RV250 benchmark - slower???

[T2k]

Just straight from Rage3D:

"Radeon 9000 Benchmarks!!

Just got off a friend from Hong Kong, A Weekly Machine PC Weekly has posted some benchmarks on the new Radeon 9000 ( aka RV250 ). The Card is a Grandmars Radeon 9000 Pro and it is engineering sample. I am thinking that some webmaster may have this in hand for testing as well but I decided to post this for those who are anxious to see it.

As his scanner is not working so no picture will be post, but I am sure other readers from HK could verify the truth of this benchmark.

Here is some Basic info on RV250

Manufacturing Process: 0.15 - micro
4 Rendering Pipeline
Memory Bandwidth 9.6GB/sec with 600+ ( DDR ) MHZ
Core Speed 300MHz +
Pixel Shader 1.4
Vertex Shader 1.1
HydraVision
Integrated 2 RAMDAC ( 165MHz TMDS Transmitter and a Digital RGB Port )
Pin Compatible with Radeon 7500
AGP 4X

Since Radeon 9000 is Pin Compatible with Radeon 7500 which mean manufacture could easily upgrade their existing production line to Radeon 9000. And by doing that could also lower the cost of production.

The design of Radeon 9000¡¦s PCB is slightly different from the Radeon 8500, since Radeon 9000 is a integrated solution design, which means that there is less component on the board,
Such as the Rage Theater and the secondary external RAMDAC. By design Radeon 9000 is cheaper to make than the Radeon 8500 which makes it more attractive to customers.

On the current engineering sample it is running at a 290 MHz (Core&Mem)
The Chip is version A12 and the size is the same as RV200

Benchmark

3D Mark 2001 SE 330
RV200 RV250 MX460 8500LE
1024 x 768 5549 7548 5931 8475
1280 x 1024 4315 5987 4591 6915
1600 x 1200 3286 4652 3506 5376
Aquamark
1024 x 768 25.2 41.6 25.2 46.8
1280 x 1024 20.8 29.1 19.8 36.9
1600 x 1200 15.5 21.1 14.6 26.9
Quake 3
1024 x 768 116.4 160.7 163.8 183.1
1280 x 1024 73.7 105.6 115 139.9
1600 x 1200 52.7 73.4 81.3 92
Jedi Knight II
1024 x 768 87.9 92.2 92.6 90.9
1280 x 1024 65.7 86.9 80.5 91.3
1600 x 1200 46.8 66.2 60.6 82.2

Tested with:
Athlon 1800+
256 DDR 266
WinXP
Driver 7.74

I personally think the benchmark is a bit doggy since RV250 is very simliar to R200 than why is the benchmark so different??

But please bear this in mind this is a enginerring sample with very early driver......"

Duno, but if a card has same parameters but faster than another how could be SLOWER in the very same test???

I'm confused...

 

[MfA]

So its certain that ATI's numbering has also fully degenerated into "if the number is this high it has to be good" mode?

 

[hughJ]

not sure what's worse.. a number that doesn't really represent anything, (and possibly slightly misleading).. or the Ultra Pro Super Delux edition naming scheme..

 

[T2k]

So its certain that ATI's numbering has also fully degenerated into "if the number is this high it has to be good" mode?

What you gonna say?

 

[Joe DeFuria]

If the RV250 is in fact a 4 pipe board, and assuming the same architecutre as R-200, I'm puzzled as well.

RV250 at 290/290 falls significantly behind R-200 at 250/250 ?

On the other hand, the 290/290 RV-250 (9000 "pro") seems clearly as good or better than the MX460, which should be the direct competitor for this board. If the "non pro" version of the RV-250 (clock speed?) competes with the MX 420/440 directly on price, we'll finally have a decent DX8 board at value pricing and quantities.

 

[Dolemite]

What's in a name?

Guys, whatever they decide to call it, we're all going to know its exact specifications and performance anyway, so we're not going to be misled into thinking its a DX9 part.

Its true, the average consumer may not know these things, but probably also doesn't know that ATI at one time related DirectX version to its naming conventions. Full compatibility with DX9 vs. DX8 probably also won't make a damn bit of difference to the average consumer either.

I do think the Geforce4 MX thing is something of a travesty, not only because it implies something more capable than it is and may well mislead the average consumer, but also because nVidia had the opportunity to produce a shader-capable budget chip and opted not to. Sure, a GF4MX with shaders might have killed GF3 sales, but I think it would have really helped in the adoption of shaders and the need to support them in games.

 

[T2k]

But I'm still confused:

If a card has same capabilities, a refreshed one of another but faster how could be SLOWER in the very same test???

There should be something secret...

 

[opy]

Why do you sure it has the same capabilities as 8500?

What about HyperZ? Maybe it has a total lack of it. Don't forget that 9000's manufacturing is much cheaper than the 8500's according to the sources. How? I think it has significantly less transistor count than 8500. This could be the reason of the higher clock speed.

 

[Dave Baumann]

If its being sold as a low cost part but it has basically the same functionality and more integration then something must have gone - it could be something like the cache sizes have been slashed or something in that region.

 

[Doomtrooper]

But I'm still confused:

If a card has same capabilities, a refreshed one of another but faster how could be SLOWER in the very same test???

There should be something secret...

It has alot LESS transistor count than a R200, a value part that is giving excellent speed and Direct X 8.1 Full Compliance Pixel Shader Suppport for the same price as a Geforce MX 460 ??? What else do people want here...as stated a 100 times this chip is designed for the VALUE segment, and in some cases will outperform a R200 if Hercules etc..adopt BGA ram and exotic cooling and sell it as some Ultra Rv250, all for 120 US dollars.

 

[DemoCoder]

Both naming schemes suck. With the GF4, we are to assume that all of the models share the same core features, just different performance. That was true of the former cards, where the GF2MX was just a vastly lower performance model of the GF2.

With the Radeon, we were to assume that the first digit gave us the DirectX feature set that was being supported. In the worst case, all models with the same leading digits would essentially have the same core rasterization features, just different performance. Creating a 9000 that is DX8 and a 9500 that is DX9 is hellishly confusing just like the GF4 MX. Unless they name the R300 the "10000", consumers will be confused. Only hard core users like us will know the real deal.


The Pro/Ultra monikers were not confusing, and were essentially just different performance models of the same basic card.


Video cards should work like cars. If BMW introduced a new model called the M6 and a cost reduced 4-door model called the M6i, I would expect them to have the same basic engine, just different handling/performance. I would not expect to have the engine from the entry level vehicles.

Why would I expect this? Because I will see slick car commercials with the M6, but realize that I have a family and I need a sedan. So in buying the cheaper, M6i, I will feel like I am getting some of the benefit of the real M6, just in different packaging and slightly worse performance.

Likewise, consumers are going to look at the GF4 or the R300 and think that by buying the cheaper version, they are somehow getting a lower performance /cheaper version and they are not.
[/code]

 

[Dolemite]

I had never even heard of this "first digit = DX version" deal until the past month or so when speculation about the new names surfaced.

To me, that guarantees that Joe Consumer is going to be absolutely clueless about the issue. In many ways I think that's OK, though. Half the computer gamers I know wouldn't know DirectX if it was f*cking their mom, let alone what version they have installed, let alone what version their hardware is compliant to.

So what difference does it make if Joe Consumer gets a good deal on a Radeon 9000, and doesn't know he's not getting PS2.0 and full DX9 compliance that he'd get paying $300 for an R300 card?

To me, none whatsoever.

 

[T2k]

I had never even heard of this "first digit = DX version" deal until the past month or so when speculation about the new names surfaced.

You didn't hear it - and what does it prove?
You did not hear it - but it's the system since last summer.

PS: First number is the highest DX-COMPLIANCY, second one sorrespond to memory speed.

 

[Geeforcer]

 

[Dolemite]

I had never even heard of this "first digit = DX version" deal until the past month or so when speculation about the new names surfaced.

You didn't hear it - and what does it prove?

It doesn't prove anything, but the point I was making is:

1. The people whom DX compliancy matters to already know the score.
2. To people whom DX compliancy makes no difference to, naming conventions will be equally meaningless.

Sure it might be helpful if there was some industry standard way of signifying such compliance levels, but what chance is there of that in this industry where we have to fight over HLSLs and the sort?

Trust me, the world is used to meaningless numbers in product names. I remember back in the fifth grade when the poor kids who got free lunch wore the most generic, uncool shoes, called (and be ready to be mad) XJ-9000's. Did the number 9000 have any meaning? Perhaps it was their family's yearly income? Or the number of BJ's their mom gave each year? Nope. No meaning. Sorry!

Sure, ATI set a kind of stupid precedent to break, but I really doubt its going to cause the widespread confusion you guys are implying. And with the Fire GL 8700 and 8800, I'm not sure where an RV250 board would fit in the 8xxx series...but does it really matter? Its all just PR stupidity anyway, nomatter what they would have called it.

Does Radeon mean anything? What about GeForce?

 

[T2k]

You're right, I was too lazy to look for...

Anyway, I wasn't correct maybe with the memory speed but the thing is about 2nd digit something similar, isn't?

 

[Doomtrooper]

People want to argue over a name ??? I really don't know what to say besides rediculous.
Sure ATI naming scheme is no different than Nvidias Geforce 4 MX, yet there is a difference at least ATI is delivering a video card for the value market that will not continue to keep the general card population in DX7 days.

 

[Waltz King]

This is not the first time ATI ignores its own naming scheme. The Radeon 7000 lacks a T&L-engine, so it's not a full DX7-part like the original Radeon (AFAIK now called Radeon 7200) or any GeForce-based card.
So the Radeon 7000 should be called Radeon 6x00 or something like that.

And there is also the Radeon 8500 LE - ATI says that the last three digits represent the relative performance of the chip within its generation. The 8500 LE has got the same numbers as the regular 8500, so it should show the same performance - but it doesn't, it is slower.
So the Radeon 8500 LE should be called 8400 or something like that.

So this branding of the RV250 as "Radeon 9000" (if true) is not a real surprise as ATI does not seem to care about the own "logical" naming scheme they are so proud of...

Of course it doesn't really matter - most users will not know and will not care as well, and we (being the well-informed hardcore users) just cannot be fooled

(Sorry for my not-that-perfect English, guys!)

 

[jvd]

well

well the slide says generation , so really if radeon was the 7th gen , the radeon with no tnl would be the slowest thus the 7000 , the 7500 would be the fastest thus the 7500 number. the radeon 8500 well , i can't defend the le crap so whatever ... but then again the rv250 is another gen part so thus the 9 and the r300 would be 10... who knows if i'm right

 

[Mintmaster]

It has alot LESS transistor count than a R200, a value part that is giving excellent speed and Direct X 8.1 Full Compliance Pixel Shader Suppport for the same price as a Geforce MX 460 ??? What else do people want here...as stated a 100 times this chip is designed for the VALUE segment, and in some cases will outperform a R200 if Hercules etc..adopt BGA ram and exotic cooling and sell it as some Ultra Rv250, all for 120 US dollars.

At least someone here has rational thinking. You are very right in nearly everything you said. RV250 is out to bring DX8.1 to the mainstream - the real mainstream. Geforce3 Ti200 could hardly be called mainstream (I think they're just getting rid of stock now), nor could the Ti4200. I don't know what the launch price will be for the RV250, but you can bet that prices will drop down to around $50-$60 within a few months.

Essentially ATI is planning on knocking out the Geforce4MX, and they will do a very good job of it. This card, as well as its competition (SIS Xabre, NVidia's NV18 I think, when it comes out), will make developers create games with significant pixel shader usage.