Geforce GTX 1070 vs 2x Radeon RX 480 CrossFire (PCLab)

[Bacon1]

Extra 25W for higher clocks on average, seems reasonable.

By a lot you mean 14-15W

For a tiny OC yes, that was @ 2062 Mhz / 2128 MHz Memory vs 1987 MHz / 2027 MHz Memory

3-4% core OC, 5% memory for extra 15w

Still like to see actual gaming power usage #s as those seem to be higher than 3dmark.

The benchmarks however were @ 2101 / 2420 or 5.7% core OC, 20% memory

So those will use quite a lot more than 15w extra, which is why its dishonest to use results from multiple different reviews just so you can cherry pick the results.

But we went from a 150 TDP (actual 160 from your post) to almost 200? That's a 30% increase in power usage. More when factoring in the higher OC from techpowerup review. Guess the 6+8 pin helps there.

Considering that card is $450 it makes it a great deal over the 1080 but very odd to compare to reference 480s.

 

[Silverforce11]

I didn't need to read the article, saw PCLab, knew right away it would be horribly negative on AMD GPU. It's like posting ABT benches. Do you know why nobody post their benchmarks for discussion? Think real hard.

You want to see a proper review?

https://www.youtube.com/watch?v=DGdhW_GF3dg

2x RX 480 4GB $199 each, not bad at all for $398. More now that we know all the reference RX 480 is actually 8GB.

 

[Sweepr]

So those will use quite a lot more than 15w extra, which is why its dishonest to use results from multiple different reviews just so you can cherry pick the results.

Both Guru3D TDP numbers and PCLab's test at the OP used a stock MSI Geforce GTX 1070 Gaming X, so I'm not sure what you're complaining about.

But we went from a 150 TDP (actual 160 from your post) to almost 200? That's a 30% increase in power usage. More when factoring in the higher OC from techpowerup review. Guess the 6+8 pin helps there.

''On this sample I was able to overclock the core to +125 MHz from 1607 MHz base clock (OC Mode) boosting to 1987 MHz, to a final boost clock of 2126 MHz. The memory boosted up over 200 MHz to 2230 MHz or 8290 MHz GDDR5.''

Only 14-15W extra for a nice overclock that would probably put it at Geforce GTX 1080 FE level, at lower power/heat than Radeon RX 480 CrossFire and without multi GPU scaling headaches / worse frame times.

Considering that card is $450 it makes it a great deal over the 1080 but very odd to compare to reference 480s.

Why is it odd? AMD itself compared their cards to Geforce GTX 1080 but conveniently left out the Geforce GTX 1070, which offers better perf/$. Try Geforce GTX 1070 FE if you'd like (reference vs reference), won't change much.

 

[DooKey]

I didn't need to read the article, saw PCLab, knew right away it would be horribly negative on AMD GPU. It's like posting ABT benches. Do you know why nobody post their benchmarks for discussion? Think real hard.

snip

Kinda why lots of people scoff at those that worship at the SA throne. Think real hard.

 

[Silverforce11]

Kinda why lots of people scoff at those that worship at the SA throne. Think real hard.

Do we have regular posters here who worship Semi-Accurate? It's not common here to have SA tagged threads for discussion. In fact, I would say its poor etiquette to bring up such bias sources and it's an auto-loss for people who present evidence from these same sources.

I can tell you right now, PClab faked the Hitman results. There's zero chance a 980 beats a 390X in that title DX11 or DX12.

Compare to other credible sites:

https://www.computerbase.de/2016-06/radeon-rx-480-test/5/#diagramm-hitman-2560-1440

https://www.techpowerup.com/reviews/AMD/RX_480/16.html

http://www.guru3d.com/articles_pages/amd_radeon_r9_rx_480_8gb_review,11.html

They have a history of producing data that do not match other reputable sites. Just like ABT. So if they are making up numbers as they see fit, why are people here using these clown sites for a source?

 

[Bacon1]

Both Guru3D TDP numbers and PCLab's test at the OP used a stock MSI Geforce GTX 1070 Gaming X, so I'm not sure what you're complaining about.



''On this sample I was able to overclock the core to +125 MHz from 1607 MHz base clock (OC Mode) boosting to 1987 MHz, to a final boost clock of 2126 MHz. The memory boosted up over 200 MHz to 2230 MHz or 8290 MHz GDDR5.''

Why is it odd? AMD itself compared their cards to Geforce GTX 1080 but conveniently left out the Geforce GTX 1070, which offers better perf/$. Try Geforce GTX 1070 FE if you'd like (reference vs reference), won't change much.

The graph is from before he manually OCd, he was testing the built in "OC" mode.

After finding the small 24/7 overclock above, its time to really put it on the card and see what we can come up with. I was pretty low with the power % in MSI afterburner so I raised the power to 75% to see what would happen. My starting core voltage according to GPU-Z was 1.03V. My ending voltage after applying settings was 1.06V… that is within a typical spread to me. It appears the voltage didn’t change. Either that or GPU-Z isn’t reading it right and MSI AB doesn’t seem to read it either (v4.3.0 Beta 4). Bummer

So the extra power use when heavily OC'd like TPU will be much more than 15w.

 

[DooKey]

Do we have regular posters here who worship Semi-Accurate? It's not common here to have SA tagged threads for discussion. In fact, I would say its poor etiquette to bring up such bias sources and it's an auto-loss for people who present evidence from these same sources.

I can tell you right now, PClab faked the Hitman results. There's zero chance a 980 beats a 390X in that title DX11 or DX12.

Compare to other credible sites:

https://www.computerbase.de/2016-06/radeon-rx-480-test/5/#diagramm-hitman-2560-1440


https://www.techpowerup.com/reviews/AMD/RX_480/16.html

http://www.guru3d.com/articles_pages/amd_radeon_r9_rx_480_8gb_review,11.html

They have a history of producing data that do not match other reputable sites. Just like ABT. So if they are making up numbers as they see fit, why are people here using these clown sites for a source?

Exactly, why do people here use clown sites like PClab and SA? It happens and yet it's not isolated to just one or two posters. Unfortunately both sides use their biased sites. Anyone trying to pretend it isn't shills from both sides is disingenuous.

 

[Sweepr]

The graph is from before he manually OCd, he was testing the built in "OC" mode.

So the extra power use when heavily OC'd like TPU will be much more than 15w.

TPU got 2101/2420 MHz. This website managed 2062/2128 MHz, which should still be very close to Geforce GTX 1080 FE (stock) levels of performance, and generally ahead of RX 480 CF at lower power/heat - without any scaling concern.

>120W gap if you add 14-15W from the OC above. It's not like you need to OC though, stock Geforce GTX 1070 (especially custom like MSI's Gaming X) already stacks up really well.

 

[iiiankiii]

Scary.

Not at all. I'm not saying the RX480 is efficient in respect to the latest. I'm saying a sub 500 watt 2X GPU system is NOTHING NEW. It's actually considered efficient. Are you new to the PC enthusiast space? There are a brick ton of people out there with 2x GTX970/980/980TI/TitanX with well over that power consumption. Heck, most dual GPU sets out there right now are more power hungry. SCARY and shocking, right?

 

[nurturedhate]

Not at all. I'm not saying the RX480 is efficient in respect to the latest. I'm saying a sub 500 watt 2X GPU system is NOTHING NEW. It's actually considered efficient. Are you new to the PC enthusiast space? There are a brick ton of people out there with 2x GTX970/980/980TI/TitanX with well over that power consumption. Heck, most dual GPU sets out there right now are more power hungry. SCARY and shocking, right?

Marketers gotta push that marketing. Gotta make the boss happy.

 

[Gorbugal]

Transparent as ever. Using the most skewed results, implying sub 500w dual card, total system power usage is "scary". Cancerous.

 

[parvadomus]

Oh man, I really want to work as a Nvidia riot, they must pay very good... they are even doing extra hours lately XD

 

[guskline]

Makes no sense to buy a 1080!

Sorry to disagree with you but I love the GTX 1080.

I wanted the fastest single card for my 5960x. I'm done with the multi gpu setups.

At some point faster gpus will be released but the GTX 1080 even without OCing is blistering fast on my 3440 x 1440 widescreen.

 

[Osiris-TEP]

Oh man, I really want to work as a Nvidia riot, they must pay very good... they are even doing extra hours lately XD

Way to cherry pick the original data from a site with, as Silverforce stated, internal locus against AMD; in my profession we call this confirmation bias.

BTW--
Overtime is always paid in cold, hard CASH!

 

[frozentundra123456]

Transparent as ever. Using the most skewed results, implying sub 500w dual card, total system power usage is "scary". Cancerous.

So instead of that "cancerous" post, you would recommend a dual card solution (with all its scaling issues) that whether you like it or not uses twice the power and costs approximately the same? OK then. Me thinks you protest too much.

 

[antihelten]

So instead of that "cancerous" post, you would recommend a dual card solution (with all its scaling issues) that whether you like it or not uses twice the power and costs approximately the same? OK then. Me thinks you protest too much.

That's a beautiful straw man you got going there.

Gorbugal made made no recommendations regarding 1070 vs 480, one way or the other. The only "recommendation" he made was that people should stop using biased sources when trying to make their point, why is that so hard to understand?

 

[Jaydip]

So people saying this is biased prove it! it's as simple as that, benchmarks can vary wildly depending on scenes.

 

[IllogicalGlory]

So people saying this is biased prove it! it's as simple as that, benchmarks can vary wildly depending on scenes.

Probably because the results they post are outliers in every sense of the word, and only ever in favor of one side. Defending that site and its results is laughable - and it comes from the same people who make personal attacks against other forum members for consistently posting news stories that prop the other team.

"Benchmarks vary wildly" - that is certainly true and it's easy to (cherry)pick from those wild variations results that fit the agenda.

 

[Jaydip]

Probably because the results they post are outliers in every sense of the word, and only ever in favor of one side. Defending that site and its results is laughable - and it comes from the same people who make personal attacks against other forum members for consistently posting news stories that prop the other team.

"Benchmarks vary wildly" - that is certainly true and it's easy to (cherry)pick from those wild variations results that fit the agenda.

No need to defend anything, I said if you believe they are wrong prove it, throwing wild accusations doesn't make it right as well. It is very funny when I see they consistently get hardware from AMD while still in "NV's pocket" , AMD had blacklisted many reviewers but this site was not among them. I remember the same story about TR when they are called shill site for NV but their editor went to work for AMD

 

[Rvenger]

Way to cherry pick the original data from a site with, as Silverforce stated, internal locus against AMD; in my profession we call this confirmation bias.

BTW--
Overtime is always paid in cold, hard CASH!

Looks like we are resorting to trolling and attacking other members.

I guess we're all done here. Tomorrow I will be tallying up the 30 day bans so stay tuned.

-Rvenger