AnandThenMan
Diamond Member
- Nov 11, 2004
- 3,991
- 627
- 126
I don't recall anyone saying that?
According to you:
Pros: Nvidia
Cons: AMD
I kid I kid, people are too wound up here lately.
You make it sound like you could prevent that.
Pros: Nvidia
Cons: AMD
I kid I kid, people are too wound up here lately.
- Sep 5, 2003
- 19,458
- 765
- 126
I linked you 3 reviews that measured the entire system power consumption and in all 3 the system with the 7950 used less power than a system with a 670. So my point is, you have not provided a conclusive enough sample size to state that "most review sites are showing a ~30 watt discrepancy to TPU".
The reality is the 7950 consumes less power on average than a GTX670 does. The reality is most Kepler cards have similar voltage but most reviews are using reference 7950 cards. You can't seriously compare the power consumption of a reference 7950/7970 card to an after-market one. That's the WHOLE point I keep trying to explain to you guys. You linking that 670 used 36W less than a 7950 is irrelevant since the review uses a 7950B card. Who cares!
Also, you said when you pair a GPU within a system, all the components work together. That is true but then if you have to discuss that NV cards tend to be more more CPU limited.
Now you are bringing in way too many variables. This is why when we look at average and peak #s for a GPU in a 3770K @ 4.7ghz system at TPU, it gives us the GPU power consumption on a very fast system. That's about as unbiased as it gets.
TPU's test system:
Intel Core i7-3770K @ 4.7 GHz
http://www.techpowerup.com/reviews/Sapphire/HD_7950_Vapor-X/5.html
What that means is reviews that used stock Core i processors for example are somewhat holding back the full speed of GTX670/680 cards. At TPU, we don't have to worry about any of this since they isolate the GPU power consumption for us. So we see the average and peak power consumption for the GPU alone.
The main take away is both methods are fully valid, but both have limitations.
The reason I brought up after-market 7950/7970 cards is because some people keep linking power consumption #s of reference 7950/7970 cards, when those don't even matter for people looking to buy after-market 7950 cards for overclocking. I already showed that an MSI TF3 7950 uses just 167W of power at 1025mhz. So what's the point you discussing HD7950 vs. GTX670 power consumption, using a reference B 7950 card? It's irrelevant on this enthusiast forum since those are not the type of cards people here will buy for OCing.
For most enthusiasts looking to spend $300-320, they are. 20% more performance for almost the same price as a 660Ti OC. Most people would take that in 100s of games than PhysX in 1-2 games a year.
HD7950 @ 1150mhz > HD7970GE. Why did I just make this comparison? Using your logic then, you just suggested that you would choose a GTX660Ti level of GPU performance over HD7970 GE level of gaming performance at the same price level? Is that correct?
Last edited:
SirPauly
Diamond Member
- Apr 28, 2009
- 5,187
- 1
- 0
Personally can't speak for most enthusiasts!
There is no right time for all to buy and one can't speak for most enthusiasts, sorry! Personally purchased a GTX 670 when the HD 7950/GTX 670 were around the same. Market changes -- prices adjust -- strong competition creates more value and innovation. Personally glad to see 28nm price/performance improve.
- Sep 5, 2003
- 19,458
- 765
- 126
You haven't answered my question SirPauly This goes back to the topic of the GTX660Ti:
For overclockers:
Is Videocard A that delivers 20% more performance OC vs. the competitor's OC card at a similar price a better value (7950 OC vs. 660Ti OC)?
For non-overclockers:
Is Videocard A that delivers 90% of the performance for 25-30% lower price vs. the competitor's card a better value (HD7870 vs. 660Ti stock)?
For someone who wants MSAA, PhysX, why not spend a little more for the GTX670?
Are GTX660Ti price cuts reasonable?
Looks to me like GTX660Ti is overpriced. AIB's think it needs price cuts. Can't see how this is hurting the consumer. This is great for us!
Of course an overclocked 7970 GE will be a 7950 OC. My point was that HD7950 OC can surpass stock GTX680/HD7970 GE in performance on average. GTX660Ti cannot. If you look at the benchmarks, HD7950 OC = GTX670 OC, but costs $70-80 less. Why would any overclocker running an enthusiast PC gaming rig with Core i5/i7 buy the 660Ti over the after-market 7950 card?
Similarly stated, someone who doesn't overclock and wants better price/performance and better performance/watt would get the HD7870.
Someone who wants MSAA, PhysX would pay a little more for GTX670. A $300 price-level is too high for a card to tank with Mods and MSAA - that's exactly what 660Ti does. Also, if games get demanding, there is no reserve room to increase performance by much since a large number of 660Ti's GPU boost to 1150-1250mhz out of the box. See reviews.
Also, the AIBs have made a valid point that after-market 660Tis can often be priced very close to a GTX670. For example. Keep in mind that outside of North America, in many countries of the world, NV cards cost more. That makes them even worse value against 7870/7950. It's pretty hard to portray this when people deny that AMD offered better price/performance since 2009 at most price levels unless you waited to buy your card on some special sale. On average, AMD competed on price/performance since the take-over and the failed 2900XT/3870 series. This is a well-known fact, not a hypothesis. This is why the new management dropped this strategy and raised prices. People can't be crying that AMD raised prices and at the same time deny that they offered the best price/performance. That's a contradictory position.
I never said no NV card was worth buying. I said if the same criteria being presented here is used now: performance/watt, then sure, no NV card was worth buying since AMD had superior performance/watt. Say that isn't so?
When HD4870 launched for $299, GTX260 was $399. Pretty interesting though how people ignored AMD's undeniable price/performance advantage and now the same people are claiming AMD is ripping us off for raising prices (but NV isn't since they supposedly delayed the real flagship GK110 because AMD under-delivered). If NV could have launched GK110 by now, why didn't they? NV had no problems selling 8800GTX Ultra for $830. Looks to me that selling GK110 even at $830 was not profitable this year (unless NV intends to launch it sometime in the next 4 months), or maybe it was physically impossible to launch until Q4 2012 (Dec 2012 is the expected launch date for K20).
Pretty interesting to hear that GTX480/580 at $499 were not overpriced. Let's take a look:
GTX480 draws 270W of power at load vs. 143W for the 5870. That's a 126W power penalty. GTX480 ($499) and cost 35% more than HD5870 ($369) despite being only about 15% faster.
http://tpucdn.com/reviews/Palit/GeForce_GTX_660_Ti_Jet_Stream/images/power_peak.gif
Using performance/$ or performance/watt metric, GTX480 was terrible and yet certain people were fine with the 126W power penalty. I supposed this was justified by Compute, higher performance, PhysX....OK moving on, but the general trend was that power consumption was ignored. Now suddenly power consumption is the centre of attention and compute is irrelevant. Amazing.
Further, after-market HD7950s @ 1025mhz draw 167-175W of power and gives a gamer ~GTX680 level of performance for $180 less. This is ignored. GTX460 OCing while drawing 236W of power was perfectly fine against the HD5870. Amazing.
Even when HD7950 @ 1150mhz is 15-20% faster than a GTX660Ti OC for $10 more, the power consumption difference between these 2 cards is going to be less than 80W at load. So technically that means HD7950 looks a lot better than the 660Ti compared to how GTX480 did vs. the 5870. You still get a massive compute advantage, the same 15-20% performance advantage, more VRAM (same as 480 vs. 5870) minus the 35% price premium. Yet, the 7950 is not recommended over the 660Ti. Amazing.
HD6950 @ $250-300 unlocked into a 6970. GTX580 was about 20% faster than the 6970, give or take, but cost 67-100% more than an unlocked HD6950. Damn right, GTX580 was overpriced from a performance/$ perspective (especially since it couldn't even convincingly outperform the 6970 at 1600P).
If people stayed consistent in their message, so many threads wouldn't have to drag on for 10+ pages. I bought 470s because I don't care about power consumption and I care about overclocking. The same people who are now talking about power consumption as being critical still bought Fermi cards last 2 generations. This does not compute since AMD had the best performance/watt for 3 generations in a row. If power consumption is suddenly so important, these gamers would not have bought a single NV card since 2009. This is where the inconsistency comes from. The same for overclocking: why did it count for GTX460/560Ti/470s but not for 7950/7970s? Does not compute since when you overclocked 460/560Ti/470 cards, power consumption increased through the roof!
If someone prefers NV, they should just admit it and not beat around the bush. Twisting historical facts is not getting us anywhere. Otherwise, every thread is going to be derailed to show only aspects videocards that show NV winning. I've not even seen anyone here admit that FX5000 was an utter failure. Even when NV gets blown away, it's still not admitted. This generation NV lost the performance crown, not admitted despite HD7970 GE after-market costing less. Looks like GTX680 needs a price cut to have similar price/performance.
More amusing is when AMD launched ahead of NV and lowered prices, they are called "desperate." When there are rumored price cuts on NV cards because they are not selling well, a defence is mounted why NV's pricing as justified. Double standard?
And apparently high NV prices are AMD's fault only. They have nothing to do with NV trying to maintain their profit margins because of rising 28nm wafer prices. It can't be possible that it costs AMD less to manufacture a 212mm^2 Pitcairn chip than it costs NV to manufacture a 294mm^2 GTX660Ti chip?
The story continues:
Bitcoin mining on AMD cards - ignored, despite 100% proven to work @ a low rate of electricty costs.
Quiet after-market overclockable 7950/7970 cards - ignored. Reference cards that can't be purchased are linked to skew power consumption and noise level numbers.
Massive compute advantage and double precision performance for distributing computing projects - ignored.
Any performance advantage in games where AMD wins - ignored because AMD Gaming Evolved titles don't count, but TWIMTPB titles count because TWIMTPB titles are "more popular".
I don't think anyone of us had a problem when GTX480/580 were faster with Tessellation. As gamers we want to see situations where our cards are performing well and not performing well. That actually helps us.
Consistency is key.
This goes back to the topic of the GTX660Ti:For overclockers:
Is Videocard A that delivers 20% more performance OC vs. the competitor's OC card at a similar price a better value (7950 OC vs. 660Ti OC)?
For non-overclockers:
Is Videocard A that delivers 90% of the performance for 25-30% lower price vs. the competitor's card a better value (HD7870 vs. 660Ti stock)?
For someone who wants MSAA, PhysX, why not spend a little more for the GTX670?
Are GTX660Ti price cuts reasonable?
Looks to me like GTX660Ti is overpriced. AIB's think it needs price cuts. Can't see how this is hurting the consumer. This is great for us!
Of course an overclocked 7970 GE will be a 7950 OC. My point was that HD7950 OC can surpass stock GTX680/HD7970 GE in performance on average. GTX660Ti cannot. If you look at the benchmarks, HD7950 OC = GTX670 OC, but costs $70-80 less. Why would any overclocker running an enthusiast PC gaming rig with Core i5/i7 buy the 660Ti over the after-market 7950 card?
Similarly stated, someone who doesn't overclock and wants better price/performance and better performance/watt would get the HD7870.
Someone who wants MSAA, PhysX would pay a little more for GTX670. A $300 price-level is too high for a card to tank with Mods and MSAA - that's exactly what 660Ti does. Also, if games get demanding, there is no reserve room to increase performance by much since a large number of 660Ti's GPU boost to 1150-1250mhz out of the box. See reviews.
Also, the AIBs have made a valid point that after-market 660Tis can often be priced very close to a GTX670. For example. Keep in mind that outside of North America, in many countries of the world, NV cards cost more. That makes them even worse value against 7870/7950. It's pretty hard to portray this when people deny that AMD offered better price/performance since 2009 at most price levels unless you waited to buy your card on some special sale. On average, AMD competed on price/performance since the take-over and the failed 2900XT/3870 series. This is a well-known fact, not a hypothesis. This is why the new management dropped this strategy and raised prices. People can't be crying that AMD raised prices and at the same time deny that they offered the best price/performance. That's a contradictory position.
I never said no NV card was worth buying. I said if the same criteria being presented here is used now: performance/watt, then sure, no NV card was worth buying since AMD had superior performance/watt. Say that isn't so?
When HD4870 launched for $299, GTX260 was $399. Pretty interesting though how people ignored AMD's undeniable price/performance advantage and now the same people are claiming AMD is ripping us off for raising prices (but NV isn't since they supposedly delayed the real flagship GK110 because AMD under-delivered). If NV could have launched GK110 by now, why didn't they? NV had no problems selling 8800GTX Ultra for $830. Looks to me that selling GK110 even at $830 was not profitable this year (unless NV intends to launch it sometime in the next 4 months), or maybe it was physically impossible to launch until Q4 2012 (Dec 2012 is the expected launch date for K20).
Pretty interesting to hear that GTX480/580 at $499 were not overpriced. Let's take a look:
GTX480 draws 270W of power at load vs. 143W for the 5870. That's a 126W power penalty. GTX480 ($499) and cost 35% more than HD5870 ($369) despite being only about 15% faster.
http://tpucdn.com/reviews/Palit/GeForce_GTX_660_Ti_Jet_Stream/images/power_peak.gif
Using performance/$ or performance/watt metric, GTX480 was terrible and yet certain people were fine with the 126W power penalty. I supposed this was justified by Compute, higher performance, PhysX....OK moving on, but the general trend was that power consumption was ignored. Now suddenly power consumption is the centre of attention and compute is irrelevant. Amazing.
Further, after-market HD7950s @ 1025mhz draw 167-175W of power and gives a gamer ~GTX680 level of performance for $180 less. This is ignored. GTX460 OCing while drawing 236W of power was perfectly fine against the HD5870. Amazing.
Even when HD7950 @ 1150mhz is 15-20% faster than a GTX660Ti OC for $10 more, the power consumption difference between these 2 cards is going to be less than 80W at load. So technically that means HD7950 looks a lot better than the 660Ti compared to how GTX480 did vs. the 5870. You still get a massive compute advantage, the same 15-20% performance advantage, more VRAM (same as 480 vs. 5870) minus the 35% price premium. Yet, the 7950 is not recommended over the 660Ti. Amazing.
HD6950 @ $250-300 unlocked into a 6970. GTX580 was about 20% faster than the 6970, give or take, but cost 67-100% more than an unlocked HD6950. Damn right, GTX580 was overpriced from a performance/$ perspective (especially since it couldn't even convincingly outperform the 6970 at 1600P).
If people stayed consistent in their message, so many threads wouldn't have to drag on for 10+ pages. I bought 470s because I don't care about power consumption and I care about overclocking. The same people who are now talking about power consumption as being critical still bought Fermi cards last 2 generations. This does not compute since AMD had the best performance/watt for 3 generations in a row. If power consumption is suddenly so important, these gamers would not have bought a single NV card since 2009. This is where the inconsistency comes from. The same for overclocking: why did it count for GTX460/560Ti/470s but not for 7950/7970s? Does not compute since when you overclocked 460/560Ti/470 cards, power consumption increased through the roof!
If someone prefers NV, they should just admit it and not beat around the bush. Twisting historical facts is not getting us anywhere. Otherwise, every thread is going to be derailed to show only aspects videocards that show NV winning. I've not even seen anyone here admit that FX5000 was an utter failure. Even when NV gets blown away, it's still not admitted. This generation NV lost the performance crown, not admitted despite HD7970 GE after-market costing less. Looks like GTX680 needs a price cut to have similar price/performance.
More amusing is when AMD launched ahead of NV and lowered prices, they are called "desperate." When there are rumored price cuts on NV cards because they are not selling well, a defence is mounted why NV's pricing as justified. Double standard?
And apparently high NV prices are AMD's fault only. They have nothing to do with NV trying to maintain their profit margins because of rising 28nm wafer prices. It can't be possible that it costs AMD less to manufacture a 212mm^2 Pitcairn chip than it costs NV to manufacture a 294mm^2 GTX660Ti chip?
The story continues:
Bitcoin mining on AMD cards - ignored, despite 100% proven to work @ a low rate of electricty costs.
Quiet after-market overclockable 7950/7970 cards - ignored. Reference cards that can't be purchased are linked to skew power consumption and noise level numbers.
Massive compute advantage and double precision performance for distributing computing projects - ignored.
Any performance advantage in games where AMD wins - ignored because AMD Gaming Evolved titles don't count, but TWIMTPB titles count because TWIMTPB titles are "more popular".
I don't think anyone of us had a problem when GTX480/580 were faster with Tessellation. As gamers we want to see situations where our cards are performing well and not performing well. That actually helps us.
Consistency is key.
Last edited:
SirPauly
Diamond Member
- Apr 28, 2009
- 5,187
- 1
- 0
AnandThenMan
Diamond Member
- Nov 11, 2004
- 3,991
- 627
- 126
- Sep 5, 2003
- 19,458
- 765
- 126
Budget gamers FTW! :biggrin:
Minecraft.
Strange things in the Steam HW Survey under "All Videocards"/"DX11 GPU" if we infer anything out of it. The data suggests HD7970 has so far sold more units than HD7770 or HD7870 while HD7750 is not even in the chart of DX11 GPUs. Makes sense considering only about 4% of consumers buy $400+ GPUs.
HD6970 had 0.53% market share in June 2012, 0% in July 2012, 0.53% in August 2012....makes sense.
Last edited:
- Sep 5, 2003
- 19,458
- 765
- 126
Here is food for thought. If Steam's numbers are an accurate reflections of sales, what happened last quarter?
- AMD saw gains in the discrete desktop category (2.5%).
- Nvidia's desktop discrete shipments dropped 10.4% from last quarter
http://www.techpowerup.com/170575/G...Over-Last-Quarter-and-5.5-Over-Last-Year.html
Steam's Hardware Survey shows total market share for August 2012:
All Videocards
GTX680 = 0.53%
AMD HD7000 series = not a single card makes the list.
Conclusion: GTX680 sold more than the entire HD7000 series desktop discrete GPU line-up or at the very least a $500 GTX680 sold more units than any HD7750/7770/7850/7950/7970 videocard. Makes sense........
Last edited:
Lonbjerg
Diamond Member
- Dec 6, 2009
- 4,419
- 0
- 0
STEAM numbers were just fine for AMD to use when they had the main DX11 market-share...now you say AMD is wrong? :sneaky:
AnandThenMan
Diamond Member
- Nov 11, 2004
- 3,991
- 627
- 126
No they were 100% right to use the info in a press release because it made them look good. But that doesn't make the numbers any more reliable.
Statistics is fun! You can phrase numbers to represent whatever you want!
Fake Example:
Q1:
Total market size: 100 million units
Company A has 99% market share. Ships 99 million units.
Company B has 1% market share. Ships 1 million units.
Q2:
Total market size: 80 million units
Company A has 90% market share. Ships 72 million units.
Company B has 10% market share. Ships 8 million units.
These true statements can be written:
1. Company A unit shipments decreased 27.3% from last quarter.
2. Company B market share increased 9%.
OMG Company A is dead/dying/on life support!
A random web survey shows Company A selling 9 times as many units as Company B. OMG web survey must be useless/biased!
Fake Example:
Q1:
Total market size: 100 million units
Company A has 99% market share. Ships 99 million units.
Company B has 1% market share. Ships 1 million units.
Q2:
Total market size: 80 million units
Company A has 90% market share. Ships 72 million units.
Company B has 10% market share. Ships 8 million units.
These true statements can be written:
1. Company A unit shipments decreased 27.3% from last quarter.
2. Company B market share increased 9%.
OMG Company A is dead/dying/on life support!
A random web survey shows Company A selling 9 times as many units as Company B. OMG web survey must be useless/biased!
What's important from those figures is market share. Company B went from 1% to 10% in one quarter. Company A needs to find out why and fix it.
Personally, I would buy them while I still could.
Personally, I would buy them while I still could.
You make it sound like you could prevent that.
In reality, an absolute free market is a pipe dream. Consumers are never absolutely rational and informed actors. Sadly, marketing has the same or more sway than the product itself.
Yes, it makes sense. AMD's revenue was down from Q1 -> Q2. nVidia's geforce revenue jumped 15% from Q1 -> Q2. nVidia said that they are even supply contraint in the high end because of the demand. There is a reason why AMD cards lost nearly 30% of value in the last 4 months.
Look at the number of AMD related threads on the first page of the VC&G forums. People are buying AMD cards.
AMD has cut prices and I hope nVidia cuts prices across the board. I hope it forces AMD to respond again. These cards are still too expensive IMO this late in the model year.
AMD has cut prices and I hope nVidia cuts prices across the board. I hope it forces AMD to respond again. These cards are still too expensive IMO this late in the model year.
Keysplayr
Elite Member
- Jan 16, 2003
- 21,219
- 56
- 91
I remember buying 2nd tier GeForce 6800GT for 400.00 at comp USA. @ Same price as a GTX670 2nd tier card is today. It's not like we haven't seen these prices many times before folks. And I'm talking both sides. I paid 449.00 for a 2900XT. Almost 500 for a 5870.
What's different?
What's different?
toyota
Lifer
- Apr 15, 2001
- 12,957
- 1
- 0
NIGELG
Senior member
- Nov 4, 2009
- 852
- 31
- 91
Cognitive dissonance in a nutshell.
Weren't there some special edition 5870s with additional VRAM and stuff?
EDIT: Yup, Sapphire Toxic edition with 2GB Vram. Cost ~450$ on launch.
SirPauly
Diamond Member
- Apr 28, 2009
- 5,187
- 1
- 0
HD 7750 is on the list. 7750, 7770, 7850 and 7870 were around .76 compared to .08 for the HD 7970. The HD 6970 is 1.04.
http://store.steampowered.com/hwsurvey/directx/?sort=chg
NIGELG
Senior member
- Nov 4, 2009
- 852
- 31
- 91
The Steam survey is worthless...whether it's QUOTED by AMD,AMD fanatics or Nvidia fanatics.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 24K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter