• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

GF 100 (Fermi) Technology Preview

Status
Not open for further replies.
Associate
Joined
2 Mar 2009
Posts
1,653
Location
London
Anyone seen this?

2n9vub9.jpg


Review link here http://www.guru3d.com/article/nvidia-gf100-fermi-technology-preview/1 (Image is on page 4)

If this is true, it looks like HD5870's prices is coming down soon.:p
 
Last edited:
GF100 will be up against 5970. Not the 5870. The mere fact that they compare their GPU to a smaller RV870 says that they cant win the halo back.

And i wonder when we are going to see the GX2 version of the GF100 given thats its supposingly power hungry.
 
Considering that Fermi is supposed to do tessellation on the shaders, these performance gains will only be in specific scenarios.

When the shaders are under heavy load, tessellation performance will probably have a similar drop compared to the gains it's made.

From what I've heard, the 5 series' tessellation performance should be much more consistent as the same tessellation block is used in all 5 series chips and is completely independent from the shader clusters.

Great if it's as good as it looks on paper, however, I feel like this is just like one of those performance charts that don't start at 0 and make a 10% gain look like a 50%+ gain. *cringe*
 
Not when Fermi retails at £400 :p

My word, someones optimistic indeed :p

AS said, we have no idea how that will translate into performance in game as theres nothing in their papers that suggests tesselation gets its own set of hardware, its shared hardware most likely, run physx/graphics load through at the same time and we have no idea what will happen.

Also all lower end versions from a 360gtx, down to a 310gts will have lowered tesselation performance as the hardware is linked to shader cluster's, of which lower end cards have less. AMD have an equal sized tesselator(dedicated hardware) on a 5870 and a 5350(or whatever their bottom end card ends up being) meaning a 5850/5830/5770 can give exactly as much tesselation performance as a 5870 which is great for budget users.

Now that is 1.6x in best case scenario, not only will that be lower on lower SP versions, it will be significantly worse on lower clocked versions, with rumours abound that the only ones to be available, realistically ever, being 448sp parts at some 25% lower clock speeds than those alledgedly tested in these previews, and that 1.6x advantage, becomes a 1.2x advantage in best case, if those numbers drop hard in real world usage, suddenly a 380gtx won't be able to out tesselate a 5350.......... not exactly good.

Back to that £400, I sincerely doubt the card benchmarked in these previews(which is likely a 1.6Ghz shader clocks and 512sp part) will be anything under £450, could quite easily be well over £500 and basically not available. A far slower 448sp 1.2Ghz part might be all you can ever buy, and thats likely to cost well in excess of £350.
 
I love the fact that they have no shame in hiding the actual scores. I also love how they only show a specific part of the benchmark. This says a lot on its own.
 
ye its hardly confidence inspiring to show performance figures from half? a benchmark, having said that i couldn't care less about the lower end parts giving significantly reduced performance with tesselation, its gtx 380 + 3D vision all the way for me
 
As others have said this could just be a specific part of the benchmark.

Even benchmarks / games have better performance in certain sections depending on the card rendering them.

Until I see some real world tests im not going to be convinced by how good the card could be.
 
most negative comments in this thread come from Ati owners

yup, 5850, of course, theres a 8800gtx up on the shelf in my room in a box I kept putting off selling.

I have no problem buying whatevers good, however, price/performance is my key concern. I don't care if a single core from ANY company is the fastest, if you can match that performance with really any number of slower cores........ for a lower price, thats where my money goes.

If you read any threads, other than giving Nvidia a little support and trying to explain to Nvidia guys who are waiting and AMD guys, exactly why the delay(partly their fault, a lot to do with TSMC), I also mention the idea of Nvidia going for a new architecture, which they DO need. They need to switch up to an AMD strategy, small cores, cheaper to make, cheaper to sell, sell more of them and make more profit.

AMD's revenue/profit has gone through the roof this year, Intels gone up hugely, Nvidia went up (due to previous year being a recession) but no where near as much. If Nvidia go for a smaller style core, then we all get more choice, more price competition and more value.

If Nvidia's 260gtx was able to be sold at 4890 prices with similar profit margins, it would have meant price competition for us all.

I'm all for Nvidia success giving us ALL better performance, anyone that thinks it will happen with Fermi is deluded though. You can't make a core thats 60% bigger, cheaper, its literally impossible and therein lies the problem.
 
I am interested to see what these can do ... I tend to agree with the notion of cherry picked results as shown here (if only because real numbers are rare).

If the tesselation is sharing compute power in the shaders then I ask if the Heaven benchmark is using the shader too at that point and if so would this not actually show indicated performance for real-world tesselation?

(On a side note does anyone know the tesselation algorithm used by Nvidia?)
 
Last edited:
Status
Not open for further replies.
Back
Top Bottom