What I'm afraid of is that they are measuring IPC gain in some AVX/AVX2 workload, since Zen/Zen+ is limited there to 128-bit, instead of the full 256-bit that AVX(2) can offer. Zen2 is fixing that, however its rarely used by games, for example, so it wouldn't result in any gains there. Getting full AVX1/2 is still a noteworthy gain, but folding that into some generic IPC increase would be .. misleading, since its not a generic gain, but only in applications that use it specifically. What I'm really interested in is what all-core clocks it can get for a 24/7 OC under water/AIO. If thats in any way decent, I'm definitely interested in a 12-core variant. But I definitely want 4.7+ all core for that, since I can run my "aging" 14nm Intel 10-core on that today already.

nz3777:

BS unless it can do 6Ghz iam not buying it! Joking aside I was expecting something Huge from Amd zen2 but even with these ipc improvements I would still rather have a 9900k. Zen is way cheeper I give em that much less then half price.

Engineering sample, and don't know if that is base, or all-core boost. Lets say it is base, what exactly is not "huge" about that? I'm only pointing this out because it sounds like you are reading this as though its not an engineering sample as well as knowing what the all-core boost is to make your opinion

I’m going to be the coolest kid on the block with a starter 6 core processor. It’ll be since upgrade for my 1600x.

S V S:

AMD "leaks" have a serious history of vastly overstating actual performance. It is hard to believe this is accurate.

Yes, everything before Ryzen has quite a bit of marketing jargon to create more of a buzz...true, but the IPC improvements of Ryzen actually were better than they promised. __ I do hope more switch to AMD, we need better competition in PC stuff, as we can see the competition in the phone business is making sure just about all actors are trying to make the best phone possible or at least the performance keeps on going up 4.5Ghz and the extra IPC sounds nice. If it is 8 cores then no need to switch up for quite some time, because PS5 will unlikely run at 4.5Ghz anyhow.

The waiting game intensifies.

My primary computer is Broadwell 5930 at 4.6 GHz. I am not hurting at the moment but my next CPU purchase will probably be AMD. In the 486 days I called up AMD and asked if AMD's 486 was pin compatible. They let me speak to an engineer who explained the AMD 486 was not actually pin compatible with Intel's. Sadly I went with Intel that time, but the whole 486 platform was a disaster. I had a Young motherboard that never worked. I'll take a good 286 any day. I never used an 8080 but I did use an 8085. I was into mini computers and mainframes before that.

Clock-Speed isn't everything that makes a CPU better. Zen2 is a new architecture based on 7nm, plus it is patched against Spectre and it isn't affected by Meltdown nor Spoiler, while Intel CPUs still are. Even with Zen and Zen+ AMD already has the upper hand against intel, as in benchmarks I see them mostly at top in multi-core performance, not intel. Might be true that single core performance on intel is better than AMD up to some extent (as of right now). However, there again AMD has a fair price policy instead, thus you can keep your kidney! For those who are gaming-oriented who must have the better single-core performance advantage regardless of the price policy: Let's not forget that a few months ago a Zen2 prototype model was running head in head with an 9900K, while Zen2 was consuming half the power that the 9900K does, which you can safely assume that Zen 2 was very most likely not running on full potential at all. Anyway, having no high hopes or expectations isn't necessarily meant to be something bad. It may as well boost the "surprise" effect when the final product launches.

Aura89:

9900k doesn't have an IPC advantage over 2700x. Sure, it some areas it does, and some areas it doesn't, and some areas they are equal. This is called, equal, as nothing is ever "exactly the same" What the 9900k DOES have over the 2700x is clock 9900k = 3.5Ghz base, 5Ghz 2-core boost, 4.7Ghz 8-core boost and i believe a 4.4 All-Core Boost 2700x = 3.7Ghz base, 4.3ish 2-core boost, 4.2ish 8-core boost and 4.1ish all-core boost If they 2700x could match the boost frequencies of the 9900k, they would be very similar in performance, as again, their IPC is very similar. Yes, there would be areas where one would win out over the other, again, this is normal, nothing can do everything exactly the same being different architectures. Yes, there are likely some scenarios where one wins out significantly. Remember, GPUs do this as well. So lets say this rumor is true, 15% IPC with 4.5Ghz, like above stated, that'd be similar to 5.175Ghz Zen+, which would be similar to 5.175Ghz 9900k. We don't know if this is a base speed or not. Usually from what i have seen, rumors come out with base speeds, so that leads me to believe that, it's possible 4.5Ghz will be base. But lets say its not, lets go with worse case scenario, if this rumor is true, and the all-core boost is 4.5Ghz, remember, the 9900k is 4.4Ghz all-core boost, and doesn't have the advantage of a 15+ IPC improvement. Obviously, all of this is rumor, and we need to wait till its release, but please, stop spreading misinformation around, as your statement that the 9900k has 15% more IPC then 2700x, is purely wrong. 9900k has a FREQUENCY advantage over the 2700x, and IF Zen 2 were to even MATCH that frequency advantage with the rumors 15% IPC increase, it will not be "slightly slower than a 9900k" Even if you believe that the 9900k intel has a 5% IPC advantage over the 2700x, which i wouldn't say is untrue, in some circumstances, as again, it's IPC increase between architectures, especially competing companies architectures, is circumstantial, a 15% overall average IPC increase from Zen+ to Zen 2 with similar clock speeds to the 9900k would not be "slightly slower than a 9900k" Get your facts straight.

He did write 2700K 🙂 And I assume that's what he meant by IPC increase?

S V S:

AMD "leaks" have a serious history of vastly overstating actual performance. It is hard to believe this is accurate.

Google translate of the source HH linked to seems to indicate that it's a 15% improvement "overall" compared to Zen+. I think people are overreacting with theories of 15% IPC improvement on top of clock improvements. Just because the Eng Sample clocks at 4.5 GHz doesn't mean a whole pile. The silicon lottery swings both ways even with production improvements over time and improvements to drivers, etc. Retail chips could perform the exact same as this hypothetical ES. Edit: Zen+ not Zen2

Sooo, they will "actually" run 32 gb of DDR4 3200 @ its rated speeds now???

screwtech02:

Sooo, they will "actually" run 32 gb of DDR4 3200 @ its rated speeds now???

Well i have been running 32 gb of DDR4 3300Mhz Cas 14 Hynix ram on Zen+ 2700x for some time without a single problem so i would say they won't have any problems running 3200Mhz and beyond on Zen 2 😉 . https://i.imgur.com/C7S3BYC.png https://i.imgur.com/mrVqAZ8.png

Netherwind:

He did write 2700K 🙂 And I assume that's what he meant by IPC increase?

Doubtful. I mean sure if thats the case, but nothing of this article says anything about 2700k, it wouldnt relate to zen2, etc. If he meant 2700k, that is extremely confusing

screwtech02:

Sooo, they will "actually" run 32 gb of DDR4 3200 @ its rated speeds now???

They already do (zen/zen+)

This looks pretty sweet if true. I have been dying to build a new PC. Zen 2 is at the top of my list. Take my monies AMD

chispy:

Well i have been running 32 gb of DDR4 3300Mhz Cas 14 Hynix ram on Zen+ 2700x for some time without a single problem so i would say they won't have any problems running 3200Mhz and beyond on Zen 2 😉 . https://i.imgur.com/C7S3BYC.png https://i.imgur.com/mrVqAZ8.png

Think about at Threadripper with 35ns memory latency, that would be epyc 😀 The only drawback for AMD now is the high latency in games, unless you are gpubound. My Threadripper 1950x i about 67ns, and my 9900k is 37ns. 1950x is in my F@H crunching computer, so latency is not imortent at all. Can't wait for Threadripper gen 3

What's with this nonsense I'm seeing about 15% IPC putting AMD on par with Intel or still preferring Intel even if AMD's ahead? AMD is pretty close to Intel as far as IPC in real world performance goes as it is, right now, with Zen+. 15% over Zen+ would put Intel at the bottom of a garbage can. Was being bent over and violated since 2006 not enough for you guys? You should be creaming your pants over the thought of AMD possibly taking the performance crown back from those shitters at Intel.

H83:

15% IPC increase means AMD has a real winner with ZEN2! It also means Intel is going to lose their performance crown after a very lengthy rein...

Maybe Intel can bribe giant companies like Dell to exclusively sell their inferior products at higher prices than what AMD's stuff would be sold at. That should do the trick. Then when they lose a lawsuit resulting from it 2 decades later, it'll be a tiny fraction of what it gained them, not to mention the extremely dangerous and weak position it'll put AMD in. Then Intel can shit on their customers with bloated monopoly prices, and 1-3% real world IPC gains per generation that's essentially the same crap re-branded for all eternity. Then if any product is accidentally too good they can stop soldering the IHS to cripple OC potential, and follow it up by having a shill write a BS article spewing fake news about how Intel have to stop soldering or it'll cause micro cracks that are actually dangerous. Morons will believe it. Someone should let Intel know about this diabolical plan. Think of where the world would be at technologically if that didn't actually happen. Companies will continue doing highly illegal scumfuck tactics like that forever so long as there isn't some hardcore prison time given as a penalty, the chump change they pay in lawsuits is always calculated ahead of time as a part of business. I want to see Intel stockholders and CEOs behind bars for 30+ years after being fined ONE HUNDRED PERCENT of what they attained illegally, or sometimes they'll even take the prison time if it's an insane enough gain. There is no justice until that happens. A broke minimum wage worker can end up in jail for petty theft of something worth $1, but these scum sucking pigs that manipulate the economies of the world, and in this case technologically cripple the world, serve zero time for billions of dollars attained through illegal methods.

Neo Cyrus:

Then if any product is accidentally too good they can stop soldering the IHS to cripple OC potential, and follow it up by having a shill write a BS article spewing fake news about how Intel have to stop soldering or it'll cause micro cracks that are actually dangerous. Morons will believe it. Someone should let Intel know about this diabolical plan.

Microcracking is 100% a real thing - whether or not Intel could figure out a better way around it is another story but it's a well researched issue and multiple companies/industries have published papers on it.

Neo Cyrus:

Think of where the world would be at technologically if that didn't actually happen.

I'm not sure if we would be much further tbh. You have to realize most of the breakthroughs in computing don't come from the companies themselves but from academia. For example AMD's zen architecture - aside from MCM (which has been around long before Zen) doesn't really do anything too different than Intel does under the hood - they basically just took all the known advancements and shoved them in there. That's why manufacturing was such a big deal for these companies because everything else architecturally is pretty well known and similar between companies. It's also why you kind of see the same advancements being added to different vendors in tandem - for example Nvidia/AMD in GPUs often come out with similar features/technologies around the same time because they are both pulling the ideas from a common area (research at academic institutions). I don't really expect to see massive gains by either company in general IPC after this. It's going to be incremental for some time outside of specific instructions like AVX. The rest of your post I agree with. Intel has historically been a shitty company and the punishment for their anti-competitive behavior wasn't even close to offsetting the advantage they got from doing it. I can totally see them doing it again. What's a few billion for 80%+ marketshare of the computing industry?

Denial:

Microcracking is 100% a real thing - whether or not Intel could figure out a better way around it is another story but it's a well researched issue and multiple companies/industries have published papers on it.

We all know it's a real thing, but threatening enough for Intel to avoid soldering the IHS altogether, on everything, starting from the 3000 series... coincidentally right until they started losing market share to AMD? Hell no. I may not be a materials engineer, but I know enough to see Intel BSing. As far how much we'd be ahead, who knows. But if Intel had to compete with ANYTHING they'd have been forced to innovated so many years ago that we can only imagine what new thing would be around. It may not have been mind blowing, but the raw performance we'd have at any given price would definitely be ahead of the "here's 4 cores at about the same frequency for the next 12 years + or -2% IPC" paradigm that Intel shoved up our asses with zero shame. Hell, maybe there'd even be some more research into that elusive reverse SMT. We're going to hit that multi threading wall eventually, and we're already near the physical limits of silicon. The mid and high end market for CPUs was a literal monopoly for so many years that it really was a worst case scenario as far as technological advancement went. Even towards the low end, AMD was rarely an option if OCing was taken into consideration. I think we're quick to forget what an abomination Bulldozer was. To this day it seems to me that whoever was pushing Bulldozer was either trying to sabotage the company, or completely incompetent.

D3M1G0D:

Yup. As far as I know, Intel still solders their Xeons. If microcracking was a serious issue, I would have expected them to have stopped using solder with their premium products - instead, it's only their consumer products where they use paste. This fact alone should tell you that it's a false flag, and that Intel is most likely using paste because it's cheaper, not because solder is dangerous.

Multiple xeons are not soldered and there can be a ton of different reasons why some are and some aren't. I highly doubt Intel is going to retool it's fab lines for TIM and re-validate the entire line to save several cents on TIM vs solder. They are doing it likely because its saving them money on x number on failed processors due to microcracking. I'm sure they ran some internal study where they said "by switching to TIM our temps go up 10c which cuts the life span by 2 years, outside the warranty period, but we save .5% of processors within the warranty period" and pulled the trigger on it.

Fox2232:

Well, what AMD vs. intel is doing differently is way they put it together. How they handle data transfers, how they handle cache accesses. And similar things.

Yeah but that's just a side-effect of the overall design. It's not like AMD is doing anything where Intel is like "wow how are they doing that?!?!" It's all well understood it's just different because the MCM approach is different. A future Intel MCM design at a high level will function similarly. I'm not saying these companies implement things identically but the stuff they do implement from a high level is well understood and documented long before its in an actual product. At a low level they may do it slightly differently because it benefits their overall design goal - or for example AMD who values security over performance puts a bunch of extra checks in it's speculative execution (spectre stuff).

Denial:

Multiple xeons are not soldered and there can be a ton of different reasons why some are and some aren't. I highly doubt Intel is going to retool it's fab lines for TIM and re-validate the entire line to save several cents on TIM vs solder. They are doing it likely because its saving them money on x number on failed processors due to microcracking. I'm sure they ran some internal study where they said "by switching to TIM our temps go up 10c which cuts the life span by 2 years, outside the warranty period, but we save .5% of processors within the warranty period" and pulled the trigger on it. Yeah but that's just a side-effect of the overall design. It's not like AMD is doing anything where Intel is like "wow how are they doing that?!?!" It's all well understood it's just different because the MCM approach is different. A future Intel MCM design at a high level will function similarly. I'm not saying these companies implement things identically but the stuff they do implement from a high level is well understood and documented long before its in an actual product. At a low level they may do it slightly differently because it benefits their overall design goal - or for example AMD who values security over performance puts a bunch of extra checks in it's speculative execution (spectre stuff).

Micro cracking, lol show me a 8 year old 2500/2600k even ones thats been overclocked to over 4GHz encounter micro cracking, the reality is by the time micro crackings occurs they are long gone as people would have moved onto something better. If a 8 year old 2600k overclocked isn't encountering it then clearly it not a issue for 99% of the users.

Dazz:

Micro cracking, lol show me a 8 year old 2500/2600k even ones thats been overclocked to over 4GHz encounter micro cracking, the reality is by the time micro crackings occurs they are long gone as people would have moved onto something better. If a 8 year old 2600k overclocked isn't encountering it then clearly it not a issue for 99% of the users.

Does the 2500/2600K experience a relatively high rate of voiding? Was the thermal density of the process/design of the chip as prone to failure as ivybridge? How many users of failed 2500/2600K would even know their failed processor failed due to microcracking without an electron microscope? How many batches of processors don't pass Q/A validation and end up in consumers hands to microcrack in the first place? Why did Xilinx, Intel and others publish research papers showing this as an issue? Why are their companies selling million dollar inspection and sorting systems design to Q/A these chips for voids if this isn't an issue? What sounds more far fetched "Intel wants to cut down on chip defects based on research done by multiple companies and internal studies so they retooled away from solder" or "Intel wants to stop .01% of their users from getting slightly better performance out of their processor but 75% of those .01 people are going to delid anyway but regardless they're going to spend millions on retooling/revalidating to prevent them from doing that and make up multiple research papers about it that just happen to coincide with other companies in the industry" I don't even understand why the latter is still an argument it's tinfoil hat levels of dumb. Also 1% of processors Intel ships (if it doesn't effect 99% of them) is millions of processors.