Nvidia fan thinking of turning

humbug · 4 Sep 2012 at 15:04

andybird123 said:
so you bought it early when the prices were still high? even worse then

Compared to Nehalem?

LtMatt · 4 Sep 2012 at 15:05

andybird123 said:
I find it hard to take purchasing decisions too seriously from someone who actually paid money for a phenom x6...

Laugh though you may, he gets a higher minimum fps in battlefield 3 multiplayer than people with an i5 sandybridge cpu.

Rusty0611 · 4 Sep 2012 at 15:08

Medium settings/no AA

Also I moved from i5 2500k to i7 3770k and noticed no such bump in minimum FPS.

Kaapstad · 4 Sep 2012 at 15:11

LtMatt said:
Laugh though you may, he gets a higher minimum fps in battlefield 3 multiplayer than people with an i5 sandybridge cpu.

I like the graph my old i7 975 is doing well on it.

LtMatt · 4 Sep 2012 at 15:12

Rusty0611 said:
Medium settings/no AA

Also I moved from i5 2500k to i7 3770k and noticed no such bump in minimum FPS.

I think its medium graphics because they were trying to highlight the cpu performance.

Bf3 48-64 man multiplayer loves cores/threads, so the four core chips get much lower minimums.

humbug · 4 Sep 2012 at 15:16

I don't just use it for gaming, i use a lot of multi threaded number crunching APP's and the Hexa Core AMD is perfect for that, especially in 64Bit Linux.

In Windows gaming i play mostly FPS like BF3 where its no worse than a 2500K, so i'm better off staying where i am for now.

TheRealDeal · 4 Sep 2012 at 15:40

Rusty0611 said:
I don't have the time to police it - I have too much VRAM scaring to pour water on .

Indeed.

Psychodil: this is why the review is half decent it's no more than that. I don't see it any better than any of the other reviews. It's just more recent and the only one.

With a proper memory overclock I think the results would fall in line with what previous observations have shown.

I don't think you can take anything from the review really with that level of overclock. Like Matt said it might not have gone that high that still doesn't detract from the point.

What the review shows is at those settings they're level. It isn't really intellectually acceptable as an argument to generalise from that - with the memory overclock issue there - that the cards are truly level.

Here is the actual review from hardocp of the gtx670 used in the other review. It seems the memory on this card would not go any higher.

http://www.hardocp.com/article/2012/06/26/asus_geforce_gtx_670_directcu_ii_top_video_card_review/3

What i find interesting is a more recent review on a gigabyte gtx670oc card from hardocp. The memory oc he managed was lower than the direct cu.

Then theres this review of another gtx670 where the memory oc was not great.

http://www.hardocp.com/article/2012/06/04/galaxy_geforce_gtx_670_gc_video_card_review/3

and the stock original gtx670

http://www.hardocp.com/article/2012/05/10/nvidia_geforce_gtx_670_video_card_review/7

3 different gtx670's and no massive memory overclock.

Whatever way you look at it the review showed a nice oc'ed gtx670 v a nice oc'ed 7950. We have already went into this though so the point i am making here is the memory overclock he achieved was not to bad when comparing 2 other cards from the same site. He has showed with other cards he knows how to push the memory so i doubt hes holding it back.

Anyhow it shows getting huge memory clocks on a gtx670 is not a given.

Rusty0611 · 4 Sep 2012 at 15:50

I know - we've already been through that.

All I'm saying is that there are numerous 670's in the 670 official thread which are benching closer to 7 Ghz and above. And considering the 660Ti is clocked far higher it does suggest he got a poor clocker with regards to the memory.

It doesn't invalidate the results, it just means that this can't be taken as a "670 = 7950" review with the question marks surrounding the level of overclock.

In the post you quoted I haven't disagreed with anything anything you've gone on to say. With respect, it doesn't matter what overclocks he achieved on the memory in comparison to other overclocks he has achieved. What matters is what users are actually seeing/getting.

I'm obviously not saying every 670 is going to push 7.5 Ghz but what I'm saying is that the memory overclock is poor.

psychodil · 4 Sep 2012 at 15:50

^^ Good info, thankyou.

It should also be said that while they seemed to have a less than average overclock on the memory of the 670 in that review, the core clock they achieved was probably above average.

At least it should hopefully put to bed the thoughts that BF3 performs much better on nvidia cards.

Rusty0611 · 4 Sep 2012 at 15:58

psychodil said:
^^ Good info, thankyou.

It should also be said that while they seemed to have a less than average overclock on the memory of the 670 in that review, the core clock they achieved was probably above average.

At least it should hopefully put to bed the thoughts that BF3 performs much better on nvidia cards.

I know you haven't said it outright but they don't offset each other. My point is that because of the low memory overclock we're now effectively guessing what it would be like with a better memory overclock as we don't have the data to be sure. Which is why I post everytime somebody uses this review without the appropriate caveats

.

Indeed. I think overall 12.8 has now all but closed the gap.

It's a shame it took them until now to really get the performance out the 7*** series that it offers but nonetheless still a good thing.

humbug · 4 Sep 2012 at 16:12

Rusty0611 said:
I know you haven't said it outright but they don't offset each other. My point is that because of the low memory overclock we're now effectively guessing what it would be like with a better memory overclock as we don't have the data to be sure. Which is why I post everytime somebody uses this review without the appropriate caveats .

Indeed. I think overall 12.8 has now all but closed the gap.

It's a shame it took them until now to really get the performance out the 7*** series that it offers but nonetheless still a good thing.

So you will agree the GTX 670's and 7950's are in the same performance bracket as eachother, this includes BF3.

Rusty0611 · 4 Sep 2012 at 16:16

humbug said:
So you will agree the GTX 670's and 7950's are in the same performance bracket as eachother, this includes BF3.

I think that's blindingly obvious? If you had read my posts rather than ignoring the stuff about, you know, graphics cards you would have seen I had been saying that all along

.

To a larger extent, the 7950, 7970, 670 and 680 are all in the same performance bracket while at a high level of overclock.

At a guess I would say slowest (7950) to fastest (7970/680) is no more than about 10-15% depending on the game.

humbug · 4 Sep 2012 at 16:18

Rusty0611 said:
I think that's blindingly obvious? If you had read my posts rather than ignoring the stuff about, you know, graphics cards you would have seen I had been saying that all along .

To a larger extent, the 7950, 7970, 670 and 680 are all in the same performance bracket while at a high level of overclock.

At a guess I would say slowest (7950) to fastest (7970/680) is no more than about 10-15% depending on the game.

Just checking, chill Winston

tommybhoy · 4 Sep 2012 at 16:20

What about the 7950's that hit 1.25+GHz?....

...overclocking lottery is what it falls down to on the day, and that's the clocks that were used.

Rusty0611 · 4 Sep 2012 at 16:24

I don't think anyone is disputing that Tommy.

What is being questioned is the use of this as evidence of '7950 = 670' when the memory overclock achieved was poor in comparison to what other users are getting whereas 1.2 Ghz on the 7950 can't be considered poor. I'm not saying it can't go higher, I'm just saying there is well over 1 Ghz difference between the 660Ti memory and the 670 memory.

At those clocks shown the cards are about as equal as you can get. But the problem of using taking one review as gospel is that you can't balance out the overclocking lottery somewhat. One duff memory clocker effectively throws the results out.

psychodil · 4 Sep 2012 at 16:26

Perhaps they used low memory overclocks on the 670 so that the GALAXY 660 would look more competitive in comparison. The side effect being that the 7950 was shown more favourably against the 670. I dont think they cared as the review was clearly just to plug Galaxy's name as much as possible

Rusty0611 · 4 Sep 2012 at 16:28

psychodil said:
Perhaps they used low memory overclocks on the 670 so that the GALAXY 660 would look more competitive in comparison. The side effect being that the 7950 was shown more favourably against the 670. I dont think they cared as the review was clearly just to plug Galaxy's name as much as possible

Good point. Didn't think of that you little conspiracy theorist you

humbug · 4 Sep 2012 at 16:29

Its a conspiracy..... can i take pleasure in that to? please?

psychodil · 4 Sep 2012 at 16:29

Rusty0611 said:
I don't think anyone is disputing that Tommy.

What is being questioned is the use of this as evidence of '7950 = 670' when the memory overclock achieved was poor in comparison to what other users are getting whereas 1.2 Ghz on the 7950 can't be considered poor. I'm not saying it can't go higher, I'm just saying there is well over 1 Ghz difference between the 660Ti memory and the 670 memory.

At those clocks shown the cards are about as equal as you can get. But the problem of using taking one review as gospel is that you can't balance out the overclocking lottery somewhat. One duff memory clocker effectively throws the results out.

Well overall that review showed the 7950 having a very slight average lead over the 670 at the clocks tested. With an appropriate overclock on the memory on the 670, we can assume the performance gap is either reduced, or it will slightly overtake the 7950. Either way they will be within a few % of each other on average.

Rusty0611 · 4 Sep 2012 at 16:37

humbug said:
Its a conspiracy..... can i take pleasure in that to? please?

It's a valid point though and potentially a consideration. The review is to showcase the Galaxy 660Ti (which let's be honest - it does pretty well) and the other cards are in a base of comparison nothing more.

psychodil said:
Well overall that review showed the 7950 having a very slight average lead over the 670 at the clocks tested. With an appropriate overclock on the memory on the 670, we can assume the performance gap is either reduced, or it will slightly overtake the 7950. Either way they will be within a few % of each other on average.

Agreed. I think sitting somewhere between a 7950 OC and a 7970 OC on pure performance terms. However, as I said somewhere else it's all a little moot because for somebody with a blank slate, in my view there is only one choice and that ain't the 670!