Zen hype getting over the hill now

[Stereodude]

For those of us who know, i3s are terrible. Alt-Tabbing is excruciating, so is multitasking on an i3. Those with HT don't fair much better but it does help alleviate just a bit of the issues.

Thanks for the morning laugh.

 

[a3venom]

You know how we can tell you are limited in your knowledge in processors, because you live by i3. For those of us who know, i3s are terrible. Alt-Tabbing is excruciating, so is multitasking on an i3. Those with HT don't fair much better but it does help alleviate just a bit of the issues. Benchmarks never tell the whole story, they only play the game. They don't have the other programs most of us, excluding you and the other i3 owners, have running whilst we play our games. I play all my games at 75Fps, with 95% holding hard to 75 Fps. Not one time has my 8350 ever hindered my game play. I have been ripping my DVD collection the last few days to take them on my vacation, and with all 8 cores at 95-99% I am typing this now, played small games and watched Ntflix at the same time without a hitch.

So maybe you should refrain from posting in a thread that you have no knowledge in nor interest other than to flame.

But you agree that a 8 core with all its power juiced up, liquid cooling and shit and the most expensive ram money can buy - cannot match a dual core in FPS? Never said you want to RIP DvDs and encode videos in background, just playing games. Don't deflect the argument, embrace the truth that i3 is much faster in games - and being faster in games is really good marketing for a CPU, even though that is far from their sole purpose.

You AMD fans are supporting the worst decision AMD ever made, and would gladly take a 10 core 10 T 25% IPC increase over a 4 core + HT 50% IPC increase - if ZEN has to choose between the two. lulz

 

[JustReason]

But you agree that a 8 core with all its power juiced up, liquid cooling and shit and the most expensive ram money can buy - cannot match a dual core in FPS? Never said you want to RIP DvDs and encode videos in background, just playing games. Don't deflect the argument, embrace the truth that i3 is much faster in games - and being faster in games is really good marketing for a CPU, even though that is far from their sole purpose.

You AMD fans are supporting the worst decision AMD ever made, and would gladly take a 10 core 10 T 25% IPC increase over a 4 core + HT 50% IPC increase - if ZEN has to choose between the two. lulz

No. Never said an i3 will always get better fps, actually the majority of the time it wont. Min and Avg don't tell the whole story, we need frame time graphs over a period of time to determine that. Add the OCing to the discussion and it isn't the same game.

No person with reason will ever push an i3 as a gamer CPU, definitely not today. Now, given a person without preference, I would suggest an i5 or i7 (k variants) over AMD counterparts. But if the question is whether an 8350 or APUs are worth it or capable of gaming then definitely yes.

 

[drescherjm]

No person with reason will ever push an i3 as a gamer CPU, definitely not today.

And yet we see a lot of pushing here of APUs that are not as powerful (CPU wise) as an i3.

 

[a3venom]

No. Never said an i3 will always get better fps, actually the majority of the time it wont. Min and Avg don't tell the whole story, we need frame time graphs over a period of time to determine that. Add the OCing to the discussion and it isn't the same game.

No person with reason will ever push an i3 as a gamer CPU, definitely not today. Now, given a person without preference, I would suggest an i5 or i7 (k variants) over AMD counterparts. But if the question is whether an 8350 or APUs are worth it or capable of gaming then definitely yes.

Explain how a 7870k APU will be the best option to someone who has only $150 to spend on a CPU and $40 on MOBO. You cannot explain a merit of the APU in this case except that it takes up less space and power than a dGPU + CPU from the same company. APUs are borderline retarded for this reason, unless you buy some of the cheapest ones, which are again lulz.

i3 vs FX = this, DX 12 as of today, the gap is even bigger, a stronger hyperthreaded dual core will be better than an ancient octa core.

 

[pxc]

His basic assumption is incorrect. AMD never said Zen was 40% faster per thread. AMD stated of Zen: "40% more instructions per clock". On a processor with SMT, that leaves a lot of leeway for where that 40% number comes from. Anyways, AMD never claimed 40% was across a wide range of workloads, so picking Cinebench and multiplying by 1.4 is beyond wishful thinking and well into fantasy.

And the other problem:

Has Zen even taped out yet? The last time AMD mentioned it, it still hadn't been.

 

[Pieter3dnow]

His basic assumption is incorrect. AMD never said Zen was 40% faster per thread. AMD stated of Zen: "40% more instructions per clock". On a processor with SMT, that leaves a lot of leeway for where that 40% number comes from. Anyways, AMD never claimed 40% was across a wide range of workloads, so picking Cinebench and multiplying by 1.4 is beyond wishful thinking and well into fantasy.
And the other problem:
Has Zen even taped out yet? The last time AMD mentioned it, it still hadn't been.

Not only his base math is wrong but also the rest of it. But he tries to make a point rather then proving Zen is that fast in single thread you can get "near" (yes not entirely true) but the factors of gigahertz and processor design means that even tho you have an idea does not mean it will follow your logic ...

The New Citavia Blog: AMD Zen and K12 (ARM) tapeouts confirmed by LinkedIn profile

 

[razor1]

major caveat there, if that was true I remember when that linked in post was up, that guy would have gotten canned or sued or both. So.......

 

[pxc]

The New Citavia Blog: AMD Zen and K12 (ARM) tapeouts confirmed by LinkedIn profile

There were several rumors that Zen taped out last year, but all were invalidated when AMD officially said it still hadn't taped out in December and would tape out in the next several months. From the December 9, 2015 conference call transcript:

"Zen was a clean sheet design that started few years ago. We are in the final figure of executing and the milestone that you want hear us talk about is Zen taping out, which should be over the next several months. And then putting samples in the hands of our customer and then starting portfolio of revenue in 2017."

It's very possible it still hasn't taped out yet 3 months later. AMD, or legitimate sources inside it, would be leaking info about that milestone if it passed. The silence is deafening.

 

[razor1]

Well in all consideration it probably did tape out a while back, CPU validation cycles are much larger than GPU's, so I would expect it to be taped out by now, hopefully.

 

[Pieter3dnow]

There were several rumors that Zen taped out last year, but all were invalidated when AMD officially said it still hadn't taped out in December and would tape out in the next several months. From the December 9, 2015 conference call transcript:It's very possible it still hasn't taped out yet 3 months later. AMD, or legitimate sources inside it, would be leaking info about that milestone if it passed. The silence is deafening.

When talking about server parts right. I recall this because someone else posted this as well...

 

[Darakian]

New rumor. New date. October
AMD Zen 8 Core 95W Summit Ridge CPUs Launching In October - Already Going Through Validation

 

[Spirit_Retro]

Well in all consideration it probably did tape out a while back, CPU validation cycles are much larger than GPU's, so I would expect it to be taped out by now, hopefully.

It's taped out already or they are out of business.

My suspicion (based only on having been in the tech industry since the beginning of time) is that they are keeping things quiet for one of two reasons:

1. Zen is a massive win, performs far better than anticipated.

2. Zen significantly sucks.

If it was a middle of the road result they would be promoting it already.

 

[Spirit_Retro]

i3 vs FX = this, DX 12 as of today, the gap is even bigger, a stronger hyperthreaded dual core will be better than an ancient octa core.

The lack of real world practical knowledge is keeping young Padawan chained to ignorance.

 

[gamerk2]

Explain how a 7870k APU will be the best option to someone who has only $150 to spend on a CPU and $40 on MOBO. You cannot explain a merit of the APU in this case except that it takes up less space and power than a dGPU + CPU from the same company. APUs are borderline retarded for this reason, unless you buy some of the cheapest ones, which are again lulz.

i3 vs FX = this, DX 12 as of today, the gap is even bigger, a stronger hyperthreaded dual core will be better than an ancient octa core.

It's worth noting the CPU results are compressed due to GPU bottlenecks, so Intel's actual lead is larger then those charts indicate.

Granted, I wouldn't recommend an i3 for a gaming rig, due to latency concerns, but in terms of maximum possible throughput, i3 > FX 8xxx.

 

[razor1]

It's taped out already or they are out of business.

My suspicion (based only on having been in the tech industry since the beginning of time) is that they are keeping things quiet for one of two reasons:

1. Zen is a massive win, performs far better than anticipated.

2. Zen significantly sucks.

If it was a middle of the road result they would be promoting it already.

Taped out doesn't mean they have final silicon in hand, they might get final silicon soon though for an end of year launch.

 

[a3venom]

The lack of real world practical knowledge is keeping young Padawan chained to ignorance.

FPS numbers > Real world analogies.
Just found out what's a padawan today lol

 

[Pieter3dnow]

New rumor. New date. October
AMD Zen 8 Core 95W Summit Ridge CPUs Launching In October - Already Going Through Validation

I have seen the base rumour for this on SA forums and that is all it is the post was made somewhere in January. Wouldn't be the first time "news" gets recycled

/g/ - Technology » Thread #52492812

 

[Humpernator]

I don't know what to think.

 

[a3venom]

I was thinking - since Zen will have DDR4 and (say) Ivybridge IPC. Wouldn't that make it faster in gaming than Haswell + DDR3? Because going from 2133 to 3200 Mhz in DDR4 makes a significant difference in perf numbers, specially on slower CPUs in Skylake as seen on digitalfoundry - so this could easily be faster than Haswell in a lot of titles.
Any thoughts?

 

[Darakian]

I have seen the base rumour for this on SA forums and that is all it is the post was made somewhere in January. Wouldn't be the first time "news" gets recycled

/g/ - Technology » Thread #52492812

Dang. Thought things were speeding up.

 

[Dan_D]

I was thinking - since Zen will have DDR4 and (say) Ivybridge IPC. Wouldn't that make it faster in gaming than Haswell + DDR3? Because going from 2133 to 3200 Mhz in DDR4 makes a significant difference in perf numbers, specially on slower CPUs in Skylake as seen on digitalfoundry - so this could easily be faster than Haswell in a lot of titles.
Any thoughts?

You can't really compare Skylake to Haswell and assume DDR4 is the reason for that improvement. Haswell-E also has DDR4 memory support. Skylake experiences more performance gains in games with memory clock increases than Haswell-E does compared to Haswell. This tells me that the secret is in Skylake's architecture and not simply due to DDR4 support. In other words, we can't make architectural assumptions about how Zen will behave when coupled with DDR4 and compare that to Intel CPUs. There may be specific applications where Zen excels compared to Intel CPUs. At this juncture, everything is pure speculation.

More specifically, I think it's foolish to think that Zen is automatically going to be better than Haswell at gaming just because it has DDR4 support. Zen very well could be that good, but I wouldn't hold my breath on that one. If it is that good, I don't think it will be because of DDR4 support alone.

 

[Spirit_Retro]

I don't know what to think.

It's all speculation at this point. But since I don't actually *need* a new proc, I'm waiting on jumping ship to Intel. If I needed a new proc I'd already have switched. I just want to give AMD a chance to come out swinging because competition is good, and if they fail there will not be any.

Then we'll all be kinda screwed.

 

[Dan_D]

Then we'll all be kinda screwed.

AMD isn't really competitive anymore and hasn't been for years. At this point Intel competes more with itself than they do AMD. Intel has to try and get people to buy new processors, chipsets and whatever else they sell when the old stuff is still working just fine. At this point it's often hard to justify upgrades as the improvements are relatively small for the high cost of doing them. All the people who held onto / still have their Core i5/i7 2500/2600K CPUs are good examples of this. Those CPUs are ancient in terms of technology, but they still get the job done. Even Skylake isn't a massive increase in gaming performance. Intel wanted to win over gamers and get them to spend money which is why Skylake's advertising was aimed at gamers and enthusiasts. Intel spouted things like "designed for overclocking" and "gaming" when Skylake was revealed to the market in full.

Since the introduction of the Core 2 Duo, Intel has been very aggressive with its price points. While the HEDT segment still commands relatively high prices across the upper end of the range, the entry level CPU is priced quite reasonably. In fact, the 5820k overlaps the Core i7 6700K. Other offerings in the mainstream product line have a broad price range. Intel's kept it's highest end mainstream part within the $300-$400 range for years now.

Intel hasn't seriously been threatened by AMD in more than 10 years time. In terms of revenue, they never have been. AMD has been less and less competitive since K8, and yet Intel's processor prices have spread out into more brackets and we have more to choose from than we ever have before. AMD could die tomorrow and I don't think a single processor would have a price shift. Ever since the sub-$1,000 PC came about, the market hasn't been able to support really expensive mainstream CPUs. I'd wager the average price of a PC is probably sub-$600 outside of enthusiast circles. 15 years ago when AMD was at it's most competitive, you couldn't get PC's that cheap with Intel CPUs in them.

I'm not saying that competition isn't needed or that AMD has no market impact, but again Intel already competes with themselves more than they do AMD.

 

[Spirit_Retro]

AMD isn't really competitive anymore and hasn't been for years. At this point Intel competes more with itself than they do AMD. Intel has to try and get people to buy new processors, chipsets and whatever else they sell when the old stuff is still working just fine. At this point it's often hard to justify upgrades as the improvements are relatively small for the high cost of doing them. All the people who held onto / still have their Core i5/i7 2500/2600K CPUs are good examples of this. Those CPUs are ancient in terms of technology, but they still get the job done. Even Skylake isn't a massive increase in gaming performance. Intel wanted to win over gamers and get them to spend money which is why Skylake's advertising was aimed at gamers and enthusiasts. Intel spouted things like "designed for overclocking" and "gaming" when Skylake was revealed to the market in full.

Since the introduction of the Core 2 Duo, Intel has been very aggressive with its price points. While the HEDT segment still commands relatively high prices across the upper end of the range, the entry level CPU is priced quite reasonably. In fact, the 5820k overlaps the Core i7 6700K. Other offerings in the mainstream product line have a broad price range. Intel's kept it's highest end mainstream part within the $300-$400 range for years now.

Intel hasn't seriously been threatened by AMD in more than 10 years time. In terms of revenue, they never have been. AMD has been less and less competitive since K8, and yet Intel's processor prices have spread out into more brackets and we have more to choose from than we ever have before. AMD could die tomorrow and I don't think a single processor would have a price shift. Ever since the sub-$1,000 PC came about, the market hasn't been able to support really expensive mainstream CPUs. I'd wager the average price of a PC is probably sub-$600 outside of enthusiast circles. 15 years ago when AMD was at it's most competitive, you couldn't get PC's that cheap with Intel CPUs in them.

I'm not saying that competition isn't needed or that AMD has no market impact, but again Intel already competes with themselves more than they do AMD.

I disagree. From my standpoint we'd be farther along with even faster processors if Intel had competition.

In any industry, when competition fails, consumers lose. We're losing right now. Most of us think we're winning. But most of us are wrong.

 

[Zarathustra[H]]

You're a flamer!

I kid, but the article is nothing but blind speculation. We'll all see what AMD has to offer in a few months time. Until then zen is Schrodinger's core.

Yeah, I think we all learned to take pre-market buzz with a truckload of salt after the Bulldozer launch fiasco. (Or at least we should have)

I'll start getting excited when I see official benchmarks from trusted sites.

 

[Chimpee]

Yeah, I think we all learned to take pre-market buzz with a truckload of salt after the Bulldozer launch fiasco. (Or at least we should have)

I'll start getting excited when I see official benchmarks from trusted sites.

Yea, it is better to have low expectations and be surprised by the results then having high expectations and only to be disappointed with AMD.

 

[Dan_D]

I disagree. From my standpoint we'd be farther along with even faster processors if Intel had competition.

In any industry, when competition fails, consumers lose. We're losing right now. Most of us think we're winning. But most of us are wrong.

Making a faster processor is not the problem. The problem is that Intel's priorities have shifted over the last decade or so. At one point we had rapid progress, not merely because of competition with other vendors, but because more performance was desirable above all else. Increasing transistor count, processor size, and general performance was relatively easy. Things slowed down when performance per watt became the goal. Performance per watt is the most important thing right now because that's a big deal in dense server farms and datacenters where power consumption and cooling the equipment become major concerns. In the mobile market, performance per watt translates to battery life. Those two markets have an aligned set of goals despite the gap in usage scenarios and workloads.

Lowering power consumption while increasing performance is much harder than simply making something faster. The desktop market is a secondary concern. This is why we essentially get repurposed mobile and server chips. The hardware we get is generally sufficient for the work loads we have. A focus on pure performance would be foolish for Intel or AMD at this point.

 

[gamerk2]

Making a faster processor is not the problem. The problem is that Intel's priorities have shifted over the last decade or so. At one point we had rapid progress, not merely because of competition with other vendors, but because more performance was desirable above all else. Increasing transistor count, processor size, and general performance was relatively easy. Things slowed down when performance per watt became the goal. Performance per watt is the most important thing right now because that's a big deal in dense server farms and datacenters where power consumption and cooling the equipment become major concerns. In the mobile market, performance per watt translates to battery life. Those two markets have an aligned set of goals despite the gap in usage scenarios and workloads.

Lowering power consumption while increasing performance is much harder than simply making something faster. The desktop market is a secondary concern. This is why we essentially get repurposed mobile and server chips. The hardware we get is generally sufficient for the work loads we have. A focus on pure performance would be foolish for Intel or AMD at this point.

Pretty much this. The competition isn't for the fastest desktop chips anymore, it's about the highest performance, cost effective, low power mobile platforms. Intel isn't competing against AMD, they're competing against Qualcomm.

 

[Zarathustra[H]]

Making a faster processor is not the problem. The problem is that Intel's priorities have shifted over the last decade or so. At one point we had rapid progress, not merely because of competition with other vendors, but because more performance was desirable above all else. Increasing transistor count, processor size, and general performance was relatively easy. Things slowed down when performance per watt became the goal. Performance per watt is the most important thing right now because that's a big deal in dense server farms and datacenters where power consumption and cooling the equipment become major concerns. In the mobile market, performance per watt translates to battery life. Those two markets have an aligned set of goals despite the gap in usage scenarios and workloads.

Lowering power consumption while increasing performance is much harder than simply making something faster. The desktop market is a secondary concern. This is why we essentially get repurposed mobile and server chips. The hardware we get is generally sufficient for the work loads we have. A focus on pure performance would be foolish for Intel or AMD at this point.

Yep, I'd argue that it is a combination of the two.

1. Decreased competition from AMD, coupled with growing mobile, and shrinking desktop sales means Intel doesn't have to focus as much on competing for the desktop, and can shift its focus to mobile/server markets, using a unified design for all markets.
2. Server and Mobile markets demand performance per watt.
   
3. While there is some correlation between advances in performance per watt and max desktop performance in that if you are producing less heat you can up the voltage more and clock things up more, this isn't a perfect correlation. Optimizing for performance per watt can result in a clock speed ceiling, which we have seen more and more on newer Intel chips, and is why my older Sandy-E chip can hit higher clocks at max overclock than some newer Ivy and Haswell designs.

Even so, I'm not sure this is a huge deal. Where we are with desktop computing today, full out CPU performance really doesn't matter outside of bragging rights, except for a small subset of users who are into encoding/rendering and scientific compute type applications.

You certainly don't need a beast of a CPU for web/email/youtube, and if you play games, even a mid range Intel CPU is probably more than enough for all but the most demanding systems, at which point the added PCIe lanes for SLI are typically more interesting than the raw CPU power.

In today's CPU market you just need to be "good enough". Above that "good enough" level, for most applicatiosn it really doesn't matter if you get any faster.

For most things most people do (web/email/youtube) AMD's current designs are good enough. Even most games will run fast enough to be GPU limited on an faster AMD CPU these days. There are - however - some exceptions, where even the fastest AMD chip, overclocked as high as it will go, just won't provide the same experience as an Intel chip.

This is where Zen needs to achieve. It doesn't need to trade blows with Intel at the very top end, it just needs per core performance to keep up with a mid to high end 6 year old Sandy Bridge chip. If it can do that, it will be viable.

 

[razor1]

Well I agree with most of what you stated, but if AMD and Intel are fairly equal in all metrics, then they will sell well, if there are any negatives to either AMD or Intel then which ever has weakness will be scrutinized by marketing. So even though there is no need for the extra processor performance/ power usage ratio over its competition, there are always compelling reasons for the ability to market them.

Because Zen and Intel's core products will be on similar (still think Intel might have an advantage here from a cost perspective and possible tech lead, but much less than before), a price war will happen if the metrics of power consumption and performance are close.

AMD's recent problems, a large contribution to it is AMD's cutting into their margins, larger die size chips pitting against Intel's smaller chips because of a performance level. So if they are able to compete on a mm to performance ratio, their margins will increase, but marketshare probably won't, there needs to be more of reason to push OEM's and system builders to their products. Price war possible, but that will regress again by cutting into margins, which I don't think AMD would want to do, because Intel has deep enough pockets to sustain a price war.

 

[Zarathustra[H]]

Well I agree with most of what you stated, but if AMD and Intel are fairly equal in all metrics, then they will sell well, if there are any negatives to either AMD or Intel then which ever has weakness will be scrutinized by marketing. So even though there is no need for the extra processor performance/ power usage ratio over its competition, there are always compelling reasons for the ability to market them.

Because Zen and Intel's core products will be on similar (still think Intel might have an advantage here from a cost perspective and possible tech lead, but much less than before), a price war will happen if the metrics of power consumption and performance are close.

AMD's recent problems, a large contribution to it is AMD's cutting into their margins, larger die size chips pitting against Intel's smaller chips because of a performance level. So if they are able to compete on a mm to performance ratio, their margins will increase, but marketshare probably won't, there needs to be more of reason to push OEM's and system builders to their products. Price war possible, but that will regress again by cutting into margins, which I don't think AMD would want to do, because Intel has deep enough pockets to sustain a price war.

Yeah,

I can only speak for myself, but in the spirit of supporting competition, as long as I can get a per core speed upgrade over my i7-3930k I'd consider an AMD Zen upgrade.

I don't have a direct NEED for it, as the i7-3930k has provided to be the best high end CPU/Motherboard investment I've ever made, since it's been viable since 2011 and shows no need for an upgrade, but its starting to get old. Some of the caps on the motherboard are starting to get wiggly, and some of the palstic tabs and slot parts have started cracking from repeated re-installations. It runs a bit hot overclocked, and it doesn't have modern features like m.2 slots, or native NVME booting.

So, I don't really need any huge increase in performance (and even if I stuck with Intel I wouldn't get one) so as long as Zen is at least competitive, even if it doesn't take the performance crown, I'd consider the upgrade.

I just don't want to lose any PCIe lanes or cores, and I'd like at least a symbolic boost in per core performance.

The only bummer would be having to replace all my RAM with DDR4. I have 64GB in this machine.

 

[Pieter3dnow]

This is where Zen needs to achieve. It doesn't need to trade blows with Intel at the very top end, it just needs per core performance to keep up with a mid to high end 6 year old Sandy Bridge chip. If it can do that, it will be viable.

Really ?
It does not need to be better in IPC then Intel because these people keep yelling "IPC IPC IPC", so confused right now (sarcasm)?

From AMD side of things they need to get good yields and a very good line of motherboards. That should allow them to make a decent enough effort to try and get back permitted that they reach the performance threshold. From watching that video about Zen by Adoredtv they can push out a much better amount of cpu per wafer this alone would make things less bleak.

 

[razor1]

the problem with that video, IPC doesn't translate to complete performance, so what ever that video shows, doesn't show you what the performance might even be if IPC is 40% more. And its ridiculous to think the average computer buyer even knows what IPC is, I wouldn't be surprised the sales people at your local sam's or best buy probably don't even know what IPC is lol.

As I stated, margins will get better as long as there is no price war. If there is a price war, margins well guess what will be similar to now *Intel's will drop though*, unless there is balance for what Intel and AMD would want to go down to which is not exactly legal depending on how its done over course (that is called price fixing). If AMD wants to gain marketshare they need a compelling reason to push OEM's and system builders to their side. If that isn't performance its gotta be price (total system price, which lower power consumption/ performance ratio is important). Well what does that equate to?

 

[razor1]

Yeah,

I can only speak for myself, but in the spirit of supporting competition, as long as I can get a per core speed upgrade over my i7-3930k I'd consider an AMD Zen upgrade.

I don't have a direct NEED for it, as the i7-3930k has provided to be the best high end CPU/Motherboard investment I've ever made, since it's been viable since 2011 and shows no need for an upgrade, but its starting to get old. Some of the caps on the motherboard are starting to get wiggly, and some of the palstic tabs and slot parts have started cracking from repeated re-installations. It runs a bit hot overclocked, and it doesn't have modern features like m.2 slots, or native NVME booting.

So, I don't really need any huge increase in performance (and even if I stuck with Intel I wouldn't get one) so as long as Zen is at least competitive, even if it doesn't take the performance crown, I'd consider the upgrade.

I just don't want to lose any PCIe lanes or cores, and I'd like at least a symbolic boost in per core performance.

The only bummer would be having to replace all my RAM with DDR4. I have 64GB in this machine.

Yeah performance is always good to have, special for people that need it,

Also for low end to mid range system buyers for what ever they are doing, there really isn't a need to push performance than what it is right now, for OEM's power draw is very important as it does drop the cost on other parts of the system, I don't know the break down of cost between cpu/motherboard/ etc profit margins that AMD and Intel get, but with mid range and low end systems, I wouldn't be surprised to see more margins on the system than the CPU itself that AMD gets right now.

 

[Pieter3dnow]

the problem with that video, IPC doesn't translate to complete performance, so what ever that video shows, doesn't show you what the performance might even be if IPC is 40% more. And its ridiculous to think the average computer buyer even knows what IPC is, I wouldn't be surprised the sales people at your local sam's or best buy probably don't even know what IPC is lol.As I stated, margins will get better as long as there is no price war. If there is a price war, margins well guess what will be similar to now *Intel's will drop though*, unless there is balance for what Intel and AMD would want to go down to which is not exactly legal depending on how its done over course (that is called price fixing). If AMD wants to gain marketshare they need a compelling reason to push OEM's and system builders to their side. If that isn't performance its gotta be price (total system price, which lower power consumption/ performance ratio is important). Well what does that equate to?

Was using the yields/production as an example. The rest of the video did not interest me that much.

I don't care what it leads to as long as I can spend my money on a 8 (or more) core Zen. The way I see it the desktop market is lost already , for the reasons Zarathustra already listed.

 

[razor1]

good yields and good motherboards matter but not as much as performance/watt, they have to have a similar ratio to intel otherwise OEM's and system builders won't be making cheapo systems with AMD, they won't be able to get prices as low and Intel, over all with all other system parts (if AMD processors use more power) those parts have to be beefier. Still have hope that AMD can pull a rabbit out of its hat, but in reality yeah its hard to stomach.

I do think they can match performance per watt at least in the low end systems.

 

[defaultluser]

You can't really compare Skylake to Haswell and assume DDR4 is the reason for that improvement. Haswell-E also has DDR4 memory support. Skylake experiences more performance gains in games with memory clock increases than Haswell-E does compared to Haswell. This tells me that the secret is in Skylake's architecture and not simply due to DDR4 support. In other words, we can't make architectural assumptions about how Zen will behave when coupled with DDR4 and compare that to Intel CPUs. There may be specific applications where Zen excels compared to Intel CPUs. At this juncture, everything is pure speculation.

Precisely. My thoughts on why:

Extra memory bandwidth is wasted if the processor can't use it. Intel went to great lengths to improve I/O in Skylake, increasing buffer sizes, cache prefetch algorithms (so good they broke benchmarks!), load/store bandwidth. They also improved hyper threading performance, and added an one more execution unit.

Most of these game engines that show improved performance use 8 or more major threads, so improvements to hyperthreading mean you can service more of those threads simultaneously. In order to service those threads with spare execution units, the instructions and data have to be in cache. The improved prefetch makes that happen, but at the cost of increased memory bandwidth utilization.

The prefetch is key to everything else in the improved pipeline getting fed, and the reason why Skylake performance improvements vary from game-to-game. But more ram bandwidth really helps things maximize the effects of the new prefetcher!

 

[Zarathustra[H]]

Really ?
It does not need to be better in IPC then Intel because these people keep yelling "IPC IPC IPC", so confused right now (sarcasm)?

From AMD side of things they need to get good yields and a very good line of motherboards. That should allow them to make a decent enough effort to try and get back permitted that they reach the performance threshold. From watching that video about Zen by Adoredtv they can push out a much better amount of cpu per wafer this alone would make things less bleak.

I don't remember saying anything about IPC at all.

IPC is one half of the performance equation. Theoretically you can achieve absolute performance by any combination of IPC*Clock, but history has proven (both with Intel's Netburst and AMD's Bulldozer) that making up for low IPC with high clocks generally doesn't work out so well.

What I am saying is that absolute performance - for most users - just needs to pass a certain threshold. It doesn't need to be the absolute fastest as gamers are likely going to be GPU limited anyway, and non-gamers are primarily doing browser/email/YouTube type loads which don't take much CPU power.

There are fringe groups who do a lot of encoding/rendering that will take advantage of higher end chips, but for most other people they are a waste.

 

[Derfnofred]

A lot of my scientific stuff (COMSOL, primarily) relies heavily on memory bandwidth over CPU speed. In fact, I rarely see a sim use more than 60% processor load on my quad/hex core computers (depending on which workstation I'm on). The x58 970 stock is faster than my overclocked i7 860, primarily because of that 3rd memory channel.

Again, a beast of a system isn't solely relying on the CPU. Amdahl will always have an opinion, haha.