Anand's GTS250 review

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

taltamir

Lifer
Mar 21, 2004
13,576
6
76
I went into the AMD zone... it reminded me of alien / christ conventions... it was like the twilight zone, a distinct disconnect from reality... And that was when I was 100% AMD user (because they were just a better deal back then)
 

OCGuy

Lifer
Jul 12, 2000
27,224
37
91
Originally posted by: nemesismk2
I trust Anandtech reviews for an accurate GeForce GTS 250 review. I also trust Tech Report, Guru3D, AMD Zone, PC Perspective and PC Stats. I have used all of these over the last 8 years or so! :)

LOLOLOLOLOL!



:laugh:
 

josh6079

Diamond Member
Mar 17, 2006
3,261
0
0
The only thing I can think of is that a different time demo was ran with the GTS 250 1GB, since all of the other framerates on the other cards are identical to the first Multi GPU benchmarks.

MultiGPU Update Review

GTS 250 Review

We can't be sure, because Anandtech doesn't outline what they did in either review to get the numbers. All they show is which drivers and hardware were used.
 

chizow

Diamond Member
Jun 26, 2001
9,537
2
0
Originally posted by: phexac
In all other reviews, that is not the case. GTX260 is way out ahead of the midrange cards, where is should be. What's wrong with anand's review?
I've noticed this as well in the past and asked Derek about some of the variances. I think part of the problem is that Derek doesn't force Vsync off in the Nvidia drivers. He hasn't done this in some time, ever since he found it had some weird impact in Crysis with regard to frame rates. Its possible Nvidia's driver is employing some kind of frame rate smoothing or normalization if Vsync is not specifically forced off. I've asked him in the past to do some comparisons just to make sure there's nothing going on there but I don't think he has. Most other sites do force it off in the drivers.

In the first multi-GPU review I also asked him about the results for 2 of the games. It wasn't an issue with only the GTX 260, as it effected results for all high-end parts. Fallout 3 was essentially frame capped at 60 FPS average. Derek said he had turned off the frame rate cap by setting iPresentInterval=0, but he also used some custom tweaks that may have capped FPS inadvertently. Similarly with COD5, there's clearly some kind of 60 FPS frame cap he acknowledges with single-GPU only. Multi-GPU scales over 100% as a result. Again, most other sites do not show any such limit for COD5.

Anyways, it is what it is, if you don't trust it there's plenty of other reviews to compare. Overall I think Derek did a much better job this time around without any of the glaring errors from the last 2-3 months. Hopefully he doesn't revert to using archived benches older than a month or so though, as advancements in driver performance have been making significant gains for the last few months for both IHVs.
 

postmortemIA

Diamond Member
Jul 11, 2006
7,721
40
91
Originally posted by: nemesismk2
Originally posted by: garritynet
Originally posted by: postmortemIA
so, does the n-th iteration of turd polishing work?

When has the 8800/9800 ever been a turd?

I don't agree with calling the 8800/9800 a turd as well because my XFX 8800 GT gave me great performance.

I agree that 8800 were great. I still have two 8800 cards. However, they decided to stop making new designs after that.

This is what happened to USA auto BIG3 in late 90s and early 00s: they decided they don't have to do R&D anymore for most cars, but just polish them lil bit every year, do the marketing, and keep selling them. Eventually, they become very uncompetitive.

Same thing with NVIDIA, no DX10.1 support clearly means those are still 8800 series, just made bigger. Marketing is making products now.


 

Wreckage

Banned
Jul 1, 2005
5,529
0
0
Originally posted by: postmortemIA
Marketing is making products now.

They still have the fastest GPU and video card. Not to mention game physics.

So what's worse. NVIDIA still running old stuff or that ATI's new stuff can't beat it.

 

Paratus

Lifer
Jun 4, 2004
17,524
15,568
146
Wreckage if they don't make you a focus group member there is no justice in the world.

Your efforts are tireless. ;)
 

OCGuy

Lifer
Jul 12, 2000
27,224
37
91
Originally posted by: Paratus
Wreckage if they don't make you a focus group member there is no justice in the world.

Your efforts are tireless. ;)

Was there something inaccurate about what he said? Prove him wrong with some facts, no the focus group broken record.
 

Paratus

Lifer
Jun 4, 2004
17,524
15,568
146
There's nothing wrong with what he said. He's a completly biased Nv fan, yet pretty good about keeping his opinion separate from his facts.

I think they should offer him a spot in the focus group for the effort he puts forth for Nv.

He'd be good, more like keys than rollo, IMO.
 

Paratus

Lifer
Jun 4, 2004
17,524
15,568
146
On topic theres nothing physically wrong with the GTS250. The new board is nice and it competes well.

The only two marks it has against it is the new name on old, but current, tech and more importantly price.

It's not convincingly faster than the cheaper 4850 and the 4870 512 may end up close to the same price.

posted via Palm Life Drive
 

BladeVenom

Lifer
Jun 2, 2005
13,365
16
0
Originally posted by: SSChevy2001
Originally posted by: novasatori
huh interesting

in AT review 260=4850 or so and in the TR review 260=4870

:confused:

to be fair tho there aren't many benches
It's not interesting at all. The TR review is using a factory OCed Evga FTW Edition in this review

It speed are 670/1404/2304MHz, instead of the stock 576/1242/2000MHz. Yet the 4870 they use is running stock speeds, go figure. :confused:

+1. That doesn't make them look like biased fanboys, does it.
 

Flipped Gazelle

Diamond Member
Sep 5, 2004
6,666
3
81
Originally posted by: BladeVenom
Originally posted by: SSChevy2001
Originally posted by: novasatori
huh interesting

in AT review 260=4850 or so and in the TR review 260=4870

:confused:

to be fair tho there aren't many benches
It's not interesting at all. The TR review is using a factory OCed Evga FTW Edition in this review

It speed are 670/1404/2304MHz, instead of the stock 576/1242/2000MHz. Yet the 4870 they use is running stock speeds, go figure. :confused:

+1. That doesn't make them look like biased fanboys, does it.

I don't believe TR are fanboys, I think they were just lazy in that review.
 

Wreckage

Banned
Jul 1, 2005
5,529
0
0
Originally posted by: Paratus
There's nothing wrong with what he said. .... yet pretty good about keeping his opinion separate from his facts.

You do realize that you contradicted yourself from the first sentence to the second.

It doesn't take a focus group to see the clear facts in front of you.

Besides, I gladly smack down all the FUD here for free. :beer:
 

crimson117

Platinum Member
Aug 25, 2001
2,094
0
76
From here:

Not only that, but the flat-out dishonesty is that Nvidia gave its board partners 'special' boards to send to reviewers. They are not allowed to give out their own vanilla cards, they MUST use the special set supplied by Nvidia.

Why is this dishonest? Want to bet that those boards have cherry-picked chips and RAM that clocks to the moon? That they will do everything better than any card you will ever be able to buy? Basically, Nvidia supplied ringers to the press that are not representative of what you can buy, and forced OEMs to give them to review sites without telling them. The technical term is 'mushrooming', feed them [scatological reference deleted] and keep them in the dark.
 

Zap

Elite Member
Oct 13, 1999
22,377
7
81
Basically, Nvidia supplied ringers to the press that are not representative of what you can buy

Trust Charlie to put a spin into it.

NVIDIA supplied the review cards: True. The reason is that they specified 1GB cards with memory clocked at 2.2GHz. No partner had cards available (yet) with the required 0.8ns memory. Thus, NVIDIA had to make the cards themselves.

NVIDIA cards are not representative of what you can buy: False. A card you can buy at retail clocked at 738/1836/2200 will work just like one of the NVIDIA provided boards at 738/1836/2200.

With that being said, partners may also have cards that are "factory overclocked" as well as cards with slower (and thus cheaper) RAM. This is nothing new and not just an NVIDIA thing. Anyone here remember back when the Radeon 3870 were the top ATI cards? Those used GDDR4 clocked at... 2.2GHz or so? A couple of partners used cheaper GDDR3 clocked at 1.8GHz. Was that just a big scheme by ATI to rip off consumers? No.

That practice continues to this day. Radeon 4650 can come with DDR2 clocked at 800MHz... all the way to GDDR3 clcoked at 1800MHz. GeForce 9500 GT is similar, with DDR2 versions clocked at 667MHz as well as GDDR3 versions clocked at 1600MHz.

Why is anyone suprised at this? You think NVIDIA or ATI sent out DDR2 versions of those cards for review? Or did ATI send out GDDR3 versions of the Radeon 3870 for review?
 

SickBeast

Lifer
Jul 21, 2000
14,377
19
81
Originally posted by: chizow
Originally posted by: phexac
In all other reviews, that is not the case. GTX260 is way out ahead of the midrange cards, where is should be. What's wrong with anand's review?
I've noticed this as well in the past and asked Derek about some of the variances. I think part of the problem is that Derek doesn't force Vsync off in the Nvidia drivers. He hasn't done this in some time, ever since he found it had some weird impact in Crysis with regard to frame rates. Its possible Nvidia's driver is employing some kind of frame rate smoothing or normalization if Vsync is not specifically forced off. I've asked him in the past to do some comparisons just to make sure there's nothing going on there but I don't think he has. Most other sites do force it off in the drivers.

In the first multi-GPU review I also asked him about the results for 2 of the games. It wasn't an issue with only the GTX 260, as it effected results for all high-end parts. Fallout 3 was essentially frame capped at 60 FPS average. Derek said he had turned off the frame rate cap by setting iPresentInterval=0, but he also used some custom tweaks that may have capped FPS inadvertently. Similarly with COD5, there's clearly some kind of 60 FPS frame cap he acknowledges with single-GPU only. Multi-GPU scales over 100% as a result. Again, most other sites do not show any such limit for COD5.

Anyways, it is what it is, if you don't trust it there's plenty of other reviews to compare. Overall I think Derek did a much better job this time around without any of the glaring errors from the last 2-3 months. Hopefully he doesn't revert to using archived benches older than a month or so though, as advancements in driver performance have been making significant gains for the last few months for both IHVs.

If Derek tested with Vsync enabled, you would see a framerate cap at whatever refresh rate he used. Seeing as there were scores in excess of 116FPS at 1680x1050 for NV cards in that review, I'd say that you either:

- don't know what Vsync does (!)
- don't know how Derek writes his reviews

Please don't spin things to make it look as though the NV cards are being held back unless you know what you're talking about.
 

evolucion8

Platinum Member
Jun 17, 2005
2,867
3
81
I find very odd the benchmarks results in the AT review, I checked out reviews from other 6 sites and they all pretty much state the same thing:

1 - In VRAM limited situations, the GTS 250 1GB outperforms the 9800GTX+
2 - In non VRAM limited situations it's only as fast as a 9800GTX+, and sometimes slighly slower because of the VRAM latency or some strange wizardly
3 - When higher levels of Anti Aliasing are used with high resolutions, the HD 4850 shines except when is VRAM limited, those scenarios in the GTS 250 will not be limited by the amount of VRAM, but instead in raw power.
4 - Is competitive with the HD 4850, and it's slighly faster overall when compared to the 9800GTX+, but the HD 4850 1GB gives some headaches to that card.
5 - Power consumption is considerably lower, it has a smaller PCB and cool temperatures.
6 - It can be SLIed with other 9800GTX/9800GTX+ with the same framebuffer, the price is about right, but since ATi lowered the HD 4870 512MB to $149.99 and some other SKU's, there's no G92 card that can outperform the HD 4870 card.
 

Cookie Monster

Diamond Member
May 7, 2005
5,161
32
86
Originally posted by: evolucion8
I find very odd the benchmarks results in the AT review, I checked out reviews from other 6 sites and they all pretty much state the same thing:

1 - In VRAM limited situations, the GTS 250 1GB outperforms the 9800GTX+
2 - In non VRAM limited situations it's only as fast as a 9800GTX+, and sometimes slighly slower because of the VRAM latency or some strange wizardly
3 - When higher levels of Anti Aliasing are used with high resolutions, the HD 4850 shines except when is VRAM limited, those scenarios in the GTS 250 will not be limited by the amount of VRAM, but instead in raw power.
4 - Is competitive with the HD 4850, and it's slighly faster overall when compared to the 9800GTX+, but the HD 4850 1GB gives some headaches to that card.
5 - Power consumption is considerably lower, it has a smaller PCB and cool temperatures.
6 - It can be SLIed with other 9800GTX/9800GTX+ with the same framebuffer, the price is about right, but since ATi lowered the HD 4870 512MB to $149.99 and some other SKU's, there's no G92 card that can outperform the HD 4870 card.

+1 agreed. This is exactly what ive noticed.
 

imported_Scoop

Senior member
Dec 10, 2007
773
0
0
The 1GB BFG GTS 250 is an factory OC'd card in case you couldn't figure out from the pictures. Derek didn't mention this in the review so I don't know if he lowered the clocks or not to match stock. I checked out few other reviews and two of them, Tweaktown and Guru3D, had a different design in their test, both having a card which had two 6-pin power connectors (different cards though). So there's the difference between designs. Overclockers Club had the same BFG card in their tests and had similar results to those of AT in regards to how it compares to GTX 260 core 216. There were a few games @ OC Club where the GTX 260-216 gave significantly better performance and these games weren't tested at AT.

If people are waiting for a mainstream GTX 260/280, this is the card. I don't see nvidia putting out a card that would perform between GTS 250 and GTS 260. There's no room for that. I think the GTS 250 1GB looks like an awesome card for mainstream systems. Lots of performance for low power. Looks to beat the HD 4850 in most games so the only thing is the price and that depends on where you live. Where I live, Nvidia overprices their cards like no other, you can get the HD 4870 for less than the GTS 250 here. But I could definitely see myself spending 20 bucks more to get this particular BFG model instead of the 4850.
 

chizow

Diamond Member
Jun 26, 2001
9,537
2
0
Originally posted by: SickBeast
If Derek tested with Vsync enabled, you would see a framerate cap at whatever refresh rate he used. Seeing as there were scores in excess of 116FPS at 1680x1050 for NV cards in that review, I'd say that you either:

- don't know what Vsync does (!)
- don't know how Derek writes his reviews

Please don't spin things to make it look as though the NV cards are being held back unless you know what you're talking about.
I'm well aware of what Vsync does thanks, if you bothered to read, or better yet understand what I wrote you'd see what I'm talking about is clearly evident from the benchmarks and confirmed by Derek in the review comments on page 4 and 5 of his Multi-GPU review.

I never claimed Vsync was on in his testing, what I said was that unless you specifically forced it off, the driver itself could be smoothing frame rates. And before you go spouting off again, the driver does already do this, it has to in order to predict how many frames ahead it should pre-render and actually allows you to adjust this setting to a degree with a setting in the driver.

It should also come as no surprise that some of the recent gains in performance actually focus directly on extracting gains through improving multi-threaded driver improvements, allowing the driver and CPU to generate frames sooner as the GPU is able to render, reducing delay and increasing frame rates.

The reason this is relevant is because the OP specifically wondered why the results for the GTX 260 were off from what they've seen in other reviews. Again, this goes back to Derek not forcing Vsync OFF in the drivers, which other review sites do. At the same time, its clearly obvious Vsync in the drivers can result in very different results than Vsync in-game, which leads me to believe there's an extra layer of frame rate manipulation via the drivers that sacrifices performance (potentially lower FPS) for gameplay (less input lag).

And here's the back story of why Derek doesn't force Vsync off in the driver anymore:

Derek's Vsync Poll and Crysis investigation

And if you want to see the difference between Vsync in the drivers and Vsync in-game you can play with that as well. I personally find forcing Vsync in the driver results in much less input lag than in-game Vsync, which leads me to believe the driver is smoothing frame rates first and not just sampling 60 frames from max pre-rendered frames in-order.