Overclock.net › Forums › Graphics Cards › Graphics Cards - General › 7950 or gtx 670?
New Posts  All Forums:Forum Nav:

7950 or gtx 670? - Page 10

post #91 of 237
Quote:
Originally Posted by jtom320 View Post

All I know is I d on't believe Nvidia is as far ahead as AMD in tesselation when AMD takes Metro and Crysis 2 and then proceeds to completely spank them in Heaven 3.0.
Also the 7970/680/670 comparison is more interesting as AMD's current scaling is better and the 7970 is easily the best overclocker of the bunch. I had my reference asus card pushing 1325 stable with a mid-sized voltage bump.
The problem if you ask me with current Nvidia offerings by the way is that they locked down voltage and because of that you will see poor overclocks unless you are lucky enough to get good silicon. See you used to be able to get bad silicon but push a ton of volts through it and get a decent overclock. Now with the 6** series it's entirely dependent on luck.

My 7970 can easily hit 1350 with 1.2 volt bump and 1900 memory with 1.65 volt. Maybe amd is not equal in tesellation but its better than the previous generations. A lot better
Workstation
(4 items)
 
  
CPUMotherboardGraphicsMonitor
Xeon E5-2690 Supermicro 2011 Nvidia GP100/ Vega FE Dell ultrasharp 4k 
  hide details  
Reply
Workstation
(4 items)
 
  
CPUMotherboardGraphicsMonitor
Xeon E5-2690 Supermicro 2011 Nvidia GP100/ Vega FE Dell ultrasharp 4k 
  hide details  
Reply
post #92 of 237
Quote:
Originally Posted by sugarhell View Post

My 7970 can easily hit 1350 with 1.2 volt bump and 1900 memory with 1.65 volt. Maybe amd is not equal in tesellation but its better than the previous generations. A lot better

Really? Easily 1350 with just 1.2 volts? Can we see some benchmarks with those clocks?
post #93 of 237
Quote:
Originally Posted by SeanPoe View Post

First of all, i specifically said it runs at full boost all the time if heat isn't an issue, so i didn't contradict anything. Learn the definition of the word. Secondly, the boost is dynamic, it doesn't ONLY go no-boost or full-boost, it increases by steps of ~13mhz. So if heat was an issue, it would still be boosting, but just not as much.
The reason this information is irrelevant is because a overclocked 7950 in the same scenario as an overclocked 670, will suffer from heat MORE because a 7950 produces more heat. So if the 670 EVER has to throttle itself down then the 7950's overlook will have to be manually reduced as well. The fact that a 670 produces less heat means even in a situation where the 7950 is too hot at a high-overclock, the 670 might still be able to run at full boost. So you still don't understand how it works with comment liks that.
And that graph is inaccurate and incomplete. I also like how you used it out of context, classy touch thumb.gif. He's listing the boost clock speed for half of those not the maximum clock speed with boost (there's a huge difference). You can ask the author yourself and he confirm that half of those entries are incomplete as he didn't have the maximum boosted clock data.

Yep yep, or maybe i know how drastically ambient temps and climax affect an overclock.. As experience... Day 2 Day here in chicago or Day to Day in Puerto Rico 24/7 = 365 tropical..

I can put the pc outside without the need of cooling on a 10 degree weather here in chicago and get good oced performance or even just running the pc with 65f degrees here.. cant say the same over the tropical side when i get at least 32.2c degrees all year around degrading my margin of an overclock.. We are talking at leas of 15c of difference my friend on just ambient temps.. That without including the rest of the hardware added heat to the pc itself which takes longer to cool off on a hotter climate..

Benches with no boost please.. XD

Btw how the hell 670 runs cooler?
http://images.anandtech.com/graphs/graph5818/46564.png
idk about that
Edited by zGunBLADEz - 5/14/12 at 1:25pm
post #94 of 237
Quote:
Originally Posted by DimmyK View Post

Really? Easily 1350 with just 1.2 volts? Can we see some benchmarks with those clocks?

with a crappy 965 the bottleneck is high as hell xd
Workstation
(4 items)
 
  
CPUMotherboardGraphicsMonitor
Xeon E5-2690 Supermicro 2011 Nvidia GP100/ Vega FE Dell ultrasharp 4k 
  hide details  
Reply
Workstation
(4 items)
 
  
CPUMotherboardGraphicsMonitor
Xeon E5-2690 Supermicro 2011 Nvidia GP100/ Vega FE Dell ultrasharp 4k 
  hide details  
Reply
post #95 of 237
Quote:
Originally Posted by zGunBLADEz View Post

Yep yep, or maybe i know how drastically ambient temps and climax affect an overclock.. As experience... Day 2 Day here in chicago or Day to Day in Puerto Rico 24/7 = 365 tropical..
I can put the pc outside without the need of cooling on a 10 degree weather here in chicago and get good oced performance or even just running the pc with 65f degrees here.. cant say the same over the tropical side when i get at least 32.2c degrees all year around degrading my margin of an overclock.. We are talking at leas of 15c of difference my friend on just ambient temps.. That without including the rest of the hardware added heat to the pc itself which takes longer to cool off on a hotter climate..
Benches with no boost please.. XD
Btw how the hell 670 runs cooler?
http://images.anandtech.com/graphs/graph5818/46564.png
idk about that

I realize all that, but you need to understand that high ambient temperatures hinder the 7950 more than the 670 simply because the 670 runs cooler. So, what i've tried to convey a couple times now, is that if the 670 struggles to keep itself at max boost then the 7950 will also struggle (and will need to be down-clocked too). That's why the 670 isn't being unfarily represented when it has boost turned on.

Yeah, that graph is showing CORE temperatures at low factory-OC. Core temperature has little to do with total heat expenditure though. That's just shows how efficient the core is at transferring its heat to the heatsink -- NOT how much total heat the heatsink has to dissipate into the atmosphere (ie, the case). The 7950 has to double its power consumption to go from ~900c to 1200c (369w at load), heat expenditure is closely tied to power consumption, the more power something consumes the more of that power is converted to heat. The 670 only has to increase it's power consumption by 4% to go from 915c to 1250c (265w at load) , that's why it produces so little heat compared to the 7950 and that's why the 670 will always be running at boost speeds.
post #96 of 237
You guys need to keep it more civil, otherwise I'm shutting down this thread. Which would be a shame cause it's generally a good discussion.

Now ... Gunz ... it's not possible to 'turn off the boost' on Kepler cards. So your continued insistence on benches without boost has become a bit nonsensical ... it is no't like Intel where you can turn it on and off. So these benches you're asking for ... that ain't gonna happen.

AFA the OC Scaling question goes, one must look at ACTUAL operating frequencies to do properly the calculations. On Kepler, it's basically done just like it always has been done: % FPS change/% Clock Change.

The proper scaling formula is:
((OC'd FPS - Stock FPS)/Stock FPS)
/
((Actual OC'd Operating Frequency after 'Kepler Boost' - Actual Stock Operating Frequency after 'Kepler Boost')/Actual Stock Operating Frequency after 'Kepler Boost')

This is the same formula as should always be used for all card, in the sense that one should always use 'actual' observed frequencies... the difference for Kepler is just that Actual Clocks is not what's seen/set in Afterburner or whatever, because that misses the Kepler Boost ... which only occurs at run-time.

However, there's a number of very wrong Kepler Scaling calcs being done out there, where reviewers wrongly use the base clock of 915 to calculate the lower end, then use the max OC'd operating frequency to calculate the delta from that. They then incorrectly associate the non-OC'd FPS with 915 clocks, which is where the issue comes in. WHen you 'do it wrong', (like they did at [H] for example), Kepler scaling looks terrible.

I suspect when all'z said and done and calculated properly, Kepler OC Scaling and Tahiti OC Scaling will be very close ... although the actual raw FPS value gained will be much higher for Tahiti, because their % Overclock (the actual clock delta) overall will be much higher.
Edited by brettjv - 5/14/12 at 2:53pm
    
CPUMotherboardGraphicsRAM
xeon X5675 6-core @ 4.1ghz (1.29v, 20x205 +ht ) rampage iii extreme msi rx470 gaming X (the $159 budget king) 3 x 2gb corsair xms3 pc12800 (9-9-9-24-1T@1600MHz) 
Hard DriveOptical DriveCoolingOS
hynix 250gb ssd (boot), 2tb deskstar (apps),1tb... plextor px-712sa - still the best optical drive... corsair h8o v2 aio W10 home 
MonitorPowerCaseAudio
asus vw266h 25.5" (1920x1200) abs sl (enermax revolution) * single 70A rail 850w silverstone rv-03 XFi Titanium 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
xeon X5675 6-core @ 4.1ghz (1.29v, 20x205 +ht ) rampage iii extreme msi rx470 gaming X (the $159 budget king) 3 x 2gb corsair xms3 pc12800 (9-9-9-24-1T@1600MHz) 
Hard DriveOptical DriveCoolingOS
hynix 250gb ssd (boot), 2tb deskstar (apps),1tb... plextor px-712sa - still the best optical drive... corsair h8o v2 aio W10 home 
MonitorPowerCaseAudio
asus vw266h 25.5" (1920x1200) abs sl (enermax revolution) * single 70A rail 850w silverstone rv-03 XFi Titanium 
  hide details  
Reply
post #97 of 237
Some of you guys are exaggerating how much the boost clock is throttled back. A lot. The clockspeed goes down by a whopping 13mhz when it reaches 70C, which it won't ever reach to begin with if you make your own fan curve. As long as the card is properly stressed it will run at its max boost clock.
Beestje 2.0
(20 items)
 
Beestje 1.5
(16 items)
 
 
CPUMotherboardGraphicsGraphics
Intel Core i7 3820 Sabertooth X79 ASUS GeForce GTX 680 ASUS GeForce GTX 680 
RAMHard DriveHard DriveHard Drive
Samsung low profile x4 OCZ Vertex 3 Crucial M4 Samsung HD103SJ 
Optical DriveCoolingCoolingOS
Lite-On BD drive Megahalems rev. B Cooler Master Excalibur Windows 7 home premium x64 
MonitorKeyboardPowerCase
ASUS VG278H Logitech G710+ Corsair TX750 Silverstone Raven 2 Evolution 
MouseMouse PadAudioAudio
Corsair M60 Razer Kabuto Echo Audiofire 2 Sennheiser HD515 
CPUMotherboardGraphicsRAM
i7 920 Asus P6T Club3d HD7870 royalKing Corsair Dominator 3x4GB 
Hard DriveHard DriveOptical DriveCooling
Crucial V4 128GB Seagate Barracuda LP 2TB LG GH22LS Scythe Mugen 2b 
CoolingCoolingMonitorKeyboard
CM Excalibur x2 CM R4 x3 Samsung BX2450 Microsoft Sidewinder X6 
PowerCaseMouseMouse Pad
Seasonic M12 II 520 Antec 600 Microsoft Intellimouse Razer Kabuto 
  hide details  
Reply
Beestje 2.0
(20 items)
 
Beestje 1.5
(16 items)
 
 
CPUMotherboardGraphicsGraphics
Intel Core i7 3820 Sabertooth X79 ASUS GeForce GTX 680 ASUS GeForce GTX 680 
RAMHard DriveHard DriveHard Drive
Samsung low profile x4 OCZ Vertex 3 Crucial M4 Samsung HD103SJ 
Optical DriveCoolingCoolingOS
Lite-On BD drive Megahalems rev. B Cooler Master Excalibur Windows 7 home premium x64 
MonitorKeyboardPowerCase
ASUS VG278H Logitech G710+ Corsair TX750 Silverstone Raven 2 Evolution 
MouseMouse PadAudioAudio
Corsair M60 Razer Kabuto Echo Audiofire 2 Sennheiser HD515 
CPUMotherboardGraphicsRAM
i7 920 Asus P6T Club3d HD7870 royalKing Corsair Dominator 3x4GB 
Hard DriveHard DriveOptical DriveCooling
Crucial V4 128GB Seagate Barracuda LP 2TB LG GH22LS Scythe Mugen 2b 
CoolingCoolingMonitorKeyboard
CM Excalibur x2 CM R4 x3 Samsung BX2450 Microsoft Sidewinder X6 
PowerCaseMouseMouse Pad
Seasonic M12 II 520 Antec 600 Microsoft Intellimouse Razer Kabuto 
  hide details  
Reply
post #98 of 237
Quote:
Originally Posted by brettjv View Post

The proper scaling formula is:
((OC'd FPS - Stock FPS)/Stock FPS)
/
((Actual OC'd Operating Frequency after 'Kepler Boost' - Actual Stock Operating Frequency after 'Kepler Boost')/Actual Stock Operating Frequency after 'Kepler Boost')
This is the exact same formula as should always be used ... the difference for Kepler is just that Actual Clocks is not what's seen/set in Afterburner or whatever, because that misses the Kepler Boost ...
I suspect when all'z said and done and calculated properly, Kepler OC Scaling and Tahiti OC Scaling will be very close ... although the actual raw FPS value gained will be much higher for Tahiti, because their % Overclock overall will be much higher.

This is definitely correct, i even made the foolish mistake of doing it the other way here when i calculated it (calculating default core speed vs oc core speed (without boost, assuming this to be a nearly negligible constant) as the percent of overclock, rather than boosted core vs OC boosted core).

We don't have enough data to calculate it properly because no one is listing their stock boosted core speeds. So all we know is the OC boosted core speeds. But from what i've seen, the boost varies between +140-180 which means you can calculate the scaling (like i did in the link above in this post) by just assuming the boost is +~150 for both the default and OC core values. It's not 100% accurate but it should be accurate within 3-6% is my guess given the range on the boost.
post #99 of 237
Thread Starter 
thanks for all your replies guys, really helpful!
post #100 of 237
Thread Starter 
@ SeanPoe it doesn't matter if some benches are non ref 7950's compared to referance 670's, because that's what's for sale in my country, and thats what my choice consists of
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Graphics Cards - General
Overclock.net › Forums › Graphics Cards › Graphics Cards - General › 7950 or gtx 670?