Farewell Zen, Hello Ryzen: AMD’s Eight-Core CPU Runs At 3.4GHz+

[atomicWAR]

@alidan

I agree with a lot of what you said. If AMD can deliver a 8c/16t part that is at least close to intel, I would happily trade in my i7 3930k's six cores in for eight. I would have sooner with Intel but 1G for 2 more cores that don't give me a whole lot more (even IPC wise) has been a tough sell. Assuming AMD keeps a reasonable price on their 8c/16t part and it delivers on the promise they are showing right now. I am sold.

 

[atomicWAR]

i could not agree more synphul. That's why I am waiting eagerly for launch reviews. AMD, Intel and everyone else in the silicon game love releasing cherry picked benchmarks that look great until launch when the subtleties of a chips performance become much clearer and we are left with a sour taste in mouths.

 

[TOMpel]

if these things still goin cheap as they used to, ima cancel my 1151 for it.

 

[rhysiam]

I'm leery of a single benchmark. Hopefully more will follow. If we adhered to a single benchmark then we could cherry pick a benchmark where the fx 9590 matches an i7 5960x and call it a win. Of course that doesn't come close to painting the full picture. If looking at the webxprt benchmark scores, an fx 9590 slightly edges out the i7 6900k. Again it means very little.

In single threaded cinebench r15 the outdated 4790k beats out the 6900k and in many other tasks it's almost tied. Going back to that webxprt bench, the 4790k embarrasses the 6900k. There are several bench's where the outdated desktop i7 holds its own or surpasses the 6900k in performance and that's without overclocking it. I don't think it's time to pop the cork on the champagne just yet until more data is made available.

I certainly agree that we need independent reviews before getting too excited, but to equate this to a singlethreaded test, or mish-mash of a benchmark like WebXPRT is not really fair. Blender (rendering) and Handbrake (encoding) are both opensource, CPU intensive programs, generating exactly the sorts of real-world workloads which motivate people to spend big dollars on a high end workstation in the first place. We certainly need independent tests, but AMD holding their own in these benchmarks is exciting.

 

[sarinaide]

The i7 4790 does not beat the i7 6900 in cinebench,. Please refrain from making things up

 

[rhysiam]

The i7 4790 does not beat the i7 6900 in cinebench,. Please refrain from making things up

To be fair, @synphul specifically said "single threaded cinebench R15" , where the 4790 does win by a solid margin by virtue of the higher clock speed and single threaded workload: http://www.anandtech.com/bench/CPU/1028

 

[ledhead11]

The real bottom line with AMD and I challenge anyone to provide the real numbers to disprove.

We welcome a hundred tests. 1/3 tests we pick and totally excel over anyone else. 1/3 of the tests the competitor excels. The final 1/3 are left to completely argue by fans or be ignored. I admit that current hardware/software favors team green and the results can be lopsided, but that's changing but doesn't change the lack of true innovation going into the hype for this side of tracks. It's truly AvP, no matter who wins we lose.

Please don't get me wrong, I really want AMD to do better. As a team green person I know that true competition will happen with relevant innovation. I was just telling a co-worker today about how I foresaw their stocks gaining even though they made the brilliant strategy of not being cutting edge with anything. This, however, is contradictory to that. Who care's if you made a Octa that's barely more powerful than a tablet from 2-3 years ago!

 

[blppt]

"by the time amd chips are out intel will already have faster chips out"

Which is why I thought Zen should have been released this year, but hey, they're the ones designing and producing the thing, maybe some bugs needed to be worked out with the support chipset or something.

 

[blppt]

"Back when the fx 9590 came out it was over $900 usd. It soon fell but only to $880. Given its overall performance, it's now a $200 cpu. Amd was a bit over enthusiastic with their pricing on that one. Hopefully this time around they're more thorough in testing and base their prices accordingly. It will be much less embarrassing if it performs like a $500 cpu and is priced around $480-500 than if they slap a $1200 price tag on it only to have to quickly slash prices due to lackluster performance. Every company makes mistakes, hopefully they learn from them. I think it would go a long way in saving face."

The real issue for me with my 9590 is the nuclear reactor it needs to run. 220W TDP for a cpu that often would trail my old 3770K 77W TDP cpu in games. Even at its enticing current price point, the required water cooling and massive power draw for a cpu that most of the time gets stomped by anything from a Devils Canyon to current Skylake i7 makes these cpus just about pointless.

 

[rhysiam]

...I admit that current hardware/software favors team green and the results can be lopsided...

You're talking about "Team Green", i.e. Nvidia? This is a CPU release. Nothing to do with Nvidia or the (AMD) Radeon Technologies Group which oversee the Graphics side of their business.

In the CPU department over the last 5 years AMD haven't really been able to offer a performance competitive mid-range or better CPU. Forget 1/3rd of benchmarks, outside of very specific workloads (7Zip, for exmaple), they haven't been competitive in really any benchmarks at all. Speaking generally, those wanting anything better than a entry level PC have only had Intel to choose from. With this lack of competition, we've seen CPU performance almost completely stagnate.

This is big news because it shows (in a couple of benchmarks at least), an AMD CPU going toe to toe with a $1K Intel CPU. We need to wait to see how representative this is of Ryzen performance more generally. But compelling CPU options from AMD, undercutting Intel on price, would push Intel to actually compete again.

 

[bramahon]

I'm not the one for wishful thinking but it seems AMD has finally "risen" from its slumber! Competition rings music in the ear of an enthusiast and the absence of it in last half a decade meant many lost the thrill and eventually the attraction. Having options and alternatives is a wonderful thing, I remember the thrill of building in the Deneb/Thuban days - not class-leading but great alternatives to C2Q/Nehelam. Unfortunately, Phenom/K12 was already a mature core and AMD couldn't keep up the good works requisite to stretch further the competition. Zen/Ryzen is in its 1st iteration and needs not to suffer from the same eventualities.

Anyway, thanks to Chris for covering the event and While I'm sure he is waiting ravenously to rip through that cache subsystem in benchmarks along with Igor and co., it's been a good teaser!

 

[bit_user]

Finally more SATA Express support, just want I need for those 0 SATA Express drives available.

Given a choice between standard SATA ports and SATA Express, the latter seems like a no-brainer.

And SATA still makes sense for RAID, in case you need more space or reliability than a single drive can offer.

 

[falchard]

I found this very promising considering the only intensive thing I use my CPU for is rendering in Blender. The FX Processor currently in my machine is more GPU bottle-necked in everything else.

 

[bit_user]

finally! hooray AMD! give intel some competition for the love of god.

i'll still buy intel, and hopefully at a cheaper price.

Of course, if everyone does that, there's no way they can stay in the game.

BTW, I'm pretty sure Intel hasn't lowed prices, in a long time. When faced with a threat by AMD, they tend to release SKUs that offer more features & performance, within their existing price structure.

 

[cats_Paw]

I dont want to be too negative on zen but please remember we are kinda hyping it up a bit.
Lets wait and see.

 

[sarinaide]

I remember when a FX 8150 Zambezi debuted and put up a 2:20 in Handbrake while the 2600 put up like 1:50 or something, Zen puts up 54 seconds to a i7 6900K's 59 seconds, but of course it will get painted over as a "cherry top" without looking at how radical that is. Intel CPU's have always faired well in Handbrake.

AMD seems to have radically fixed that Cache sub system and IMC to the point that it was Bulldozers biggest weakness to being one of strength, to transcode that well deserves at least some credit instead of dismissing it. FP ops also looks to be night and day improvement over anything Bulldozer was ever capable of.

The gaming benches were said to feel the same, you can't fake feeling the same on a frostbite engine at 4K with similar graphics cards on DX11.

 

[rmpumper]

It would be nice if AMD could bring down Intel's price gouging, but I won't get hyped by a single hand picked benchmark by the notoriously deceitful AMD's PR team. Let's wait for some third party tests and at least preliminary price points for these CPUs fist before over hyping yet another product.

 

[sarinaide]

Handbrake is a third party tool lol

 

[apache_lives]

As a some what Intel fanboy i really do sincerely hope this things is competitive, Intel needs competition and i want cheap mainstream 8 core CPUs already

 

[joshyboy82]

When the VERY FIRST WORD IS MISSPELLED. It's not all caps, don't go crazy, unless your brain hits a stumble-block the INSTANT you try to read the first sentence.

 

[CaedenV]

As exciting as this is, there are a few things I am still quite worried about that AMD needs to contend with.

1) AMD is a 'cheap' brand. The way things stand now, there is just no way for them to sell a chip that costs more than $400. The problem here obviously is that CPUs are so advanced now, that to put out something for the high end gaming or pro-sumer market this chip is going to have to be in the $500-1000 range. While the AMD faithful won't have an issue with this, the mass-market mindshare won't even look at anything other than Intel in this price range.

2) AMD is a confusing brand. Sure, Intel's lineup is not exactly straight forward (really with the LGA2011 parts were an i9 instead of i7 part to differentiate the class of chips), but almost anyone can tell you that the class of chips goes from celeron and pentium on the junk side of things, and then i3/5/7 on the mainstream to high end. AMDs branding is just a mystery looking from the outside-in. I mean, they have an APU lineup... and nobody really knows what an APU is still after years of AMD trying to explain it. Then there are the FX chips that are all mashed together. They need this new gen of parts to have clear performance levels that are understood with clear name conventions. When shopping, people need to know if their FX part is equivalent to an i3 or an i7 without needing to memorize tables and comparisons.

3) AMD has confusing motherboards. Say what you will about Intel requiring a new mobo every 2 years, but it makes things so much easier when building a rig! Chip x needs socket y, and it will have all of the latest technology available to the CPU. None of this silly 'well, I bought a new chip, but found out the board I was looking at was a 5 year old model and will technically work, but give up on technology a, b and c'. No more of that! 2 years is a perfectly acceptable amount of time to come out with a new chipset for new boards that supports new chips and technologies. Stop supporting legacy crap and confusing the unwashed masses of people learning to build rigs. Sure it is nice if you know what you are doing to just replace your CPU, or to save some money by buying an older board... but those are money savings that AMD cannot afford to give, and savings that can more easily be had by people who shop on the used parts market.

4) AMD has confusing code names!!!! Intel and nVidia have code names that are easy to remember, easy to report on, and easy to pronounce. With rare exception, AMDs code names (and product names) are just work. Who cares about a bunch of islands that nobody has ever heard of? Give us names that are easy to say, easy to remember, and are asperational in some way to give the brand something to desire (outside of performance).

Not saying that they will be putting out a bad part. In fact, I just might build a new rig next year with one of these new zen/ryzen (see! why not stick with zen!?!?! is it really that hard?) chips on board. But if AMD wants to sell to the mass market, they need what they have never had: a marketing team that can be trusted to make a good long-term brand that appeals to human beings rather than engineers. No matter how good their parts are, the audience will be very limited without a proper marketing department.

 

[WFang]

Was it just me that expected them to officially pronounce it "Risen" (or at least Rizen!) as in risen from the grave? That wouldn't be a big stretch from Ryzen.

What I'm a bit annoyed with, and that seems to have completely fallen off the radar, is that back in August/September they were talking 'enthusiast models' of Zen hitting the market late December/early January, now all they talk about (for awhile) is "Q1 2017". Also, when can we expect the gag to come off from Motherboard makers?

While I enjoyed the live demo stream yesterday, I'm also a bit worried about how it doesn't even look like an 'early' Q1 availability for enthusiasts because you would think they would drop some hard dates at this point if they knew... Instead they talk about possibly higher base clocks and still working on how much 'boost' will be available... Guys... we're a couple of weeks away from the start of Q1 2017...

All of that taken together, I'm starting to think we won't have reviews until around February and good market availability until end of Q1. 🙁

 

[tashfeenmajid]

You left the most important part:

Ryzen was matching/beating Intel's flagship at ONLY 95 WATTS!!!

 

[bit_user]

1) AMD is a 'cheap' brand. The way things stand now, there is just no way for them to sell a chip that costs more than $400. The problem here obviously is that CPUs are so advanced now, that to put out something for the high end gaming or pro-sumer market this chip is going to have to be in the $500-1000 range. While the AMD faithful won't have an issue with this, the mass-market mindshare won't even look at anything other than Intel in this price range.

I'm not sure about that. It will be priced competitively, and anyone who really cares about things like CPU brand or model # will just track down some benchmarks and buy the best option in their price range.

But, in the mainstream you talk about, a lot of people don't care about CPU brand.

 

[WFang]

After reading both articles, the one from here and the one from Anandtech, I'm wondering if this "extended frequency range" is a way AMD is will try to limit overclocking and exactly how it will effect overclocking.

Don't take my hobby away now....

Quote from another article on Ars:
"That said, fully manual overclocking is still supported for power users." (article link)