Fermi's lead over Cypress shrinks w/ new drivers

[v8envy]

When you think about it, many people with enough money to buy a 5870 or 480 usually want the best and will purchase a 2nd or 3rd for SLI/CF.

Plenty of people desiring the highest performing video solution (read: 30" or multi-monitor or 3d stereoscopic) may also be sensitive to input lag, notice microstutter or play titles which either have no multi-GPU profiles or exhibit poor (possibly negative) multi-GPU scaling.

So for them the only viable choice is the single fastest GPU. Which ATM is either the GTX480 or the 5870, overclocked as high as it'll go.

I agree there's a lot of value in seeing multi-GPU results in benchmarks. Most reviews already include the highest end hardware pairs though.

 

[railven]

This article might interest you. Look at graphs for resolutions at 1080p or greater.

http://www.xbitlabs.com/articles/cpu/display/cpus-and-games-2010_4.html#sect0



Also read: http://www.tomshardware.com/reviews/balanced-gaming-pc-overclock,2625-8.html (Intel cpu scaling)
http://www.tomshardware.com/reviews/balanced-gaming-pc,2477-8.html (AMD CPU's with previous gen gaming cards)

Ah thank you very much. With that info in hand and this current article there is just no way I can let the GF get a GTX 465 over a 5850.

The power draws and heat are the killers. Not worth it for the small gains (or based on this article, losses) if you ask me.

 

[OCGuy]

Did AMD even claim to have a performance increase with the 10.5s?

 

[v8envy]

That, more than anything, is what this shows (although it makes it in no way truer than it was before, but it's worth repeating anyway).

I don't agree. What it shows is: if you put the competing cards into two otherwise identical machines with water cooling and have a typical person play a wide selection of games at playable settings they will not be able to reliably determine which hardware is in which machine.

If that's the case then why put up with one of the solutions broiling your toes while it deafens you and costs more to both buy and operate?

 

[railven]

Did AMD even claim to have a performance increase with the 10.5s?

Nope, which the article addressed. Which makes these results even more surprising and as you've seen caused Medium to not be so Happy

 

[v8envy]

Ah thank you very much. With that info in hand and this current article there is just no way I can let the GF get a GTX 465 over a 5850.

The power draws and heat are the killers. Not worth it for the small gains (or based on this article, losses) if you ask me.

Every review out there points out the 465 competes with the 5850 on price, but on performance it compares to the uberfail 5830 or an overclocked 5770. While using as much juice as a 480.

Just from that standpoint it should not be considered. Pretend it and the 5830 don't exist, look at the viable parts -- 5850, 470, etc.

 

[SlowSpyder]

Every review out there points out the 465 competes with the 5850 on price, but on performance it compares to the uberfail 5830 or an overclocked 5770. While using as much juice as a 480.

Just from that standpoint it should not be considered. Pretend it and the 5830 don't exist, look at the viable parts -- 5850, 470, etc.

With 5830's as low as $210AR they are getting closer to where they should be, but still too much in my opinion. With 5770's start at $130AR, the 5830 really needs to drop yet... But compared to where they launched at they are at least getting closer to a sane price.

 

[Grooveriding]

The real test would be to pit 2 x 480s up against 2 x 5870s. I don't think 1 x 480 is a smart buy. It is only when 2 are bought where the real benefits begin solely due to the extreme scaling which is given. When you think about it, many people with enough money to buy a 5870 or 480 usually want the best and will purchase a 2nd or 3rd for SLI/CF. This would be the real scenario to test.

I'm not a fanboy of either camp. I am first and foremost an enthusiast. Everyone who knows me knows this. I recently sold 3 x 5870s to test the waters with Nvidia's latest offering. With a single 480, I wasn't too content. Now, with 2 x 480s, I appreciate them as much as I did 3 x 5870s. Plus, I get the benefits of CUDA based GPGPU applications. This was actually the "main" reason I switched to the 480s. It was for Adobe's Mercury Playback Engine in their latest CS5 offering.

So, I beg Anandtech to do a test between 2 x 5870s and 2 x 480s. I will also plead that they do so with Nvidia's next WHQL driver. That way we can have 10.5 whql vs 257.xx whql.

I couldn't agree more and have said this as well. On it's own the 480 is a piss poor offering that is not worth it, it becomes a decent solution when you purchase two.

The 480 is about 10-15% faster than a 5870 on it's own, not this repeated nonsense of 25-30%. Stop cherry picking one benchmark again and again. There are benches where a single 5870 is faster than a 480, but can we call it a faster card, no.

I almost sold my 5870s earlier this week, I was offered $750 for the pair and I would of gone and gotten two 480s. I changed my mind when I figured I would probably only have them for 3 months until September when ATI's new cards come out, so took a pass.

Still mulling over doing it though...

 

[railven]

Every review out there points out the 465 competes with the 5850 on price, but on performance it compares to the uberfail 5830 or an overclocked 5770. While using as much juice as a 480.

Just from that standpoint it should not be considered. Pretend it and the 5830 don't exist, look at the viable parts -- 5850, 470, etc.

The GF is a known nVidia Fanboi if you would. I've already tried to talk her into getting a 5850 (evening offering to pay for half) and the GTX 470 is out of her allowed budget (and I personally refuse to pay for half of that) so the GTX 465 was her next option.

I'll be waiting for the 460 and its numbers. If anything I just might get her a GTX 275 as a place holder.

 

[OCGuy]

I purchased a 5850 and immediately regretted it when the 470 was at Compusa for $70 below MSRP the next day


Always happens to me though.

 

[v8envy]

Ah, the 465 is a card for the Faithful so it sounds perfect for her. If NV's marketing has convinced your GF that theirs is the only product worth buying then they should be compensated.

The 465 is pretty much a 275 with DX11, so from that standpoint I can't see saving $30 on an expensive card to get a 275 over a 465. Or either over a 4890 for $175, but let's not go there.

Seeing as many 470s are $306 with BCB on tigerdirect this very moment I really have a hard time understanding why saving $40 is worth the leap down to a 465. She'll save a few bucks on power over the coming years and her ears will be thankful. If she's going to buy an expensive part, talk her into buying one that makes some sense.

Edit: I see the 465 is $246 with BCB. Still a horrible value compared to a $264 5850 IMO.

 

[railven]

Ah, the 465 is a card for the Faithful so it sounds perfect for her. If NV's marketing has convinced your GF that theirs is the only product worth buying then they should be compensated.

The 465 is pretty much a 275 with DX11, so from that standpoint I can't see saving $30 on an expensive card to get a 275 over a 465. Or either over a 4890 for $175, but let's not go there.

Seeing as many 470s are $306 with BCB on tigerdirect this very moment I really have a hard time understanding why saving $40 is worth the leap down to a 465. She'll save a few bucks on power over the coming years and her ears will be thankful. If she's going to buy an expensive part, talk her into buying one that makes some sense.

Firstly, anything ATI is off limits, as she just won't listen to reason.

So going from that, I'd prefer she not get the 465 or 470 due to their heat output. She isn't running the best airflow in her CM Scout to begin with. She isn't a big PC-person and is just getting tired of the slowdown her 9800 GTX is giving her in some of her favorite games (such as Sims 3, upgrading her CPU, mobo, and RAM didn't give her the performance boost she wanted.)

I'm going to try to hold her off until the GT(X/S) 460 is announced. If anything she is either going to get a GTX 200 or nothing at all.

 

[v8envy]

Here's how to handle that situation. Get a 5850 for your machine if you don't already have one. Overclock it, and your CPU if it's not already cranked to 11.

Have her try her games on YOUR machine. Point out she's on her own with NV, but is guaranteed to have that exact level of performance with ATI. Then either pass on the 5850, or get her one as well.

 

[HumblePie]

Decent little article. More in line what I was looking for.

However, if I was doing a review website, I would do things as thorough as possible.

1) Have all cards stock and overclocked maxed air/water.
2) Use both single and multiple card setups.
3) Do both an AMD and Intel platform. Both max and "average" user setups.
4) Do all the demo and synthetic benchmarks for all resolutions past 12x10.
5) Do them 3 times each. Show min and average for all 3 and across all 3.
6) Show no AA/AF and medium level and max level of AA/AF for everything I can.
7) Devise real world game play done in each game reviewed from 3 spots random areas of the game. Make sure the test is consistent across all platforms.
8) Compile all the results.


Done this way, I doubt there would be any room for doubt or arguing anymore.

 

[railven]

Here's how to handle that situation. Get a 5850 for your machine if you don't already have one. Overclock it, and your CPU if it's not already cranked to 11.

Have her try her games on YOUR machine. Point out she's on her own with NV, but is guaranteed to have that exact level of performance with ATI. Then either pass on the 5850, or get her one as well.

The stealth upgrade has been recommended to me already haha.

I've seriously considered that route. I just might...

 

[RussianSensation]

xbit is ok, ...
ANd metro 2033 they failed to use aa because thats where the lead will increase.

They didn't fail to use AA. It's clearly stated in the Xbit article that AA Reduces image quality in Metro 2033, making the textures blurry. Tessellation was not used either because the GeForce GTX 480 is the only graphics card able to deliver a playable speed at resolutions above 1600x900.

Based on current pricing, 5850 competes with GTX465, which is a laughable position for NV. 5870 easily justifies the $30 price premium over GTX470 since the latter tanks at 2560x1600, which does not bode well for future more intensive games, as well as being more power hungry, louder and hotter.

One important aspect of the Techspot review is that: "Of course, we have changed our testing methodology for this article by using FRAPS to measure performance, while the focus was more on real-time gameplay instead of time demos."

 

[tincart]

If these results persist across a wider range of games, then the real interesting thing is exactly how close the current generation of cards are to each other in terms of performance. Outside of certain notable anomalies, of course.

 

[HurleyBird]

Very interesting results. Begs the question if they are due to better test methodology are if they are just outlier results. It's nice that stuck to newer more resource intensive games, as there is absolutely no reason to benchmark L4D on these cards like certain sites do.

As far as the guy stating that the 5870 and 480 would be best tested in pairs, I doubt it. 2x HD 5870 would already put out an uncomfortable amount of heat, but any more than 1 480 is simply impractical no matter how you look at it.... unless you're in the arctic and want to have a bloody space heater next to you under your desk

 

[Compddd]

The stealth upgrade has been recommended to me already haha.

I've seriously considered that route. I just might...

Headbutt her and then tell her she is getting a 5850 if she likes it or not.

 

[Jacky60]

I addressed the game you linked to from AT, Crysis.


Yes, typically the GTX470 and GTX480 are faster than their more expensive counterparts, and it can be up to 25%, but even in a single game it can vary from being slower to being faster.
Which means that you can put forward support for either side and play the value and performance game until you're blue in the face, and argue both sides equally, but at the end of the day, what matters more for an individual user is not that the GTX480 can be on average 25% faster if you use the right benchmark, or that the HD5850 can be faster than the GTX470 if you use the righ benchmarks,
QUOTE]


This sounds very much to me like the author is being paid by Nvidia to sound reasonable while selling a wholly inaccurate corporate message. Nvidia cards (470 / 480) ARE MORE not LESS expensive than their broadly equal or superior performing competitiors (5850/5870) and are NOT 25% faster.

 

[NoQuarter]

xbit is ok, Look at the rest of the benches, the gtx 470 is beating the 5850 by more then 5% OVERALL. Much more.
I'm not talking games that are allready old games like resident evil where it really dosent matter. There over 100fps anyway.
ANd metro 2033 they failed to use aa because thats where the lead will increase.

If you read the article they said they chose not to use AA because it worsens the textures. Apparently the nVidia Metro 2033 AA bug still exists?

 

[luv2increase]

Nvidia cards (470 / 480) ARE MORE not LESS expensive than their broadly equal or superior performing competitiors (5850/5870) and are NOT 25% faster.

In a single-GPU config, the 480 is not on average 25% faster than the 5870. In a 2-way SLI/CF config, the 480 is over 25% faster than the 5870.

Simply put, if you don't plan on using CUDA capable GPGPU applications and don't plan on using more than 1 GPU in your system, the 5870 is the way to go. However, if you plan on using CUDA capable GPGPU applications and plan on using a multi-GPU setup, the GTX480 is the way to go.

I don't know what or how Nvidia did it, but the multi-GPU scaling for Fermi is AMAZING.

I also forgot to mention 3D gaming. If you plan on that, then the GTX480 is where it's at.


When you look at things from a pure user utilization perspective, the 5870 does not succeed. I had 3 of 'em all the way up to the 10.3 preview drivers from the 9.10 rc6 drivers, and they are great cards, but they simply even come close the the GTX480.

Eye-candy is another advantage of Fermi over Cypress. The picture quality is better---> Fermi>Cypress>GT200>4xxx series ---> when it comes to picture quality.

 

[zerocool84]

What I find funny is that people are arguing if a 9 month old card is faster then a 2 month old card. Seems like Nvidia didn`t do a good enough job on their cards that such an old card can still trade blows.

 

[Madcatatlas]

So from a performance perspective, do 2x 480s in sli, beat 3x 5870s in trifire?


do they?

 

[luv2increase]

So from a performance perspective, do 2x 480s in sli, beat 3x 5870s in trifire?


do they?

Here is a quote of mine off of EVGA:

Alright folks, with the help of the 257.15 beta drivers, I've surpassed the 40k mark in 3dmarkvantage for the first time ever! I don't really care much for Vantage anymore but nonetheless, it is still cool IMO.

Here is the best run I've ever had with 3 x 5870s. Everything else is held constant. The CPU, RAM, et.. are all the same. The only thing different between the runs are the GPUs and their drivers of course.


EVGA GTX480 "Vanilla" x 2 in SLI @ 850/1700/2025 @ 1.138v
GPU Score - 35826
http://service.futuremark.com/compare?3dmv=2214135

XFX 5870 x 3 in Tri-Fire @ 1000/1295
GPU Score - 36922
http://service.futuremark.com/compare?3dmv=1693329


As you can see, the tri-fire setup beat out the SLI setup by "only" 1096 points! The tri-fire run was with more mature drivers than that of the EVGA run above as well. I guarantee that I'll surpass 36922 with the same clocks and better drivers from Nvidia here real soon.

Remember, both setups had exactly 150mhz overclock.