Pascal offers no IPC gain?

[EuphoricRage470]

What a silly article.. Performance isn't going to scale linearly with clock speed. I bet you if somebody took a Maxwell and clocked it much higher (i.e. Pascal level clocks), "IPC" would go down.

 

[pxc]

The author is making bad conclusions from inappropriate extrapolations.

 

[Elmy]

No. The performance boost is calculated relative to the 980. THE 980.

It's not replacing the 980Ti. It's replacing the 980. Maybe if you weren't so busy receiving free hardware from AMD you'd stick your head out of the sand and realize you're just talking baseless crap.

Hilariously hypocritical baseless crap as well; how big a boost is Fury X over 290x hm ? Fury X replaces 290x btw. Do they achieve anything close to the 60-70% increase that 1080 has over 980 ? Nope. But you can't talk about that because AMD will cut you off from receiving your cards

Doesn't matter whats it replacing.... People are selling their 980Ti's as we speak to " upgrade to the 1080 " .... go read the for sale threads and ebay... 980 Ti's being sold for the cheap because people are upgrading to the 1080. And they will sell their 1080's when the 1080 Ti's come out.

And when did this ever become about AMD... Going way out of context here just to attack AMD any chance you get? A little salty are we?

 

[Ieldra]

Doesn't matter whats it replacing.... People are selling their 980Ti's as we speak to " upgrade to the 1080 " .... go read the for sale threads and ebay... 980 Ti's being sold for the cheap because people are upgrading to the 1080. And they will sell their 1080's when the 1080 Ti's come out.

And when did this ever become about AMD... Going way out of context here just to attack AMD any chance you get? A little salty are we?

People are selling 980Tis to buy 1080s. Okay. So I'm supposed to form an opinion of this product, based on what people are selling in order to buy it?

This makes about as much sense as a quad-gpu system; none.

I'm not attacking AMD, I'm just pointing out your hypocrisy, it's loads of fun, you should try it sometime. A bit of introspection can only do you good, looking forward to your Polaris build logs

My friend is selling a 390x to buy Polaris 10 for his small form Factor pc. I guess that means AMD is only offering 10% improvement. QQ

 

[Elmy]

People are selling 980Tis to buy 1080s. Okay. So I'm supposed to form an opinion of this product, based on what people are selling in order to buy it?

This makes about as much sense as a quad-gpu system; none.

I'm not attacking AMD, I'm just pointing out your hypocrisy, it's loads of fun, you should try it sometime. A bit of introspection can only do you good, looking forward to your Polaris build logs

My friend is selling a 390x to buy Polaris 10 for his small form Factor pc. I guess that means AMD is only offering 10% improvement. QQ

I agree with quad gpu systems... doesn't make any sense except for setting world records.... drivers are a pain ... only a handful of games benefit from it.... But it sure looks cool

 

[JustReason]

People are selling 980Tis to buy 1080s. Okay. So I'm supposed to form an opinion of this product, based on what people are selling in order to buy it?

This makes about as much sense as a quad-gpu system; none.

I'm not attacking AMD, I'm just pointing out your hypocrisy, it's loads of fun, you should try it sometime. A bit of introspection can only do you good, looking forward to your Polaris build logs

My friend is selling a 390x to buy Polaris 10 for his small form Factor pc. I guess that means AMD is only offering 10% improvement. QQ

Actually he is right. It doesn't matter what the actual tier card it is replacing. What matters is the top card current compared to the new top card coming. The top available 9xx series is the 980Ti. The top 10xx series is the 1080. Now say you who has a 980Ti waits for the direct replacement, the 1080Ti, then that is somewhat reasonable as the time frame between the two and you are getting the top card for each series. What he is saying is that a great deal of these people are scrambling to get the 1080 by selling their current 980Ti, and likely will repeat the process when the 1080Ti comes out. And above all that the visual differences between the current 2 is likely indiscernable.

And your friends change makes a little sense in that small form factors favor more efficient GPUs mainly less heat. If it was for a desktop then he could be considered as daft as the aforementioned people.

 

[Ieldra]

Actually he is right. It doesn't matter what the actual tier card it is replacing. What matters is the top card current compared to the new top card coming. The top available 9xx series is the 980Ti. The top 10xx series is the 1080. Now say you who has a 980Ti waits for the direct replacement, the 1080Ti, then that is somewhat reasonable as the time frame between the two and you are getting the top card for each series. What he is saying is that a great deal of these people are scrambling to get the 1080 by selling their current 980Ti, and likely will repeat the process when the 1080Ti comes out. And above all that the visual differences between the current 2 is likely indiscernable.

And your friends change makes a little sense in that small form factors favor more efficient GPUs mainly less heat. If it was for a desktop then he could be considered as daft as the aforementioned people.

I was joking about my friend selling a 390x for Polaris. I don't have any friends.

Yes, it would be daft, just as daft as comparing the 1080 to a 980ti. I don't care who sells what, I don't care who buys what.

The 1080 replaces the 980. It brings a 60%+ improvement.

When the 1080ti comes you can compare it to the 980ti.

Lol, by this logic when Polaris comes out it'll be the fastest gcn3 gpu, so we should compare it to the Fury X. Oh. That'll be good for a laugh. Elmy complains about meager 25% gains over 980ti, can't wait to hear his thoughts on performance recession when comparing Polaris 10 to Fiji XT

 

[misterbobby]

I was joking about my friend selling a 390x for Polaris. I don't have any friends.

Yes, it would be daft, just as daft as comparing the 1080 to a 980ti. I don't care who sells what, I don't care who buys what.

The 1080 replaces the 980. It brings a 60%+ improvement.

When the 1080ti comes you can compare it to the 980ti.

Lol, by this logic when Polaris comes out it'll be the fastest gcn3 gpu, so we should compare it to the Fury X. Oh. That'll be good for a laugh. Elmy complains about meager 25% gains over 980ti, can't wait to hear his thoughts on performance recession when comparing Polaris 10 to Fiji XT

When a 1080 costs more than the 980 ti launched at it will damn well be compared to it. And Nvidia has said on more than one occasion that the $700 FE card will be in the "middle of the price stack" which is fucking laughable for a mid range level die that used to carry x60 and x60ti skus.

 

[Ieldra]

When a 1080 costs more than the 980 ti launched at it will damn well be compared to it. And Nvidia has said on more than one occasion that $700 FE card will be in the "middle of the price stack" which is fucking laughable for a mid range level die that used to carry x60 and x60ti skus.

I genuinely didn't know the 980ti launched at less than $599

 

[razor1]

When a 1080 costs more than the 980 ti launched at it will damn well be compared to it. And Nvidia has said on more than one occasion that the $700 FE card will be in the "middle of the price stack" which is fucking laughable for a mid range level die that used to carry x60 and x60ti skus.

It is NOT a mid range die, its a performance level die, and it is NOT an enthusiast level die either. Also the launch price of AIB partners is $599 and the 980ti was $650, the FE boards, if you want to buy one that's up to you, I don't think they are "hot" commodities.

 

[cageymaru]

When a 1080 costs more than the 980 ti launched at it will damn well be compared to it. And Nvidia has said on more than one occasion that $700 FE card will be in the "middle of the price stack" which is fucking laughable for a mid range level die that used to carry x60 and x60ti skus.

I always compare products based on price personally. That's how [H]ardocp reviews compare them also. If a card is $699 then it shall get compared to something equitable to it price wise and the closest thing is the 980ti. I think that the GTX 1080 is worth the $50 premium on the GTX 980ti even though it thermal throttles like a SOB. Yes, it is a $699 card to me as I water cool my system.

 

[misterbobby]

I genuinely didn't know the 980ti launched at less than $599

For someone that likes to use the word "daft" it sure does apply to you right now. I made it clear I was talking about the $700 FE card which is 50 bucks more than the 980 Ti launched at. And AGAIN Nvidia has made it clear that the FE is in the middle of the price stack so expect many models costing way more than 700 bucks.

 

[misterbobby]

It is NOT a mid range die, its a performance level die, and it is NOT an enthusiast level die either. Also the launch price of AIB partners is $599 and the 980ti was $650, the FE boards, if you want to buy one that's up to you, I don't think they are "hot" commodities.

It is the SAME level die that used to give us x60 and x60 ti cards before the 680 so dont play dumb. It has no business carrying a $700+ plus price tag.

 

[JustReason]

For someone that likes to use the word "daft" it sure does apply to you right now. I made it clear I was talking about the $700 FE card which is 50 bucks more than the 980 Ti launched at. And AGAIN Nvidia has made it clear that the FE is in the middle of the price stack so expect many models costing way more than 700 bucks.

Actually I used the word daft first, but your point is made. lol

 

[Ieldra]

It is the SAME level die that used to give us x60 and x60 ti cards before the 680 so dont play dumb. It has no business carrying a $700+ plus price tag.

What's a die level? I've never heard of die levels.

 

[misterbobby]

What's a die level? I've never heard of die levels.

So just keep showing how "daft" you really are.

 

[razor1]

It is the SAME level die that used to give us x60 and x60 ti cards before the 680 so dont play dumb. It has no business carrying a $700+ plus price tag.

die size ok?

die size predominately creates the cost of the GPU, and this influences the cost of the card. You want these graphics card companies to be around? You dont' want them to drop margins like AMD did in the past as they will not have enough $ to properly fuel R&D for next gen chips.

Cost isn't just wafers and manufacturing chips its also cost to create the designs, this is why per transistor its x2 the cost on these new nodes.

Why do you think nV really pushed to get performance card bracket sales up? Because mid range is no longer creating enough margins like they used to. To keep that 40% margin range they need charge more now. But they aren't increasing the cost of GPU's as fast as the cost of R&D and manufacturing has been going up. Thankfully. Otherwise the prices would be much higher.

 

[Ieldra]

So just keep showing how "daft" you really are.

Why is daft in quotation marks? You clearly don't know how to use them. Let me help you.

When you talk about gpus, and you mention "die levels", that warrants the quotation marks.

 

[misterbobby]

die size ok?

die size predominately creates the cost of the GPU, and this influences the cost of the card. You want these graphics card companies to be around? You dont' want them to drop margins like AMD did in the past as they will not have enough $ to properly fuel R&D for next gen chips.

Cost isn't just wafers and manufacturing chips its also cost to create the designs, this is why per transistor its x2 the cost on these new nodes.

Yes I know that but you also know that x104/x204 gpu chips used to be for true mid range cards. It was pretty bad when Nvidia charged 500 bucks for the 680 which would have certainly been a 660 ti had the 7970 been more impressive. Now they have really lost their minds charging $700 for a x104 equipped card with their no so great cooler.

 

[Ieldra]

Yes I know that but you also know that x104/x204 gpu chips used to be for true mid range cards. It was pretty bad when Nvidia charged 500 bucks for the 680 which would have certainly been a 660 ti had the 7890 been more impressive. Now they have really lost their minds charging $700 for a x104 equipped card with their no so great cooler.

Nvidia seem sane to me. It's the public that has lost their minds. Nobody is forcing you to buy the FE. It sucks ass. There. I said it.

The FE sucks ass, dirty donkey ass.

Don't buy that shit, save yourself $100 and buy the AIB ones.

 

[razor1]

Yes I know that but you also know that x104/x204 gpu chips used to be for true mid range cards. It was pretty bad when Nvidia charged 500 bucks for the 680 which would have certainly been a 660 ti had the 7970 been more impressive. Now they have really lost their minds charging $700 for a x104 equipped card with their no so great cooler.

Why did AMD charge $650 for the Fury X than, it was a poor card for that price don't you think? or do you have other words for it? How about $650 for the nano, I mean great its small. Isn't bigger better? Where is your logic coming from.......

Who gives a shit about the die size when its the end performance that dictates everything....

 

[Quix]

Even if all the assumptions the author makes are 100% correct (and frankly he seems like a real tosser) who cares? Performance is the only thing that matters, higher clocks, higher IPC, more RAM bandwidth. I really don't care how they do it. Why would you?

 

[razor1]

Even if all the assumptions the author makes are 100% correct (and frankly he seems like a real tosser) who cares? Performance is the only thing that matters, higher clocks, higher IPC, more RAM bandwidth. I really don't care how they do it. Why would you?

All his assumptions are wrong because there is no equalizing point to look from..... To calculate IPC you must first create a equalized comparison point then get the IPC from the performance metrics. You can't change frequency and just run the numbers doesn't work that way.

If that was the case, where does AMD get the 40% increase in IPC for Zen? You need to isolate certain things to get the IPC increase amounts.

 

[Ieldra]

Why did AMD charge $650 for the Fury X than, it was a poor card for that price don't you think? or do you have other words for it? How about $650 for the nano, I mean great its small. Isn't bigger better? Where is your logic coming from.......

Who gives a shit about the die size when its the end performance that dictates everything....

I'm building a mosaic with gpu dies and I have a 600mm² hole in it. I demand large dies. I demand large dies now.

 

[misterbobby]

Nvidia seem sane to me. It's the public that has lost their minds. Nobody is forcing you to buy the FE. It sucks ass. There. I said it.

The FE sucks ass, dirty donkey ass.

Don't buy that shit, save yourself $100 and buy the AIB ones.

And what you cant seem to comprehend is that Nvidia keeps indicating that the good aftermarket cards will be costing MORE than the FE. How many times do I have to tell you that they have said the FE is priced in the middle of stack. Pricing the unimpressive FE at 700 bucks simply allows any of the AIB partners to justifiably charge more for their much better designs.

 

[limitedaccess]

I don't really understand the focus on IPC from a practical consumer stand point. Albeit from the curiosity stand point is a different matter.

IPC started to become a "buzz word" concept surrounding CPUs as in practice we saw limitations in clock speed scaling and core count scaling and how they applied to actual end user performance.

With GPUs though neither of those issues have shown to be limitations yet.

Even in terms of CPUs its practical informativeness, and this pertains to the current AMD Zen marketing information (this is veering off topic). is rather low. At the end for the end user Zen IPC (and AMD IPC vs Intel) is not really the issue, it has always been ST (single thread) performance deficits over the last few generations. At the end what people really want to know is how Zen will compare in terms of ST vs Intel's offerings and not IPC.

 

[razor1]

And what you cant seem to comprehend is that Nvidia keeps indicating that the good aftermarket cards will be costing MORE than the FE. How meany times do I have to tell you that they have said the FE is priced in the middle of stack. Pricing the unimpressive FE at 700 bucks simply allows any of the AIB partners to justifiably charge more for their much better designs.

reviews come out in the next day or two I think you will be eating that statement lol.

 

[Ieldra]

And what you cant seem to comprehend is that Nvidia keeps indicating that the good aftermarket cards will be costing MORE than the FE. How meany times do I have to tell you that they have said the FE is priced in the middle of stack. Pricing the unimpressive FE at 700 bucks simply allows any of the AIB partners to justifiably charge more.

Well the watercooled cards will cost more, but there will be plenty at $599. Maybe wait and see when they release before... Whatever it is you call this you're doing

 

[razor1]

I don't really understand the focus on IPC from a practical consumer stand point. Albeit from the curiosity stand point is a different matter.

IPC started to become a "buzz word" concept surrounding CPUs as in practice we saw limitations in clock speed scaling and core count scaling and how they applied to actual end user performance.

With GPUs though neither of those issues have shown to be limitations yet.

Even in terms of CPUs its practical informativeness, and this pertains to the current AMD Zen marketing information (this is veering off topic). is rather low. At the end for the end user Zen IPC (and AMD IPC vs Intel) is not really the issue, it has always been ST (single thread) performance deficits over the last few generations. At the end what people really want to know is how Zen will compare in terms of ST vs Intel's offerings and not IPC.

Exactly. The author of that article is just inept.

 

[misterbobby]

reviews come out in the next day or two I think you will be eating that statement lol.

That makes no sense. I simply going by what Nvidia has been saying so if we have some good aftermarket cards at $599 then great. That is not what they have been indicating though as they seem to say the better quality cards will cost more than the FE.

 

[razor1]

modified cards will be more expensive that is the norm, but those cards will be better than the FE editions at lower costs.

EVGA super clocks are going to go for $650 bucks. Just like they have charged before for SC variants in the past. You get better cooling, higher base clocks, and all that for those $50s. The FE cards were never ment to be competition to the AIB brands.

 

[Ieldra]

modified cards will be more expensive that is the norm, but those cards will be better than the FE editions at lower costs.

EVGA super clocks are going to go for $650 bucks. Just like they have charged before for SC variants in the past. You get better cooling, higher base clocks, and all that for those $50s. The FE cards were never ment to be competition to the AIB brands.

Which is exactly why the cost so much, to avoid competing with AIBs.

In exchange you get a full Metal shroud and Vapor Chamber cooler with best in class blower design.

 

[primetime]

Which is exactly why the cost so much, to avoid competing with AIBs.

In exchange you get a full Metal shroud and Vapor Chamber cooler with best in class blower design.

I think those vapor blower cards can be pretty good for those buying 2 cards and using certain cases.....they have a certain "look" to them as well. More choices is always welcome i you ask me. People dont want to pay full price?.....Just wait a while as always

 

[CSI_PC]

I always compare products based on price personally. That's how [H]ardocp reviews compare them also. If a card is $699 then it shall get compared to something equitable to it price wise and the closest thing is the 980ti. I think that the GTX 1080 is worth the $50 premium on the GTX 980ti even though it thermal throttles like a SOB. Yes, it is a $699 card to me as I water cool my system.

Thanks for linking the video.
He just confirmed what a few of us have been saying; increase the power and temp targets and you hit a consistent boost, his card stabilised at 1785MHz and that was without changing the fan profile.
Interesting he commented that when the fan hit 67% it was still much quieter than previous models.
TBH I would look to make the fan profile more aggressive but not to 80% like he did just to show it can be kept below 80degrees (it kept droping to around 77degrees).
Just to note as some may not watch the video, increasing target temp to 92 did not mean the card reached that temp, highest temp was 87degrees with fan left on defaults.
This shows that the Boost3 algorithm needs more tweaking, or if left to the user the FE model will need a couple of tweaks albeit quite easy to apply.

Cheers

 

[cageymaru]

Thanks for linking the video.
He just confirmed what a few of us have been saying; increase the power and temp targets and you hit a consistent boost, his card stabilised at 1785MHz and that was without changing the fan profile.
Interesting he commented that when the fan hit 67% it was still much quieter than previous models.
TBH I would look to make the fan profile more aggressive but not to 80% like he did just to show it can be kept below 80degrees (it kept droping to around 77degrees).
Just to note as some may not watch the video, increasing target temp to 92 did not mean the card reached that temp, highest temp was 87degrees with fan left on defaults.
This shows that the Boost3 algorithm needs more tweaking, or if left to the user the FE model will need a couple of tweaks albeit quite easy to apply.

Cheers

Yes, but also remember that he was using an open test bench in the video. I would assume that a video card could possibly run hotter in an enclosed computer case. I've never heard a reference fan that was quiet when run at 80%; maybe Nvidia has some new fan tech also.

Glad you found it informative.

 

[CSI_PC]

Yes, but also remember that he was using an open test bench in the video. I would assume that a video card could possibly run hotter in an enclosed computer case. I've never heard a reference fan that was quiet when run at 80%; maybe Nvidia has some new fan tech also.

Glad you found it informative.

Good point but I do not think it is making too much of a difference, as I mentioned in the past other reviewers who did the same also had good stable boost while using enclosed case; gamersnexus went as far as disabling all case fans (not talking about the OC throttle due to fans left on default profile); those reviewers who showed same throttling behaviour also showed the same benefits when doing the temp/power target change.
Yeah he should had tried 75% and go from there depending upo results, but he mentioned it being quiet at 68% not 80%.
Cheers

 

[CSI_PC]

Yes, but also remember that he was using an open test bench in the video. I would assume that a video card could possibly run hotter in an enclosed computer case. I've never heard a reference fan that was quiet when run at 80%; maybe Nvidia has some new fan tech also.

Glad you found it informative.

BTW where did you see Jay using an open test bench?
I could not see it in the video but interested.
Makes no difference to end result IMO, due to the other reviews that showed exactly same trend/behaviour and fix while using a PC case.
Cheers

 

[cageymaru]

BTW where did you see Jay using an open test bench?
I could not see it in the video but interested.
Makes no difference to end result IMO, due to the other reviews that showed exactly same trend/behaviour and fix while using a PC case.
Cheers

Look behind him / to the right of him towards the end of the video. Or listen to the 12:50 - 13:20. It is an open case to give the card the best possible temperatures. That's why I said in a conventional closed case it could be much worse.

 

[CSI_PC]

Look behind him / to the right of him towards the end of the video. Or listen to the 12:50 - 13:20. It is an open case to give the card the best possible temperatures. That's why I said in a conventional closed case it could be much worse.

Nice catch,
well makes the 68% fan even more impressive for noise.
The only one I have seen that may look like a suspect 1080FE was the one used by Tom's Hardware, but strangely when all set to OC with fixed fan (shame they only did 100% but I guess looking at pure OC with offsets) it stabilised at around 1990MHz.
Cheers

 

[cageymaru]

Nice catch,
well makes the 68% fan even more impressive for noise.
The only one I have seen that may look like a suspect 1080FE was the one used by Tom's Hardware, but strangely when all set to OC with fixed fan (shame they only did 100% but I guess looking at pure OC with offsets) it stabilised at around 1990MHz.
Cheers

Well most of us here have never heard a reference card from Nvidia, AMD, ATi, 3DFX, Matrox, etc that ran at 80% fan speed and was quiet. I recycled the fans from a Corsair H100i GT onto my custom loop that I'm building, and they are driving me crazy at less than 50% speed. I only installed two of the 6 that I intended to. On a good note my temps dropped 20 degrees. But it sounds like a damn airplane. I bet some people would say they are quiet.