Polish magazines RX480 benchmarks leaked - performance under 970

[Morgan MLGman]

1 minute ago, That Norwegian Guy said:

That's not attributable to hyperthreading (unless it's an Nvidia SLi system) but rather to cache. Notice how the games where you see a benefit are open world games (apart from GTA V which lost FPS with HT, as you can expect from many games due to the lower OC potential)

HOW IS THAT RELEVANT HERE, I SAID THAT AN i7 IS NECESSARY TO ELIMINATE POTENTIAL BOTTLENECKS, AND AS YOU CAN SEE, A 6600K DOES BOTTLENECK HIGH-END GPUS BY NOT GIVING THE BEST POSSIBLE FRAMERATES AND SHOULD NOT BE USED IN GPU BENCHMARKS. I NEVER SAID IT WAS DUE TO THE HYPERTHREADING.

 

For god's sake. I was right from the beginning.

[That Norwegian Guy]

6 minutes ago, Morgan MLGman said:

HOW IS THAT RELEVANT HERE, I SAID THAT AN i7 IS NECESSARY TO ELIMINATE POTENTIAL BOTTLENECKS, AND AS YOU CAN SEE, A 6600K DOES BOTTLENECK HIGH-END GPUS BY NOT GIVING THE BEST POSSIBLE FRAMERATES AND SHOULD NOT BE USED IN GPU BENCHMARKS. I NEVER SAID IT WAS DUE TO THE HYPERTHREADING.

 

For god's sake. I was right from the beginning.

 

Because it's an important distinction to make, as the "more cache" argument is only going to gain you miniscule amounts of improvement, whereas the "hyperthreading" argument matters a FUCKTON to Nvidia SLi setups (because of shortcuts made by Nvidia in drivers) also this video of yours does reek of SLi.

 

It also matters because in a couple of months, Intel isn't going to be the only kid on the block anymore, and AMD (unlike Intel) doesn't artificially lower the cache of cut-down CPUs. In fact, the FX-8350 and the FX-6300 both have 8MB of L3 Cache, and I'm willing to bet AMD will follow the same trend with ZEN, as there's almost never a reason to have a lower-cache bin of a chip.

[Morgan MLGman]

1 minute ago, That Norwegian Guy said:

 

Because it's an important distinction to make, as the "more cache" argument is only going to gain you miniscule amounts of improvement, whereas the "hyperthreading" argument matters a FUCKTON to Nvidia SLi setups (because of shortcuts made by Nvidia in drivers)

 

It also matters because in a couple of months, Intel isn't going to be the only kid on the block anymore, and AMD (unlike Intel) doesn't artificially lower the cache of cut-down CPUs. In fact, the FX-8350 and the FX-6300 both have 8MB of L3 Cache, and I'm willing to bet AMD will follow the same trend with ZEN, as there's almost never a reason to have a lower-cache bin of a chip.

It is all irrelevant. My point is: You need an i7-6700K ATM to do proper GPU benchmarks, while eliminating as many bottlenecks as possible to let the GPU run to its full potential. That is all I meant from the beginning. 6600K is not good enough. And we both know that I'm right.

[That Norwegian Guy]

3 minutes ago, Morgan MLGman said:

It is all irrelevant. My point is: You need an i7-6700K ATM to do proper GPU benchmarks, while eliminating as many bottlenecks as possible

 

Bullshit, if you really wanted to eliminate all bottlenecks, you'd get a golden-chip 6600K because a better overclock will always walk all over both hyperthreading and cache combined for everything.

 

The only reason you'll ever see reviewers get higher scores with HT chips is because they limit the OC they make on the non-HT chip relative to what the HT chip is able to get.

[Morgan MLGman]

2 minutes ago, That Norwegian Guy said:

 

Bullshit, if you really wanted to eliminate all bottlenecks, you'd get a golden-chip 6600K because a better overclock will always walk all over both hyperthreading and cache combined for everything.

 

The only reason you'll ever see reviewers get higher scores with HT chips is because they limit the OC they make on the non-HT chip relative to what the HT chip is able to get.

I'm done, you've got your own world and your own toys, I'm living in the reality. I've shown you legitimate proof that the 6700K is better, and yet you call bullshit, without providing ANY source behind your arguments. BTW, a "golden-chip" 6700K even if it was clocked lower than the 6600K by 0,2gHz it would still outperform the 6600K in a vast majority of games.

 

I'm done explaining to you why the better CPU is a better CPU, if it's hard to comprehend for you, then I'm sorry but I can't explain it any simpler.

[Morgan MLGman]

19 minutes ago, That Norwegian Guy said:

 

BTW. Do you know what frametimes are? With an i7, you get MUCH more consistent frametimes with MUCH less drops. It doesn't impact FPS but impacts smoothness of the gameplay significantly. i5s experience frametime drops, and HyperThreading helps with that. You can observe that in the video I linked, as they show frametimes as well.

 

Gosh, arguing with you feels like playing chess with a pigeon. It's gonna throw all of the pawns out from the board, shit on it and act like it won anyway.

[That Norwegian Guy]

41 minutes ago, Morgan MLGman said:

I'm done

good

[zMeul]

@Morgan MLGman that video would've been relevant if the i7 runs were with HT on and off

also, the i7 has some substantial frametime discrepancies at times - so, it's not as a clear cut as you make it to be

i7 has bigger cache and higher clocks

 

and last point: the video shows a CPU bound scenario not a GPU bound one - RX480 is a GPU bound problem no matter how you look at it

 

---

 

lastly, people that buy a RX480 would most likely have a i5 or even an i3 and not an i7 - and since all benchmarks (with all cards) were done on the exact same platform, it's relevant

[zMeul]

Quote

All I will say is I benched a 390X Strix OC in Ashes Singularity DX12, I then benched the RX 480 at stock, 1fps difference.

http://forums.overclockers.co.uk/showpost.php?p=29710504&postcount=8972

[Briggsy]

27 minutes ago, zMeul said:

people that buy a RX480 would most likely have a i5 or even an i3 and not an i7

This bothers me a lot, as no one in their right mind is going to pair a 480 and i7 (unless its an older first gen i7 they've had forever). The pre-built PCs with a 480 will probably have a 6300 or i3 in them, maybe a cheap i5. People are going to watch/read benchmarks done with overclocked i7's and be fooled into thinking they will get that performance too. Would be nice to see realistic benchmarks done by mainstream reviewers sometimes.

[dalekphalm]

2 minutes ago, Briggsy said:

This bothers me a lot, as no one in their right mind is going to pair a 480 and i7 (unless its an older first gen i7 they've had forever). The pre-built PCs with a 480 will probably have a 6300 or i3 in them, maybe a cheap i5. People are going to watch/read benchmarks done with overclocked i7's and be fooled into thinking they will get that performance too. Would be nice to see realistic benchmarks done by mainstream reviewers sometimes.

The problem with "realistic" benchmarks, is that they do not control for variables. The very literal definition of "Your mileage may vary". If they benched with an i5, the results could be different for any number of reasons.

 

Not to mention that many users won't be using an i5. Some will have an i7. Others will have an i3 or some AMD chip.

 

Where do you draw the line? By benching with an i7 (ideally an extreme edition balls to the wall CPU), you eliminate that variable.

 

If your game performs worse than what is benched, it's not because the GPU is at fault, it's because you're being bottlenecked somewhere else.

 

If reviewers had tons of free time, they could of course bench the GPU using a variety of CPU's, while trying to control the other variables (not always possible), but most do not have that kind of spare time.

[The Benjamins]

17 minutes ago, zMeul said:

http://forums.overclockers.co.uk/showpost.php?p=29710504&postcount=8972

I think that is awesome.

[zMeul]

3 minutes ago, The Benjamins said:

I think that is awesome.

if it would've been a game that actually matters, maybe

Ashes is a bad benchmark tool and a shit game

[Briggsy]

7 minutes ago, The Benjamins said:

I think that is awesome.

Makes one wonder. The 480 has 4 Async Compute Engines while the 390/x has 8 ACE's. Such a waste if 4 are adequate at that performance tier.

[The Benjamins]

5 minutes ago, zMeul said:

if it would've been a game that actually matters, maybe

Ashes is a bad benchmark tool and a shit game

We get to see how it preforms in a bunch of games soon.

 

I think it at least shows AMD generational improvement

 

Makes one wonder. The 480 has 4 Async Compute Engines while the 390/x has 8 ACE's. Such a waste if 4 are adequate.

 

Hawaii scaled better then fiji. not sure what fiji had though.

[Briggsy]

2 minutes ago, The Benjamins said:

We get to see how it preforms in a bunch of games soon.

 

I think it at least shows AMD generational improvement

 

 

Hawaii scaled better then fiji. not sure what fiji had though.

Fiji had 8 ACE's as well.

[zMeul]

3 minutes ago, The Benjamins said:

I think it at least shows AMD generational improvement

it performs worse in DX11 titles, so that "improvement" is a bad trade off since there aren't any relevant DX12 games on the market to actually make a difference

maybe after Vega, maybe after Volta

 

DX12 is absent, so is Vulkan .... and that makes me a sad cookie monster

[Notional]

6 hours ago, zMeul said:

 

160W power draw .. that's pathetic; it's 10W more than the GTX1070

it draws the same amount as a reference GTX970, for arguably the same perf ... that GloFo 14nm FinFET node is utter shit

It uses less than a 1080 you massive FUD machine. I guess that TSMC 16nm FF+ node is utter shit. Then again the yield is so low, no one can get a hold of a 1080 to test for themselves. How ironic.

6 hours ago, LAwLz said:

That's so damn expensive.

200 dollars in the US.

300 dollars in France.

320 dollars in Sweden.

 

And people accuse Nvidia of price gouging... I get that there is VAT included in the European prices but come the fuck on... We don't have 50-60% tax.

I really hope that's not what the card will actually cost when it is released in all stores.

Those prices re perfectly fine. You have import taxes, higher living costs, meaning higher prices, and VAT added on top. No gouging compared to any other electronics as far as I can see.

[Briggsy]

actually, come to think of it, 4 Aces would put 480's compute queue at 32 compute +1 graphic, similar in size to Nvidia's queue. Interesting choice.

[pbenologa]

You see a AMD Thread. You see zMeul. You abandon the thread as fast as you can. 

[Sharkyx1]

10 hours ago, Mr.Meerkat said:

I think the biggest flaw here is the magazine saying the 380X and 970 is OCed tells us nothing about their clocks...maybe the 970 was OCed at 1600Mhz with the 380X at 1400mhz...

It's a zotac that boosts to above reference 1080 speed. 1304 or something mhz

[ApolloFury]

19 hours ago, zMeul said:

to play at sub 60FPS?! can you explain to me the logic behind making that aquisition

 

You can get a 60hz Freesync monitor for $120-150. And who said you have to play all the latest games at max settings?

[zMeul]

24 minutes ago, ApolloFury said:

You can get a 60hz Freesync monitor for $120-150. 

I really like to see that magical 150$ VRR monitor

[kelvindeschutter]

9 minutes ago, zMeul said:

I really like to see that magical 150$ VRR monitor

i know its a Dutch site and that the prices are in euros but if you translate the cheapest will be below 150 usd: https://tweakers.net/categorie/344/monitoren/producten/#filter:q1bKL0pJLXLLTM1JUbJSKijKzCpW0oEIBucXlQDFEouT4SIFqcmeQHW6hjogpcmpvpl5SlYGOkrFQAm3zJyS1KJiJatqJRMzcxMQXZaYAzTAwtzURKm2thYA

[zMeul]

5 minutes ago, kelvindeschutter said:

i know its a Dutch site and that the prices are in euros but if you translate the cheapest it will be below 150 usd: https://tweakers.net/categorie/344/monitoren/producten/#filter:q1bKL0pJLXLLTM1JUbJSKijKzCpW0oEIBucXlQDFEouT4SIFqcmeQHW6hjogpcmpvpl5SlYGOkrFQAm3zJyS1KJiJatqJRMzcxMQXZaYAzTAwtzURKm2thYA

get out of here ... AOC G2260VWQ6 VRR window 48-75Hz - no wonder it's cheap