Poll: GT300 VS. HD5870

[Mem]

Originally posted by: happy medium
gt300 will avg 22% faster and cost 27% more

I reckon 22% bigger and you'll need 27% larger case .


Personally I think around 10-15% faster,pricing and availability date are the crucial factors IMHO.

 

[dguy6789]

Originally posted by: OCguy

Originally posted by: ronnn
looking for the gt300 will likely match the 5890 when it is readily available option.

So a completely new arch and adding SIMD and DDR5 will match a doubling of specs of the 4890?

Interesting prediction. Hopefully we get to see who is right before Xmas.

The completely new NV40 merely matched the doubling of specs of the R300. My prediction is that the GT300 will be faster and more expensive than the RV870 but not faster than the 5870x2 and probably not cheaper than the 5870x2.

 

[ronnn]

Originally posted by: OCguy

Originally posted by: ronnn
looking for the gt300 will likely match the 5890 when it is readily available option.

So a completely new arch and adding SIMD and DDR5 will match a doubling of specs of the 4890?

Interesting prediction. Hopefully we get to see who is right before Xmas.

Think I was too oblique. I was trying to say that I don't expect the gt300 on store shelves until the 5890 is getting close. Likely an exaggeration, but who knows. Not at all a comment about what I expect from the 5890, besides 2gb ram. :beer:

 

[bryanW1995]

Originally posted by: zagood
I believe it will be a rebrand of the 9800GTX+.

Some guy named Charlie told me.

:laugh:

edit: @ ronnn; that would make sense for ati to actually release a 5890 with some teeth to it. faster clocks with just a few optimizations like 4890 would be nice, but faster clocks, optimizations, and 2gb would really help 5890 sales imho.

 

[Janooo]

Originally posted by: MODEL3

Originally posted by: Idontcare

Originally posted by: MODEL3
The same gap.
GT300=32ROPs.
GT300=512bit memory controller=double 5870 bandwidth. (1,2GHz GDDR5)
GT300=higher TU / ROP ratio & higher Flops/s / ROP ratio than GT200.

I think overall the same gap.

Does that mean you are giving the new architecture no presumed performance benefits over GT200 architecture, or you think GT300 architecture is the same/similar as that of GT200 insomuch as Cypress architecture is the same/similar to that of RV770? (plus all the DX11 circuits to make it work of course)

I mean the same gap, like:

The gap between GT300 and 5870 will be the same,
as the gap between GT200 and RV770.

For me GT300 will have in relation with GT200:
1,1X pixel rate
1,75X texel rate
2,5X Flops rate
2,2X vertex rate

I wrote the day of the 5870 launch that the 5870 results are strange.
And that according to my perception, there are other reasons except bandwidth limitations and driver maturity, that the 850MHz 5870 hasn't doubled its performance in relation with a 850MHz 4890.

Usually when a GPU has 2X the specs of another GPU the performance gain is 2X (of cource i am not talking about games with engines that are CPU limited or engines that seems to scale badly or are poor coded for example)
There are many examples in the past that we had 2X performance gain with 2X the specs. (not in all the games, but in many games)

From the tests that i saw in Ryan's review and from my understanding of the 5870 architecture in general, i think there are 2 more reasons that 5870 performs like that.

The day of Ryan's review, i wrote to the forums the additional reasons that i think the 5870 performs like that, but nobody replied me.

I wrote that probably 5870 has:

1.Geometry/vertex performance issues (in the sense that the classic vertex unit cannot generate 2X geometry in relation with 4890) (my main assumption)

or/and

2.Geometry/vertex shading performance issues (in the sense that the geometry shader [GS] cannot achieve 2X speed in relation with 4890)(another possible assumption)

My assumption is that ATI though that many of the future DX11 games will use vertex techniques based on the tesselator unit.

I guess there are synthetic benchmarks that have tests like that (pure geometry speed, and pure geometry/vertex shader speed, in addition with the classic pixel shader speed tests) so someone that has a 5870 can see if my assumption is true.

5870 delivers 2.6x4890 in CoH at the top resolution and settings.

 

[lopri]

IMO GT300 will be fast. I mean, it should be fast. NV didn't really make anything new since G80.. in a sense, and I'd like to think they didn't waste the past 2 years. Good performance is a given, me thinks.

However, the bar has been raised somewhat from my point of view. There are hordes of fixes/features HD 5870 brought to the table and GT300 will also be judged by those as well.

For example, according to AT review HD 5870 got rid of squealing noise under load. I've seen many users who got a brand new card yet wondered it's defective because of that noise. Well, I'd expect all future video cards to not squeal under load.

 

[imported_Shaq]

The GT300 would have to, at least, be some 20-30% faster than a 5870. AMD is going straight to the X2's and not producing a 5890. If they thought a slightly faster 5870 was all they would need to beat the GT300 that is what they would be working on first. So it is more likely that the 5850 X2 will be the competitor to the GT300, which indicates that the GT300 will be some 60-70% faster than the 5850. I believe that is about 20-30% faster than a 5870. The 5870 X2 will give AMD the lead, about 30-40%, until the GT300 GX2 comes out next year, which will be 30-40% faster than the 5870 X2. But by then there may well be a 5890 X2 which will compete directly with the GT300 GX2, but I believe the GX2 will still win slightly.

 

[Cookie Monster]

It all depends on how nVIDIA prioritized the importance of gaming performance over GPGPU performance.

 

[bryanW1995]

wow, that's pretty detailed. did you know that 4850x2 was unofficial? sapphire made it on their own.

ati is certainly working to optimize the 5870 while ALSO working on 5870x2. Those 2 are not mutually exclusive, you know. Why would they put all their chips in one basket? Nvidia has shown in the past that they are determined to claim the top dog spot. ati used to do the same, but picked a really good time to go to a small ball strategy. That strategy has forced them to be more efficient imho than nvidia with their recent cards. If ati can build a cheaper gpu and keep it competitive for the 2nd round in a row then there will be some serious teeth gnashing in Santa Clara.

 

[Kakkoii]

Originally posted by: OCguy

Originally posted by: ronnn
looking for the gt300 will likely match the 5890 when it is readily available option.

So a completely new arch and adding SIMD and DDR5 will match a doubling of specs of the 4890?

Interesting prediction. Hopefully we get to see who is right before Xmas.

Did you mean to say "MIMD"? Cause that's the purported rumor that GT300 is moving to MIMD.

 

[MODEL3]

Originally posted by: Janooo
5870 delivers 2.6x4890 in CoH at the top resolution and settings.

Like i said, i know the results.
In this particular case the 2,6X figure is for 2560X1600 with 8X AA & 16XAF.
If you drop the antialiasing at 4X the difference at 2560X1600 16XAF is 1,95X.
And the DX10 codepath of CoH is extremely pixel shader oriented.
All the above are predictable and natural results.

I am not here to state the obvious.
I am trying to figure out what are the limitations of the Cypress because it shows according to my perseption strange results.
My analysis is complex and to write down it will take me more time than it took to figure out the situation.
Sorry i am not willing to lose time.
If you disagree with me that the 5870 has (except the obvious driver and bandwidth related issues) the following issues:

Originally posted by: MODEL3
1.Geometry/vertex performance issues (in the sense that the classic vertex unit cannot generate 2X geometry in relation with 4890) (my main assumption)

or/and

2.Geometry/vertex shading performance issues (in the sense that the geometry shader [GS] cannot achieve 2X speed in relation with 4890)(another possible assumption)

My assumption is that ATI though that many of the future DX11 games will use vertex techniques based on the tesselator unit.

Then i respect your opinion, but i am not willing to do anything about it.

I just thought to point out what the reviewers with all their knowledge,year in this business, technical background, briefings from ATI, 5870 cards at hand, failed to point out.

Sure i don't have technical background, nor i was at ATI briefings and of cource i don't have a 5870 card in order to be able to test my theory.
So with all these disadvatntages i may be wrong.
The funny thing will be, to be right about it.
And no, it will be no luck at all, trust me.

 

[bryanW1995]

have you been hanging out with nemesis?

 

[MODEL3]

Originally posted by: bryanW1995
have you been hanging out with nemesis?

lol,

i would certainly like to.

I suppose you said it because of this?

Originally posted by: MODEL3
Then i respect your opinion, but i am not willing to do anything about it.

 

[Janooo]

Originally posted by: MODEL3

Originally posted by: Janooo
5870 delivers 2.6x4890 in CoH at the top resolution and settings.

Like i said, i know the results.
In this particular case the 2,6X figure is for 2560X1600 with 8X AA & 16XAF.
If you drop the antialiasing at 4X the difference at 2560X1600 16XAF is 1,95X.
And the DX10 codepath of CoH is extremely pixel shader oriented.
All the above are predictable and natural results.

I am not here to state the obvious.
I am trying to figure out what are the limitations of the Cypress because it shows according to my perseption strange results.
My analysis is complex and to write down it will take me more time than it took to figure out the situation.
Sorry i am not willing to lose time.
If you disagree with me that the 5870 has (except the obvious driver and bandwidth related issues) the following issues:

Originally posted by: MODEL3
1.Geometry/vertex performance issues (in the sense that the classic vertex unit cannot generate 2X geometry in relation with 4890) (my main assumption)

or/and

2.Geometry/vertex shading performance issues (in the sense that the geometry shader [GS] cannot achieve 2X speed in relation with 4890)(another possible assumption)

My assumption is that ATI though that many of the future DX11 games will use vertex techniques based on the tesselator unit.

Then i respect your opinion, but i am not willing to do anything about it.

I just thought to point out what the reviewers with all their knowledge,year in this business, technical background, briefings from ATI, 5870 cards at hand, failed to point out.

Sure i don't have technical background, nor i was at ATI briefings and of cource i don't have a 5870 card in order to be able to test my theory.
So with all these disadvatntages i may be wrong.
The funny thing will be, to be right about it.
And no, it will be no luck at all, trust me.

What part of silicon performs geometry? Is it doubled from 4890 to 5870 or not?
Geometry is not an issue. The BW is the first in line.

 

[OCGuy]

Originally posted by: Kakkoii

Originally posted by: OCguy

Originally posted by: ronnn
looking for the gt300 will likely match the 5890 when it is readily available option.

So a completely new arch and adding SIMD and DDR5 will match a doubling of specs of the 4890?

Interesting prediction. Hopefully we get to see who is right before Xmas.

Did you mean to say "MIMD"? Cause that's the purported rumor that GT300 is moving to MIMD.

Yes, I wrote that post @ 1:53AM PST.

 

[MODEL3]

Originally posted by: Janooo

Originally posted by: MODEL3

Originally posted by: Janooo
5870 delivers 2.6x4890 in CoH at the top resolution and settings.

Like i said, i know the results.
In this particular case the 2,6X figure is for 2560X1600 with 8X AA & 16XAF.
If you drop the antialiasing at 4X the difference at 2560X1600 16XAF is 1,95X.
And the DX10 codepath of CoH is extremely pixel shader oriented.
All the above are predictable and natural results.

I am not here to state the obvious.
I am trying to figure out what are the limitations of the Cypress because it shows according to my perseption strange results.
My analysis is complex and to write down it will take me more time than it took to figure out the situation.
Sorry i am not willing to lose time.
If you disagree with me that the 5870 has (except the obvious driver and bandwidth related issues) the following issues:

Originally posted by: MODEL3
1.Geometry/vertex performance issues (in the sense that the classic vertex unit cannot generate 2X geometry in relation with 4890) (my main assumption)

or/and

2.Geometry/vertex shading performance issues (in the sense that the geometry shader [GS] cannot achieve 2X speed in relation with 4890)(another possible assumption)

My assumption is that ATI though that many of the future DX11 games will use vertex techniques based on the tesselator unit.

Then i respect your opinion, but i am not willing to do anything about it.

I just thought to point out what the reviewers with all their knowledge,year in this business, technical background, briefings from ATI, 5870 cards at hand, failed to point out.

Sure i don't have technical background, nor i was at ATI briefings and of cource i don't have a 5870 card in order to be able to test my theory.
So with all these disadvatntages i may be wrong.
The funny thing will be, to be right about it.
And no, it will be no luck at all, trust me.

What part of silicon performs geometry? Is it doubled from 4890 to 5870 or not?
Geometry is not an issue. The BW is the first in line.

Certainly BW in general is first in line, when we examine the potential reasons.
I said that a 800-850MHz 32ROPS/1600SP design with 256bit memory controller (with the current GDDR5 ICs) will be bandwidth limited before 1,5 freakin month.
This is extremely easy to figure out.
But this kind of performance behavior isn't only BW related.
According to my perception Geometry is an issue, so i guess we have to agree to disagree about this matter.

 

[alcoholbob]

Reviewers focus so much on high AA/AF benchmarks, the 5870 might end up looking a LOT more competitive with GT300 simply due to reviewer convention. I say this because nvidia cards can take as much as a 30% hit in performance with 8xAA, whereas ATI this generation is under 10%.

 

[SlowSpyder]

Originally posted by: Astrallite
Reviewers focus so much on high AA/AF benchmarks, the 5870 might end up looking a LOT more competitive with GT300 simply due to reviewer convention. I say this because nvidia cards can take as much as a 30% hit in performance with 8xAA, whereas ATI this generation is under 10%.

People generally want to run AA/AF at high levels. Reviewers like those settings as they stress the cards more, to help highlite video performance. I'd factor that into my purchasing decision as well.

 

[dguy6789]

Originally posted by: SlowSpyder

Originally posted by: Astrallite
Reviewers focus so much on high AA/AF benchmarks, the 5870 might end up looking a LOT more competitive with GT300 simply due to reviewer convention. I say this because nvidia cards can take as much as a 30% hit in performance with 8xAA, whereas ATI this generation is under 10%.

People generally want to run AA/AF at high levels. Reviewers like those settings as they stress the cards more, to help highlite video performance. I'd factor that into my purchasing decision as well.

Indeed. If game reviewers were to benchmark every card at 1024x768, one would get the impression that they are all about the same performance level. You really have to crank up the details when you get these extremely fast cards if you don't want a CPU being the limitation for FPS.

 

[alcoholbob]

Well think of it this way, is the 4890 and the GTX285 at the same level? Because if you ran a benchmarks all at at 8xAA/16xAF, the difference between the two cards goes from 20% to 5%. Is the 4890 95% of a GTX 285?

Let's say all things equal (AA algorithm efficiencies remain the same), then even if the GTX380 is 50% faster than the 5870, if you ran every benchmark at 8XAA/16XAF, given the performance loss ratios (33% vs 10%), the GT300 would only be 10% faster. Now what if the GT300 is only 40% faster under no AA/AF conditions? Then the 5870 would be the faster card across the board by 5% for reviewers that test under heavy conditions. Now the 5870 is the "faster card."

So the question is, how much of an influence does nvidia and ati have on test conditions with individual reviewers?

 

[Idontcare]

Originally posted by: Astrallite
So the question is, how much of an influence does nvidia and ati have on test conditions with individual reviewers?

Its probably the other way around, there are soooo many tier-2 and tier-3 professional reviewers and wanna be sponsored reviewers floating around out there right now (this is a recession, unemployment is high, lots of people with lots of free-time to freelance just sitting at home) that I wouldn't be surprised if the judgement that gets clouded is on their behalf while they strive to please their temporary sponsor for spotting them some review gear.

Meaning we need not stoop so low as to imply or invoke active coercion in my opinion, more than enough people will be self-starters and jump to that all on their own hoping to get one degree closer to the source of the review hardware (and be in their good graces) for the next review cycle.

I've no doubt that AMD and NV send out a review "checklist" of things they'd like to see or recommend the reviewer take the time to checkout in addition to whatever else they have planned. A guide only makes sense as it helps people rough-out that brainstorming portion of the review design phase.

But there may not be any need for active coercion by these guys, look around and see what people are willing to say and do for free as posters on forums (not even getting review samples or anything)...is it so hard to imagine it is the reviewers who might be the source of any bias in the reviews?

 

[yacoub]

Needs an option for "The first-gen cards from both companies will be too large, run too hot, cost too much, and be too loud. I'll be waiting for the refresh in 4-6 months."

 

[RussianSensation]

Originally posted by: Astrallite
Reviewers focus so much on high AA/AF benchmarks, the 5870 might end up looking a LOT more competitive with GT300 simply due to reviewer convention. I say this because nvidia cards can take as much as a 30% hit in performance with 8xAA, whereas ATI this generation is under 10%.

More likely than not, NV will address 8AA performance with the new architecture. They must be aware that they are uncompetitive with their efficiency in that mode. Also, they have a great shot to up the memory bandwidth on their card which will help with 8AA as well.

 

[Avalon]

I think the performance gap will remain the same at around 20%.

 

[ronnn]

Kyle thinks gt300 is meaningless. A little exteme, but if true will be very hard on nvidia.