- Status
alrobichaud
Distinguished
- Nov 9, 2011
- 796
- 0
- 19,060
What a great card for the money. My only complaints are:
1. Make 5760 x 1080 a standard in these reviews.
2. Overclock the 7970 to match the GTX 680!
My last 3dmark11 run yesterday came back at X3372 running at 1125MHz and 1575MHz on stock volts with a single 7970. AMD really crapped the bed setting default to 925MHz and 1375MHz.
1. Make 5760 x 1080 a standard in these reviews.
2. Overclock the 7970 to match the GTX 680!
My last 3dmark11 run yesterday came back at X3372 running at 1125MHz and 1575MHz on stock volts with a single 7970. AMD really crapped the bed setting default to 925MHz and 1375MHz.
jaquith
Glorious
- Feb 8, 2009
- 14,235
- 27
- 49,290
Another excellent review Chris & Thanks!!!
GTX 670 clock comparisons:
Ref Core (915 MHz), GPU Boost (980 MHz), and Memory (6008 MT/s) ; reference slower than GTX 680
OC Core (1061 MHz), GPU Boost (1140 MHz), and Memory (6460 MT/s) ; beats GTX 680
Clearly somewhere in the middle+ to closer OC listed above matches the GTX 680 performance.
No doubt that PCB is truly bizarre and to me unexpected, and I really don't know what to make of it. Those who will water block it 'should' get a break with less metal, but it's going to IMO look weird.
The (one DVI-I and one DVI-D) is a tad interesting, but fortunately they appear to both be Dual Link.
The stance with nVidia on PCIe 3.0 on the SB-E is driving me batty. First the GTX 680 'can' run PCIe 3.0 then a driver update later it cannot. Recently, your review with the GTX 690 it 'does' run PCIe 3.0 -- next week?! Now the GTX 670 PCIe 2.0 out of the box. Hmm...I wonder if the same GTX 680 PCie 3.0 'tweak' works to get PCIe 3.0 on the GTX 670's?! BTW - I know in 3-WAY it won't matter (PCIe 3.0 vs PCIe 2.0) -- that's not my point! However, PCIe 3.0 vs PCIe 2.0 in 4-WAY GTX 680 does matter in the resolutions (we) use 5900x1080 - proven fact.
All in all for most folks as of now the GTX 670 is an awesome deal -- we'll need to see the Prices & Performance of the GTX 660's to know what to buy. Since the majority of folks are using HD (1920x1080) monitors it seems the GTX 670's have found their home.
Ditto on the GTX 570 so no surprises there.
Hmm...I wonder if there's a 'way' to enable via a BIOS ... let's call it an 'update'?
GTX 670 clock comparisons:
Ref Core (915 MHz), GPU Boost (980 MHz), and Memory (6008 MT/s) ; reference slower than GTX 680
OC Core (1061 MHz), GPU Boost (1140 MHz), and Memory (6460 MT/s) ; beats GTX 680
Clearly somewhere in the middle+ to closer OC listed above matches the GTX 680 performance.
No doubt that PCB is truly bizarre and to me unexpected, and I really don't know what to make of it. Those who will water block it 'should' get a break with less metal, but it's going to IMO look weird.
The (one DVI-I and one DVI-D) is a tad interesting, but fortunately they appear to both be Dual Link.
The stance with nVidia on PCIe 3.0 on the SB-E is driving me batty. First the GTX 680 'can' run PCIe 3.0 then a driver update later it cannot. Recently, your review with the GTX 690 it 'does' run PCIe 3.0 -- next week?! Now the GTX 670 PCIe 2.0 out of the box. Hmm...I wonder if the same GTX 680 PCie 3.0 'tweak' works to get PCIe 3.0 on the GTX 670's?! BTW - I know in 3-WAY it won't matter (PCIe 3.0 vs PCIe 2.0) -- that's not my point! However, PCIe 3.0 vs PCIe 2.0 in 4-WAY GTX 680 does matter in the resolutions (we) use 5900x1080 - proven fact.
All in all for most folks as of now the GTX 670 is an awesome deal -- we'll need to see the Prices & Performance of the GTX 660's to know what to buy. Since the majority of folks are using HD (1920x1080) monitors it seems the GTX 670's have found their home.
- Jul 4, 2008
- 1,878
- 9
- 19,795
[citation][nom]alrobichaud[/nom]What a great card for the money. My only complaints are:1. Make 5760 x 1080 a standard in these reviews. 2. Overclock the 7970 to match the GTX 680!My last 3dmark11 run yesterday came back at X3372 running at 1125MHz and 1575MHz on stock volts with a single 7970. AMD really crapped the bed setting default to 925MHz and 1375MHz.[/citation]
I specifically wanted to avoid comparing an overclocked GTX 670 to a non-overclocked 7970, so I didn't. If you want some numbers putting the overclocked GTX 680 against an overclocked 7970, check out this piece for more! http://www.tomshardware.com/reviews/geforce-gtx-680-sli-overclock-surround,3162.html
I specifically wanted to avoid comparing an overclocked GTX 670 to a non-overclocked 7970, so I didn't. If you want some numbers putting the overclocked GTX 680 against an overclocked 7970, check out this piece for more! http://www.tomshardware.com/reviews/geforce-gtx-680-sli-overclock-surround,3162.html
- Jul 4, 2008
- 1,878
- 9
- 19,795
[citation][nom]fomoz[/nom]The GTX 680 does? Can you please do a 3-way review at that resolution then? And a 4-way with the GTX 680 4GB[/citation]
That's a lot of hardware I currently don't have in the lab, but I'll see what I can do ;-)
That's a lot of hardware I currently don't have in the lab, but I'll see what I can do ;-)
baracubra
Distinguished
- Jan 24, 2008
- 312
- 0
- 18,790
I've just noticed, is there a particular reason that there was no mention of tesselation performance? I did a quick google image search and found a picture of a tomshardware gtx 670 tesselation graph, but upon clicking it the link takes me back to the start of the article.... What gives?
On another note: if the 670 is so close in performance to the 680, is it at all realistic to be expecting a 670 Ti version of the chip? I imagine it would have at least a >5% advantage over the non-Ti version, but then it would beat out the 680...?
On another note: if the 670 is so close in performance to the 680, is it at all realistic to be expecting a 670 Ti version of the chip? I imagine it would have at least a >5% advantage over the non-Ti version, but then it would beat out the 680...?
It's the future. Nobody has reviews like that, but this many megapixels is going to be more and more common soon due to 4K displays. It's your chance to be a pioneer and to have these reviews available for comparison since 3x 1440p is 11.06 MP and 2160p is 8.29 MP, the difference is only 33%! That's a lot less than the 78% difference between 1440p and 1080p.
This basically means that if it runs well at 7680x1440, it will run it will definitely run well on QFHD.
- Jul 4, 2008
- 1,878
- 9
- 19,795
[citation][nom]Baracubra[/nom]I've just noticed, is there a particular reason that there was no mention of tesselation performance? I did a quick google image search and found a picture of a tomshardware gtx 670 tesselation graph, but upon clicking it the link takes me back to the start of the article.... What gives? On another note: if the 670 is so close in performance to the 680, is it at all realistic to be expecting a 670 Ti version of the chip? I imagine it would have at least a >5% advantage over the non-Ti version, but then it would beat out the 680...?[/citation]
Does this link work for you? http://media.bestofmicro.com/U/P/336769/original/tessellation%20scaling.png
Does this link work for you? http://media.bestofmicro.com/U/P/336769/original/tessellation%20scaling.png
davemaster84
Distinguished
- Jun 15, 2011
- 464
- 0
- 18,810
Well I think AMD isn't in such a big trouble as many people think. The 7970 was only a little bit behind the 680 only because the 680 has higher stock clock speeds, but as well as the 670 can do the 7970 can be overclocked actually better than any kepler card, so the 7970 is still better than the 670 specially in multi monitor set ups, same goes for the 680. Anyway I'm pissed of with nvidia, how could they release the 680 and 1 month after a card that is almost the same? Come on 7% between the two cards is almost nothing and the price difference is up to 25%. I don't care about the money spend since I have a good job now but it feels more like a treason from nvidia to us it's loyal customers. At least knowing that they are cutting their own product sales (meaning an obvious either small or large earning loss) makes me feel a bit better. Anyway I wont give my 680 away , it is an awesome product and I know the 680 sli is going to work beautifuly at least from a year from now, but the 670 should have been this good sorry.
davemaster84
Distinguished
- Jun 15, 2011
- 464
- 0
- 18,810
dosdecarnitas
Distinguished
- May 23, 2011
- 161
- 0
- 18,690
bigger supply? maybe its happening like AMD did some time ago with phenoms X3 that faulty cores are rebranded same here maybe? that may be the whole reason nvidia may have a bug building the chip and as a result and solution VOILA GTX670
the only difference though, here Nvidia has the better architecture Where AMD Phenoms didnt
the only difference though, here Nvidia has the better architecture Where AMD Phenoms didnt
baracubra
Distinguished
- Jan 24, 2008
- 312
- 0
- 18,790
[citation][nom]cangelini[/nom]Does this link work for you? http://media.bestofmicro.com/U/P/3 [...] caling.png[/citation]
Yupp, thanks that worked. But I've looked through the entire review 3 times now and I still can't find it. Am i going mad or is it really not in there...?
Yupp, thanks that worked. But I've looked through the entire review 3 times now and I still can't find it. Am i going mad or is it really not in there...?
ElMoIsEviL
Distinguished
- Jan 10, 2006
- 2,361
- 1
- 19,965
Wow...
Say Chris Angelini... do you work for nVIDIA by chance?
Your game selection is highly suspicious (as per usual) and appears to be this way whenever you review an nVIDIA card. WoW? Who cares about WoW performance it can run on 5 yr old hardware with GREAT ease.
But what really gives you up is on this page: http://www.tomshardware.com/reviews/geforce-gtx-670-review,3200-15.html
I mean you really can't deny being at least partially a shill. You compare the GTX 670 and 680 at 2560x1600. Fair enough. But you omit the same results from the GTX 670 vs 7970 at that same resolution. You go on to write this...
"Based on the complete benchmark analysis, we know that Nvidia’s design does best at 1680x1050, while AMD competes most aggressively at 2560x1600. Using 1920x1080 again gives us a good gauge of playable performance at the native resolution of many popular panels."
Really? then why did you post the 2560x1600 figures right above for GTX 670 vs. GTX 680 yet omit posting the same graph for GTX 670 vs. 7970? Could it be because the 7970 pulls ahead? You claim that most panels only support 1920*1080 and that's quite true but these are high end GPUs and you DID post the results right above at 2560x1600.
You even admit that your game selection is skewed: "Of course, you could add games to this list all day long, and the balance would undoubtedly shift. In our suite, though, the GTX 670 averages more than 13% faster."
WoW shouldn't be there. It may be a popular game but... it would be like posting CounterStrike Source numbers just because that is still a popular game. Or Quake 3 while you're at it.
Also WoW doesn't use any of the new features. So all you're gauging are mostly Raster Ops (if you even know what that is). That's why nVIDIA pulls ahead (hell older nVIDIA cards surpass AMD cards due to this very fact). It ain't REAL DX11. It just runs on the API doesn't use any of the features.
Amateur.
Say Chris Angelini... do you work for nVIDIA by chance?
Your game selection is highly suspicious (as per usual) and appears to be this way whenever you review an nVIDIA card. WoW? Who cares about WoW performance it can run on 5 yr old hardware with GREAT ease.
But what really gives you up is on this page: http://www.tomshardware.com/reviews/geforce-gtx-670-review,3200-15.html
I mean you really can't deny being at least partially a shill. You compare the GTX 670 and 680 at 2560x1600. Fair enough. But you omit the same results from the GTX 670 vs 7970 at that same resolution. You go on to write this...
"Based on the complete benchmark analysis, we know that Nvidia’s design does best at 1680x1050, while AMD competes most aggressively at 2560x1600. Using 1920x1080 again gives us a good gauge of playable performance at the native resolution of many popular panels."
Really? then why did you post the 2560x1600 figures right above for GTX 670 vs. GTX 680 yet omit posting the same graph for GTX 670 vs. 7970? Could it be because the 7970 pulls ahead? You claim that most panels only support 1920*1080 and that's quite true but these are high end GPUs and you DID post the results right above at 2560x1600.
You even admit that your game selection is skewed: "Of course, you could add games to this list all day long, and the balance would undoubtedly shift. In our suite, though, the GTX 670 averages more than 13% faster."
WoW shouldn't be there. It may be a popular game but... it would be like posting CounterStrike Source numbers just because that is still a popular game. Or Quake 3 while you're at it.
Also WoW doesn't use any of the new features. So all you're gauging are mostly Raster Ops (if you even know what that is). That's why nVIDIA pulls ahead (hell older nVIDIA cards surpass AMD cards due to this very fact). It ain't REAL DX11. It just runs on the API doesn't use any of the features.
Amateur.
- Jul 4, 2008
- 1,878
- 9
- 19,795
[citation][nom]Baracubra[/nom]Yupp, thanks that worked. But I've looked through the entire review 3 times now and I still can't find it. Am i going mad or is it really not in there...?[/citation]
I linked that image on the first page rather than breaking out an entire section on tessellation performance. I have the raw before/after graphics in the album as well.
If that's something you want to continue seeing moving forward, I'll start putting them back in. Nobody had commented on tessellation performance for months, though, so I was trying to avoid cluttering things up
I linked that image on the first page rather than breaking out an entire section on tessellation performance. I have the raw before/after graphics in the album as well.
If that's something you want to continue seeing moving forward, I'll start putting them back in. Nobody had commented on tessellation performance for months, though, so I was trying to avoid cluttering things up
ElMoIsEviL
Distinguished
- Jan 10, 2006
- 2,361
- 1
- 19,965
Also you're testing a Superclocked GTX 670 vs. Stock Clocked AMD cards.
Perhaps you should test a factory OCed 7970 or 7950 vs your super clocked GTX 670??
This is why your results differ from many other websites including Techpowerup and even Anandtech. Tomshardware is starting to feel like Popular culture rather than g33k/nerd culture. Like a Justin Bieber concert.
Perhaps you should test a factory OCed 7970 or 7950 vs your super clocked GTX 670??
This is why your results differ from many other websites including Techpowerup and even Anandtech. Tomshardware is starting to feel like Popular culture rather than g33k/nerd culture. Like a Justin Bieber concert.
buzznut
Splendid
- Sep 26, 2007
- 3,650
- 2
- 23,465
Yeah, I can't afford $400 for a video card either. But this one is quite sexy, powerful and efficient.
I think you've nailed it though with the paper launches. This would be more compelling if we thought nVidia could pull it off. If stock is too low prices will be inflated. For me I hope AMD does lower its prices on the 7970/7950, that will benefit me but it won't do much for AMD.
I think you've nailed it though with the paper launches. This would be more compelling if we thought nVidia could pull it off. If stock is too low prices will be inflated. For me I hope AMD does lower its prices on the 7970/7950, that will benefit me but it won't do much for AMD.
[citation][nom]cangelini[/nom]I specifically wanted to avoid comparing an overclocked GTX 670 to a non-overclocked 7970, so I didn't. If you want some numbers putting the overclocked GTX 680 against an overclocked 7970, check out this piece for more! http://www.tomshardware.com/review [...] ,3162.html[/citation]
How would new drivers and the fact that Catalyst isn't limiting the 7970s anymore effect those numbers? I assume that you guys are getting pretty busy at this point, but this is something that we don't know. So much of the data is outdated already.
How would new drivers and the fact that Catalyst isn't limiting the 7970s anymore effect those numbers? I assume that you guys are getting pretty busy at this point, but this is something that we don't know. So much of the data is outdated already.
funny. some of the recent 680 buyers who bought those after waiting for a long time might feel like suckers.
seems like kepler can be scaled back to outperform (gaming) 7800 series with ease while retaining power efficiency.
670's gaming performance is impressive.
seems like kepler can be scaled back to outperform (gaming) 7800 series with ease while retaining power efficiency.
670's gaming performance is impressive.
- Jul 4, 2008
- 1,878
- 9
- 19,795
[citation][nom]ElMoIsEviL[/nom]Also you're testing a Superclocked GTX 670 vs. Stock Clocked AMD cards.Perhaps you should test a factory OCed 7970 or 7950 vs your super clocked GTX 670??This is why your results differ from many other websites including Techpowerup and even Anandtech. Tomshardware is starting to feel like Popular culture rather than g33k/nerd culture. Like a Justin Bieber concert.[/citation]
Nowhere is the EVGA card tested against AMD's reference board.
I've added the following to page two for the conspiracy theorists:
All GeForce GTX 670 results in the following pages were generated using a card supplied by Nvidia operating at stock clock rates (915/1502 MHz). EVGA's GeForce GTX 670 Superclocked was only used for the benchmarks on this page. Our SLI results were generated by down-clocking the EVGA board to match Nvidia's.
Nowhere is the EVGA card tested against AMD's reference board.
I've added the following to page two for the conspiracy theorists:
All GeForce GTX 670 results in the following pages were generated using a card supplied by Nvidia operating at stock clock rates (915/1502 MHz). EVGA's GeForce GTX 670 Superclocked was only used for the benchmarks on this page. Our SLI results were generated by down-clocking the EVGA board to match Nvidia's.
- Jul 4, 2008
- 1,878
- 9
- 19,795
[citation][nom]joeman99[/nom]How would new drivers and the fact that Catalyst isn't limiting the 7970s anymore effect those numbers? I assume that you guys are getting pretty busy at this point, but this is something that we don't know. So much of the data is outdated already.[/citation]
Happy to take a look at 7970's overclocking using the "uncorked" 12.4s. Soon, hopefully. Back-to-back-to-back NDAs make it hard to revisit topics like that as fast as I'd like.
Happy to take a look at 7970's overclocking using the "uncorked" 12.4s. Soon, hopefully. Back-to-back-to-back NDAs make it hard to revisit topics like that as fast as I'd like.
boodahz
Honorable
- May 7, 2012
- 43
- 0
- 10,540
What is Nvidia thinking? Instead of making new cards that they can't produce fast enough, why don't they make the old cards that they can't produce fast enough. Its all a marketing game. I guess it's good to be competitive, but not being able to produce your product is far from competitive. I recently bought a 7970 because i didn't feel like paying $600+ for a 680...
- Status
TRENDING THREADS
-
News TSMC founder says Intel should focus on AI, not advanced process technologies- Started by Admin
- Replies: 27
-
-
Question New Drive - Bizarre S.M.A.R.T results (pre-failure/old-age??) is it CMR or SMR?- Started by 93QSD5
- Replies: 6
-
News Intel CEO Pat Gelsinger retires, effective immediately — also steps down from BOD, two co-CEOs step in
- Started by Admin
- Replies: 124
Facebook
Twitter
Instagram