AMD Zen to Bring Greater than 40% IPC Increase, New Reports Suggests

[Prysin]

Never had an LGa775 8 phase motherboard then? They use the 8+1 phase configuration with the phases in series.

who cares for an outdated socket that has been phased out, what? 5-6 generations ago?

[The_Strict_Nein]

Anyone here thinking that they'll be grabbing an 8 Core Zen with an IPC comparable to Intel for less than an Intel 6 core is seriously deluded.

 

AMD will take a smaller profit than Intel, because they have to to sell any. But the advantage Intel have is that they have room to lower their profit margin. AMD are already pretty much as low as they can go.

[Dabombinable]

who cares for an outdated socket that has been phased out, what? 5-6 generations ago?

Point is, back then you got motherboards that used either used 2x 4+1, or those that used 8+1 for the VRM phases. The same still applies now (though the second digit denotes the amount of memory phases).

[shawno]

Anyone here thinking that they'll be grabbing an 8 Core Zen with an IPC comparable to Intel for less than an Intel 6 core is seriously deluded.

AMD will take a smaller profit than Intel, because they have to to sell any. But the advantage Intel have is that they have room to lower their profit margin. AMD are already pretty much as low as they can go.

I take it you have the msrp there then?

[patrickjp93]

My bad didn't know Skylake E would have that.

Skylake E won't be coming until 2017 though, but AMD Zen will be coming at the end of 2016

AMD will have about a one quarter advantage over Intel.

However AMD has also priced stuff quite aggressively so I'm sure that either Intel will have to suffer market loss to AMD or price even more aggressively, which is good for us either way .

Considering Broadwell E is only a month or so away, I wouldn't abandon the idea that Skylake-E launches right around the time Kaby Lake does.

[patrickjp93]

I take it you have the msrp there then?

I take it you forget AMD launched the FX 9590 for $1000?

[shawno]

I take it you forget AMD launched the FX 9590 for $1000?

So previous launches will dictate msrp for all future products then?

Titan z launched at 3000.00 on debut before dropping, does that mean every titan launch will be 3000? Definitley not in the case of the titan x but that only had one gpu in it.

[patrickjp93]

So previous launches will dictate msrp for all future products then?

Titan z launched at 3000.00 on debut before dropping, does that mean every titan launch will be 3000? Definitley not in the case of the titan x but that only had one gpu in it.

It's simple economics. AMD needs money. Do you really think Intel's prices are just borne out of thin air and whims? Prices are set to maximize profits, not have the highest sales. If AMD has the quality to go toe to toe with Intel, it will price along very similar lines, because AMD doesn't even have to do market research to understand that is very near the absolute optimal pricing structure.

 

Every Dual Titan will launch for 2500+. Of that you have my personal guarantee.

[shawno]

It's simple economics. AMD needs money. Do you really think Intel's prices are just borne out of thin air and whims? Prices are set to maximize profits, not have the highest sales. If AMD has the quality to go toe to toe with Intel, it will price along very similar lines, because AMD doesn't even have to do market research to understand that is very near the absolute optimal pricing structure.

Every Dual Titan will launch for 2500+. Of that you have my personal guarantee.

Intels 1k offering is x99. If zen is meant to compete with haswell and skylake making it twice the price would be absurd.

I personally hope lisa su has more business sense then that.

2500 is 500 shy still. Does that mean you concur with my statement?

[patrickjp93]

Intels 1k offering is x99. If zen is meant to compete with haswell and skylake making it twice the price would be absurd.

I personally hope lisa su has more business sense then that.

2500 is 500 shy still. Does that mean you concur with my statement?

Zen is going to compete across the board, not just against the Z, H, Q, and B platforms. You're not getting an 8-core Zen for $400.

 

No, because I don't yet know how Nvidia will react to the value of the dollar increasing so much.

[shawno]

Zen is going to compete across the board, not just against the Z, H, Q, and B platforms. You're not getting an 8-core Zen for $400.

No, because I don't yet know how Nvidia will react to the value of the dollar increasing so much.

Cant argue with that logic.

Ill fence sit a bit more before i speculate price. I dont think they would release it at 1k unless it deserved that price of course.

As always time will tell.

[Trixanity]

Cant argue with that logic.

Ill fence sit a bit more before i speculate price. I dont think they would release it at 1k unless it deserved that price of course.

As always time will tell.

Summit Ridge Zen processors with 8 cores will, as previously mentioned, compete with the X99 platform. It will be priced accordingly. It's an enthusiast chip. A 4790k or 6700k is not in the same ballpark, as many seem to assume.

(obviously if the chip proves a failure, it'll drop in price to reflect that)

Intel has a single 8 core SKU at $999. So AMD will price it according to that. Don't wanna pay that much? Fine but then you don't get 8 cores and you probably don't need it either. Pricing it even at $500 would be stupid. You have nowhere else to go, so why make your product look like the cheap option?

AMD has clearly stated that they don't want to be the cheap alternative to Intel. They want to be at performance and price parity with Intel. Now the former is still up in the air but the latter will most likely happen. At best you'll see AMD pricing just below to give an incentive to go for AMD over Intel.

What AMD really needs are lots of design wins to become more ubiquitous and to become a household name in both consumer and business circles.

While enthusiasts such as those of the LTT forums are important customers, it's the bulk orders from OEMs and other business customers that will be the most important, particularly in filling their coffers.

[Prysin]

Summit Ridge Zen processors with 8 cores will, as previously mentioned, compete with the X99 platform. It will be priced accordingly. It's an enthusiast chip. A 4790k or 6700k is not in the same ballpark, as many seem to assume.

(obviously if the chip proves a failure, it'll drop in price to reflect that)

Intel has a single 8 core SKU at $999. So AMD will price it according to that. Don't wanna pay that much? Fine but then you don't get 8 cores and you probably don't need it either. Pricing it even at $500 would be stupid. You have nowhere else to go, so why make your product look like the cheap option?

AMD has clearly stated that they don't want to be the cheap alternative to Intel. They want to be at performance and price parity with Intel. Now the former is still up in the air but the latter will most likely happen. At best you'll see AMD pricing just below to give an incentive to go for AMD over Intel.

What AMD really needs are lots of design wins to become more ubiquitous and to become a household name in both consumer and business circles.

While enthusiasts such as those of the LTT forums are important customers, it's the bulk orders from OEMs and other business customers that will be the most important, particularly in filling their coffers.

Yh. The most important market is the OEMs that sold your dad that dull looking box with "very powerful cpu and a strong graphics card"

[The_Strict_Nein]

Intels 1k offering is x99. If zen is meant to compete with haswell and skylake making it twice the price would be absurd.

I personally hope lisa su has more business sense then that.

2500 is 500 shy still. Does that mean you concur with my statement?

 

Business sense doesn't mean shit if you're not covering R&D costs.

 

I don't think enough people understand this. Let's say developing Zen cost AMD X Dollars. AMD need to figure out how many they expect to sell. So the cost of the chip will be base cost to manufacture + (X Dollars/divided number of chips expected to sell) + a little on top as profit. This is why you get price cuts in the future when they've now covered the cost of R&D and they can start selling for just base cost + little bit on top.

 

EDIT: Considering how far behind AMD currently are and how much of an improvement Zen claims to be I cannot imagine the R&D costs of Zen being cheap.

[DocSwag]

Summit Ridge Zen processors with 8 cores will, as previously mentioned, compete with the X99 platform. It will be priced accordingly. It's an enthusiast chip. A 4790k or 6700k is not in the same ballpark, as many seem to assume.

(obviously if the chip proves a failure, it'll drop in price to reflect that)

Intel has a single 8 core SKU at $999. So AMD will price it according to that. Don't wanna pay that much? Fine but then you don't get 8 cores and you probably don't need it either. Pricing it even at $500 would be stupid. You have nowhere else to go, so why make your product look like the cheap option?

AMD has clearly stated that they don't want to be the cheap alternative to Intel. They want to be at performance and price parity with Intel. Now the former is still up in the air but the latter will most likely happen. At best you'll see AMD pricing just below to give an incentive to go for AMD over Intel.

What AMD really needs are lots of design wins to become more ubiquitous and to become a household name in both consumer and business circles.

While enthusiasts such as those of the LTT forums are important customers, it's the bulk orders from OEMs and other business customers that will be the most important, particularly in filling their coffers.

I've heard they already have secured quite a few deals with computer manufacturing brands though. A few weeks back Lisa Su said that AMD has already made deals with some companies to put Zen CPUs in their upcoming products. Zen has gotta be pretty good if companies are already securing deals with AMD.

[ewan_stockwell]

hi llt community. if anyone has the equipment and time to do this please do: could someone benchmark a fx-8350 and i5-6600k with only one core enabled and underclock both processors to 3.5GHz. this would allow us to see relative ipc (instructions per clock) and see if a 40% increase in performance will make the new zen architecture cpu's performance competitive with the skylake cpu's. please do a variety of benchmarks for variety!edit: a further breakdown of single threaded performance http://www.cpu-world.com/Compare/441/AMD_FX-Series_FX-8350_vs_Intel_Core_i5_i5-6600K.html#bench 

[ewan_stockwell]

i'm not a huge expert on cpu's and i know that benchmaks don't necessarily reflect real world performance but i quick google search shows the current single threaded performance of the i5-6600k and fx-8350 http://www.cpu-world.com/Compare/441/AMD_FX-Series_FX-8350_vs_Intel_Core_i5_i5-6600K.html . this shows that even with a 40%+ increase the new zen cpu's will have, at best, similar performance to the current skylake cpu's. for amd to shift units they will need a vary competitive price as they wont have the performance to make people upgrade from their current systems. again, i know these are only benchmarks 

[ewan_stockwell]

well, ...for me personally what i would want to see from those 99$ AM4 motherboard would be a solid 6 or 8 phase VRM with beefy heatsinks and dual GPU support @ pcie 3.0 x8/x8 (ideally x16/16)

all the rest i don't care...my sata3 ssd is fast enough but yes of course full overclocking options supported on those boards including memory, cpu bus and what not.

it's okay for AMD to have a slightly slower part (let's say within 10% from current intel) but give me a fully featured board for 100$.

i wouldn't get your hopes up for a $99 new motherboard with dual GPU support

[Trixanity]

hi llt community. if anyone has the equipment and time to do this please do: could someone benchmark a fx-8350 and i5-6600k with only one core enabled and underclock both processors to 3.5GHz. this would allow us to see relative ipc (instructions per clock) and see if a 40% increase in performance will make the new zen architecture cpu's performance competitive with the skylake cpu's. please do a variety of benchmarks for variety!edit: a further breakdown of single threaded performance http://www.cpu-world.com/Compare/441/AMD_FX-Series_FX-8350_vs_Intel_Core_i5_i5-6600K.html#bench

8350 is Piledriver. According to a previous user's math, you'd have to add 64% (if I recall correctly). Even then, I don't think you can translate the numbers directly like that.

Besides that, calculations have shown around Haswell IPC aka 4970K.

Regardless of all these calculations, even if absolutely spot on, I still think real world performance will look different (for better or worse). Maybe even very different.

Perhaps some of the experts can tell me how closely IPC and the improvements to it relate to actual real world performance. For example, how much of an influence was the supposedly 5-10% increase in IPC between Intel's generational products in real world performance.

To me, it has so far looked like it's very dependent on the application whether or not it has had any significant impact.

In any case, let's assume Skylake is a little faster and that Kabylake is even faster still. For most people the difference has not mattered enough. A lot of the improvements so far seem to be more about features and energy efficiency than performance. So it's not like a minor performance deficiency compared to Intel will compromise AMDs success too much except for the few cases where you need the most performance possible no matter the cost.

Despite all this, let's just wait and see. Until AMD shows off their products, all we can do is speculate and calculate. Maybe AMD has something up their sleeve that will skew the market in their favor, although I think that's wishful thinking on my part. Only things I could think of are already known or rumored.

[i_build_nanosuits]

hi llt community. if anyone has the equipment and time to do this please do: could someone benchmark a fx-8350 and i5-6600k with only one core enabled and underclock both processors to 3.5GHz. this would allow us to see relative ipc (instructions per clock) and see if a 40% increase in performance will make the new zen architecture cpu's performance competitive with the skylake cpu's. please do a variety of benchmarks for variety!edit: a further breakdown of single threaded performance http://www.cpu-world.com/Compare/441/AMD_FX-Series_FX-8350_vs_Intel_Core_i5_i5-6600K.html#bench 

i'm not a huge expert on cpu's and i know that benchmaks don't necessarily reflect real world performance but i quick google search shows the current single threaded performance of the i5-6600k and fx-8350 http://www.cpu-world.com/Compare/441/AMD_FX-Series_FX-8350_vs_Intel_Core_i5_i5-6600K.html . this shows that even with a 40%+ increase the new zen cpu's will have, at best, similar performance to the current skylake cpu's. for amd to shift units they will need a vary competitive price as they wont have the performance to make people upgrade from their current systems. again, i know these are only benchmarks

well...first it's 40% over excavator...not piledriver which the AMD FX-8350 vishera part is based on.

secondo, it doesn't work like that you can't compare IPC you can't quantify IPC this will vary a lot based on the type of instruction set the CPU will have to deal with for example the AMD FX-8350 since you're talking about it: this is a 4 modules part, each module equipped by 2 integer units, one 256bits Floating point unit that can double as a dual 128bit floating point, and 1 instruction decoder per module) therefore, this CPU if you trow integer based operations at it, it can process 8 integer based instructions per clock cycle, at the same time you can feed it with eight 128bit single-precision floating point based operations...so it can technically process up to 16 instruction sets per cycle, in which scenario it perform rather well for it's price.

BUT, you can also have a workload based mostly on floating point calculations, and 256bit double-precision floating point at that in which case the CPU can only process 4 instructions per clock-cycle making it perform rather poorly.

[ewan_stockwell]

my, bad wrong comparison i forgot about bulldozer  and i'm aware that this is only napkin maths, not anything definitive. the point i was (poorly) making was that a 40% increase in ipc wont necessarily make the new architecture higher performance than intels chips

[Humbug]

my, bad wrong comparison i forgot about bulldozer  and i'm aware that this is only napkin maths, not anything definitive. the point i was (poorly) making was that a 40% increase in ipc wont necessarily make the new architecture higher performance than intels chips

You are getting further confused.

Bulldozer is slower IPC than Piledriver. (FX-8350 is piledriver)

Piledriver is slower IPC than Steamroller.

Steamroller is slower IPC than Excavator.

 

Excavator is 40% slower IPC than Zen. Zen is 40% faster IPC than Excavator

 

[Trixanity]

my, bad wrong comparison i forgot about bulldozer and i'm aware that this is only napkin maths, not anything definitive. the point i was (poorly) making was that a 40% increase in ipc wont necessarily make the new architecture higher performance than intels chips

I believe that you're correct in assuming that it's unlikely for AMD to deliver a CPU that bests Intel, particularly in every benchmark.

However, I think for most people the performance is good enough. Even Sandy bridge is good enough for the majority.

It's the features such as DDR4, M.2 PCIE SSD, PCIE 3.0, USB 3.1 (A and C) that are important. Even those are only a selection of features necessary for a good chip.

There's also the APU with a built-in Polaris GPU, perhaps even a high end SKU with HBM (probably 4 GB onboard). Such products would be beating anything Intel can produce. That would mean a potential solid foothold in both laptops and desktops. Filling a spot no one else can do.

Note that performance is absolutely important and I do believe both Intel and AMD should strive for big improvements, so far it just seems like we've hit a snag where improvements are small and therefore become less important which I guess is a good thing for AMD. Imagine if Intel was doing 20-30% year over year. Then AMD wouldn't be able to keep up (although if I'm not mistaken Zen+ is supposedly gonna be 10-15% higher IPC than Zen).

[ace_cheaply]

You are getting further confused.

Bulldozer is slower IPC than Piledriver. (FX-8350 is piledriver)

Piledriver is slower IPC than Steamroller.

Steamroller is slower IPC than Excavator.

Excavator is 40% slower IPC than Zen.

No, Zen is 40% improvement over Excavator. It's been said at least a dozen times in this thread alone. Excavator is around 25% faster ipc than bulldozer, making Zen around 65% faster clock for clock with bulldozer. This brings it theoretically either neck and neck with, or slightly faster than Haskell ipc. However, the New architecture changes, and clock speeds need to be known before we can say what real world perfmance might rival.

[DocSwag]

my, bad wrong comparison i forgot about bulldozer  and i'm aware that this is only napkin maths, not anything definitive. the point i was (poorly) making was that a 40% increase in ipc wont necessarily make the new architecture higher performance than intels chips

Also remember this article doesn't say 40%, it says MORE than 40%, so this means it will be the same as if not better than Skylake. Also keep in mind AMD has generally been better at hitting higher clock speeds on CPUs which means even higher performance. They developed a 4ghz octa core 125 watt CPU on 32 nm (FX 8350). Granted higher IPC could mean more heat but this still holds weight, so AMD could match Intel in IPC while having higher clocked CPUs.