Hardware EVGA GeForce GTX 650 1GB review

its a GT 640 with DDR 5 - they sold them for ages as OEM

personally i think that with a screen under 21" eg 1680x1050 or lower it would do `ok`.

 

The irony here is that this is exactly how Bit used to review cards back in the 8800GTX (and previous) days.

TBH I would prefer it if they actually went back to this method. Tailor the review to meet the card, don't just throw it through some 60 second "benchmark" and comment on the outcome.

There is very little value to be garnered from throwing an entry level card through a 1080p "max settings" benchmark and commenting that it isn't fast enough. Don't tell me what I already know, tell me what the card can actually do.

EDIT: For what it's worth my old temporary HD7770 managed around 45FPS at 1680x1050 "high" settings (no MSAA) in BF3. Perfectly playable for the most part. The GT650 should be comparable.

 

I like how they put High-Air-Bracket on any of their cards. No matter GTX650 doesn't really needed. I like the card. For me it's not pointless. It depends a lot on the resolution used and the game. Since I only play SC2, the card is pretty much valid. I hope someone makes a low-profile of it like this: http://www.palit.biz/palit/vgapro.php?id=1393 . Because Sapphire low profile 7750 has the crown there for quite some, even at the 6xxx generation. And make it dual-slot. Enough of this single-slot LP-cards , I don't care how pointless it may look to slap a dual-slot cooler on a ~50W TDP card, we are enthusiasts after all have some fun here and there.

 

Were all the nvidia cards retested with the 306 driver?

Despite nvidia's claims 5,10,15% increases in some games, I noticed the 670 didn't show any improvement in BF3.

 

http://www.ldlc.com/informatique/pi...nterne/c4684/p1e48t3o0a1+fv121-7746,8109.html

GTX650 and HD7770 prices in ascending order ... not very good for the GTX650 since they have exactly the same price.

 

What I gathered from what people said when they asked for a "best settings" rating was that they wanted to know something like "How much do I need to spend to get a good experience at resolution x" or "What settings/resolution could I use if I spend y amount on a graphics card". This is useful information, the sort of thing I've been asked several times by various people.

However, despite the potential usefulness of this setting-based rating I still think that a min/average fps system (as bit-tech use atm) is the better solution. This is because having a settings-based system would introduce a lot more subjectivity into the review. If you could get a game "playable" (which is itself a subjective term) at, say, 1680x1050 with mostly low settings but textures at medium, is that preferrable to the same "playability" at 1280x1024 with all high settings? What if you could have either high textures or high shadows giving you a "playable" experience? Which one is "better" to have? You'd have the same problem across reviews; should I buy card x which can play this game at 1336x768 at high settings, or card y which can play that game at 1280x1024 on medium settings?

The point of that last paragraph was that introducing settings as a way to differentiate between graphics cards would in fact reduce clarity, as different people would have different priorities for their graphics settings. A straight-up min/average fps system makes it much easier to decide which card is best for a particular price point, as you can see that Card A performs better in the tests than Card B (although even with this system it can still depend on which games you play the most) and so is better value for money.

I guess what'd be nice is for a round-up of graphics cards at some point to show how much you'd need to pay for say maximum settings at a variety of resolutions at a given minimum frame rate (25 or 30fps is usually what I see regarded as playable, but even then would you rather see a card recommended that can provide a smooth performance (60fps) at that resolution or one that can just pass as playable (25fps) at that resolution?). You could then use this information to work either way (how much for these settings/what settings for this amount).

Oh yeah, and trying to find the best settings for a card would take much longer than their current testing, I guess that should be considered.

Now, time to get back to what I'm meant to be working on

 

Got to agree with Adnoctum here. Of course this thing wasn't going to be able to handle AAA titles with everything cranked up at 5760x1400. However, [H] would at least show me what sort of compromises I'd be looking at if, for some reason, I was trying to play AAA titles with this card.

Seems like maybe there should be a second suite of tests for low-end cards. If I'm looking to buy a $100 GPU, show me that what sort of performance I'll get in TF2, Portal 2, Dirt 3, etc. Portal 2 and Dirt 3 are still commonly used for benchmarks, even when it gets a little silly for the high-end cards (if it's in the neighborhood of 200 FPS, maybe it should get replaced in that benchmark suite.) But, these are likely the sort of games consumers are looking to play on a $100 card.

 

I have no issue with the kinds of pure benchmark number based reviews, such as this one. I think it is needed to generate data in order to empirically rank cards in a form that is easily viewed and easily analysed. It lets you look at a glance the basic difference between cards for a given benchmark.

But, if I am being completely honest, I think it is really only half a review, and only half of what is needed to make a fully informed purchase.

Take a look at this review (because we are here, not as a criticism), on page 3 the numbers say that the GTX650 averaged 25fps with a minimum of 20fps in Battlefield 3 at 1920x1080. Now my experience of FPS (but not B3) is that is marginally playable for entertainment, if not competitively (in a single-player FPS I could live with 20-25fps).
But is it? Those two figures can hold a multitude of sins that would make a game unplayable: repetitive and repeated stuttering, infrequent periods of sustained low (20fps) frame rates, predictable slow-downs in areas of high activity.

The subjectivity of "playable" is the easiest to overcome. You state up front what "playable" means. You say that for a type of game, "playable" is the ability for a game to be played with no perceived stuttering, slow-downs or hang-ups that interrupt gameplay. It may be that the requirements to meet the term "playable" is different for competitive FPS (such as B3) compared to an Action RPG (Skyrim) or an RTS (SC2).

The level of technology and capability between the two brands are largely comparable and easily graphed at the moment, but what if one maker implements something significantly better? I'm using this as an easily explained example, not reality: the AA used in this article was MSAA. What if the GTX650 could implement FXAA with no performance impact and the resulting image was no different than if it was using MSAA? And what if AMD didn't have their MLAA or it was inferior (performance or quality)?

Because neither FXAA nor MLAA is implemented during the benchmarks, only MSAA was, there is no opportunity to examine how implementing FXAA impact the GTX650 performance. Not only that, but there is no analysis of how much of an impact MSAA has on the card. There are no benchmarks for 0x, 2x, 4x, 8x. Would turning the MSAA down to 8x make B3 @ 1920x1080 a smooth experience with the GTX650? That is a reasonable question to ask of a review, and if I was reading an [H]OCP review I would know.

It still wouldn't be possible to show what a card does at all resolutions, at all possible settings to suit a single need. But that is not is what is being suggested. All that is suggested would provide a bit more depth of analysis that would give a bit of clarity to the capability of the card, so that we can infer where possible from the available data what card would satisfy our requirements.
If I wanted to know how the GTX650 would do playing B3 at 1920x1080/4xAA/16xAF/Ultra then this review is great! It even tells me how it does at 5760x1080 in case I ever want to play B3 on 3 monitors with this card. Unfortunately, it does nothing to inform me at what would happen at 0xAA or 1650x1050.

To do this does take time. You actually have to watch what is happening instead of running a benchmark and compiling the numbers spat out into a spreadsheet. Which is why [H]OCP doesn't compare cards to every other card out there. For the nVidia GTX670 launch, [H] compared it to the HD7950 and the GTX580 and tested it in 5 or 6 games. For the HD7970GHz launch they compared it to the vanilla HD7950 and the GTX680. You may have noticed that these are the competitors for the card in question, and not an irrelevant HD6770/GTX550 in sight.

They don't give us useless information about cards that are irrelevant. If you want to see how the GTX650 compares to the GTX690, then open up the GTX690 review and compare the benchmarks. What would have been useful would have been the inclusion of the GTX550/560. Does it beat the GTX550Ti which I can buy for $20 less, in B3 @ 1920x1080? From this review I would never know, although the answer is "No, it doesn't".

Not complaining, but I suspect that the data for all the other cards were lying around from previous articles, and throwing them in costs nothing and it makes the review look more complete. Otherwise there is no reason to be testing them again just for this article.

I hope that my post doesn't come across as combative, just that I've been enjoying the conversation about testing in this thread. Your post made some compelling points and a lot of sense.

Similarly, I don't really have much to be unhappy about the article and no complaints. I got from it what I need to make an informed purchase of a graphics card: I'll either get the HD7750 I was thinking about or I'll hold on until the end of the year when prices will be cheaper and there will be chatter about the AMD mid-range replacement.
I'm actually grateful to Bit-Tech for the review, because many sites haven't even bothered, which is a bit disturbing when you think about it. The GTX650 price/performance level should be the meat of lower-mainstream gaming.
It just seems that reviews seem to be a little more shallow than they used to be, which I will admit may be a result of rose tinted glasses, as reviews used to rip into settings to find out how AA impacted performance at different resolutions and examined how good the image quality was.