my old desktop rig and my laptop and I was pleasantly surprised that the performance was close. With the HD5870m being the victor because thanks to power improvements, not only does it have stock better performance, but it can be OCd much higher with ease. Not to mention the hardware capability for DX11.

 

Maybe a while ago, but look at modern titles, the 5770 pulls mostly level with the 4870.

The shaders were tweaked for the way games were going to be rather than the ones that had gone before.

4xxx series shaders are very different from 5xxx series shaders. DX11 capabilities aside.

 

But there's more differences than just tweaked shaders. The 5770 has more transistors (1.04B vs 956M) and better, faster memory (4.8 Gbps vs 3.6 Gbps).

 

From the Wiki of the Radeon architecture:
"With the release of Cypress, the Terascale graphics engine architecture has been upgraded with twice the number of stream cores, texture units and ROP units compared to the RV770. The architecture of stream cores is largely unchanged, but adds support for DirectX 11/DirectCompute 11 capabilities with new instructions. [4] Also similar to RV770, four texture units are tied to 16 stream cores (each have five processing elements, making a total of 80 processing elements). This combination of is referred to as a SIMD core."

Link: http://en.wikipedia.org/wiki/Evergreen_(GPU_family)

So no, the stream processors aside from DX11 and newer instructions are not really that much different. They changed their approach to the stream processors in the HD6000 series.

In DX9 and DX10, you will see quite similar performance.

the HD5870m, just as the HD5770, did not follow the rule of "twice the rops, memory units" etc. Thus their architecture is largely the same, and is mostly tweaked for power efficiency and stability to have higher clocks.

This is why HD5000 varians have much higher clocks. And same reason why the HD5870m can overclock its core like crazy.

And the HD4870 uas a 256 bit bus helping it achieve higher memory bandwidth. In best cases, the extra core horespower of the HD5770 allowed it to overcome the HD4870, but having less bandwidth limited in just key scenarios.

More complex achitecture, faster ram, power management and 128 bit bus only allowed to create a cheaper, faster (core wise), less hot version of the HD4870.

 

^^^^^^^^

That man and wiki speak the truth. Get two CF 6970M's in CF it's about on par with the desktop 58070's in CF..a tad slower on the mobile end I think. However that is more than enough HP for 1080P resolution gaming....

 

58070...lol

 

Adding DX11 was not a case of tacking on features, there are fundamental changes to those shaders and in modern titles even under different DX levels it makes a difference, regardless of what some monkey on wikipedia wrote.

 

Where PC games not fundamentally be bound by APIs due to the different hardware versions, you would be right in what they could do with the added features.

I didn't say they tacked it on, but it is true that their achitecture was not dramatically changed in such series. They didn't have to, in a way. Trust me on this one, the set of instructions was expanded and things that had to be done in a way, could be done now with much ease, the hardware capabilities increased. Not changed or redone from scratch.

Due to how the DX API works, if they use the DX11 set, they have access to all those extra capabilties. They can't properly use such newer capabilities in DX9 for example. (In consoles, they happen to forgoe that because they don't even use DX, and the have the freedom to custom made every single engine and software development).

I think you are either trying to put the HD5000 series too high, or trying to diminish what the HD4000 series were. The truth is that they took what they had, upgraded the instructions it could handle, and added new ways to perform what it already could to make the architecture more efficient without changing its main structure that much. At low res, same speeds, an HD4870 and HD5770 would perform similar, with theHD5770 pulling ahead a bit, but not by any big margin. In practice, the match is quite heated and the HD4870 wins its share of the battles even today. It's an awesome card.

This is not like the nvidia case of GTX480 with the GTX460 when they changed the architecture to superscalar within the same series to increase gaming performance.

Long story short, the HD5000 series are HD4000++ and it took what was there and added more, while tweaking how it handle the old a bit. that's it. I am not saying there isn't a difference at all. there is, but you are making it sound like a huge difference, when it isn't. Are HD5000 better? of course, do they largely resemble the same architecture? Yes, they do. Are they identical? Nope. Even tho when it first came out, the HD5770 was actually a tad bit slower than the HD4870.

this isn't even about wikipedia, it's reflected in reviews of the cards, editors opinions and analisys etc.

 

Yet if you look where the 4870 and 5770 place in modern titles, even with half the memory bandwidth the 5770 can pull ahead, so I still dispute that nothing of note was done for performance.

 

HD 5770 does not have half the memory bandwith, and it is also higher clocked. Where are you getting your numbers than HD5770 is pulling ahead? What modern titles are you talking about? They perform the too close to really tell who is the best hands down. when it came out HD4870 was the better performing card. Nowadays they are the same at best specially at the resolutions we play.

And I didn't say NOTHING was done for performance, but it was done very little. The fact is, the HD5770 sold not because it was an upgrade in performance, but because it had similar performance, good price and better extras like eyefinity, dx11 etc.

 

I kinda skipped all the mumbo jumbo gfx debate. But thanks for all the information. That gfx card list helped a lot. Was trying to see where my chip was in the whole thing. ( Reason being, i saw a video of Diablo 3 maxed with a 4870 and was wondering how a laptop would compare.)

 

Excellent link!! Thanks for the info.

 

In desktop performance terms, your 555M is a 9800 GT