AMD Announces Ryzen 5 Processors; $169 Four-Cores And $219 Six-Cores, Available April 11

[JamesSneed]

The clocks are a bit disappointing on the the 4 core parts. But the 1600 and 1600X look like absolute winners in any multithreading value comparison. As far as gaming goes, how much of a performance hit will they really take dropping 2 cores + 2 threads compared to the 8 core Ryzens? Most games responded better to the higher clockspeed of the 7700k vs the increased core count on the 6900k. Assuming it can hit a 4 GHz all core OC, shouldn't it perform pretty close to the 1800X?

We will have to see some reviews to really know for sure. If the speculated crossing over CPU-Complex (CCX) is a major contributor to the lack of gaming performance a 4c(2+2 CCX) could take a bigger hit. I'm underwhelmed by these but at least AMD's pricing is solid.

 

[TJ Hooker]

Why do you people think that 6 Core Intel is hell of more expensive than same 6 core Intel with 28 PCIe? These things are not cheap to put in CPU.

The 6800k and 6850k use the same die (as do all Broadwell E CPUs, AFAIK). So the cost of putting PCIe lanes in there is the same for either CPU. The reduced lane count could be related to increasing yield when there are defects in the PCIe controller, but I'm guessing it's as much about product segmentation as anything. So chances are they're deliberately disabling some lanes on the 6800k and it's not because it costs more.

 

[captaincharisma]

all i see is AMD is going back to its old strategy with giving quantity over quality when it comes to how many cores they are giving these CPU's. time to see if the quality part where they were always lacking has improved

 

[gggplaya]

These CPUs will probably be slower than the I3 and I5 Kaby Lake CPUs when it comes to games, due to their IPC. I was hoping AMD would pull through this time. Hopefully their next CPU release will be better.

Right now, most modern titles are GPU limited, unless you're willing to shell out $800 for a 1080ti. People buying 1080ti's aren't shopping for midrange CPU's anyways. I play at 1440p and if I keep buying midrange gpu's, i'll continue to be gpu limited for several years.

The AM4 socket is supposed to be good till at least 2020. So I plan on buying a ryzen 5, engine the full frame rates I would have gotten with a core i5 anyways, due to gpu limits, but also have more cores for video editing and transcoding as well as streaming. Then in 2020, buy whatever is AMD's latest and simply drop it in.

 

[hannibal]

It is the same chip as with x1800 and other released Ryzen2 so the clocks Are about the same. It seems to be mostly up to focus to effiency (a good thing) and producing technology. Not much choise in there.
These Are excelent chips that Are not as fast in Gaming as Intel versions, but very near and that is good enough at this moment. They offer superior multi threading power compared to similar priced Intel CPU so They offer very good value for money.
The Ryzen2 will get a Little closer to Intel but after that it is a though competition in manufacturing technology and price. The speed will slow down a lot after Ryzen2 but I think that even now Ryzen is good choise!
Not the best in every situation though. Buy based on what you do with your Computer and that is just fine. The competition is back!

 

[davidchaoth]

3+3, 2+2 Alignments simply prove that:

1) The hero Jim Keller doesn't know the performance impact on using 2 or 1 CCX?
2) None of AMD senior engineer knows the performance impact on using 2 or 1 CCX?
3) None of AMD senior engineering managers knows the performance impact on using 2 or 1 CCX?
4) None of AMD engineering executives knows the performance impact on using 2 or 1 CCX?

Conclusion: AMD engineering executives are just for delivering presentations and holding chips for pictures!

Potentially even worse:

Why 3+3, 2+2? Can't yield on a complete CCX for 4+2, 4+0?

 

[Supahos]

Whole lot more likely they have a whole lot of CPUs that have defects in 1 or 2 cores and stick them in r5s to sell the silicon instead of trashing it. But spout your baseless insults if you wish.

 

[sushukka]

This year will be very refreshing in CPU/GPU markets. First Ryzen 7, now after only one month Ryzen 5 and appr. one month of that R3 and Vega announcements. Finally we have some competition and change to the normal tick-tock releases! Also I'm surprised how narrow minded people tend to be. First everybody are whining how expensive Intel and Nvidia offerings are and then when you finally have some competition which could change the situation, people are whining how the solution is lagging some percents in some tests and because of that piss on AMD. You can look Intel's and Nvidia's profits in the past comparing to AMD's and make a conclusion if those two giants are really taking all out of the monopoly they have.

Also we have only a small idea how much there are Intel optimized (or designed to fit Intel's architecture) software on the market and without any change there will be near zero possibility in the future to any competitor to show up and undermine that monopoly. It's a near miracle that AMD is performing this well with all new architecture and product lineup against Intel...and comparing the R&D budgets against Intel and Nvidia it very much is.

Personally I'm now waiting the Vega announcement and after that I'll build the "reddiest" pc there is. It probably don't have the ultimate best IPC, but it will have probably the best bang to the buck, it has good future upgardeability and will perform more than well in majority of the situations now and in the future. This is maybe little bit sentimental but the extra what I get is the feeling that I have supported AMD and haven't been forced to buy the monopoly crap from Intel and Nvidia like it has been over a decade now.

 

[Isokolon]

A 6C12T CPU for the price of an i5 is a massive statement as long as it isn't too far outperformed.

For everyone doing video editing or other core-orientated tasks Ryzen 1600(X) will be a blessing and will cost Intel some market share.
Hopefully they'll be able to compete with gaming and other single threaded tasks as well to lay some pressure on Intel.
I'd hate to see another 4c4t i5-k or another 4c8t i7 with marginally higher clocks refreshed every 8-12months.

One can only hope that AMD can manifest themselves as real competition again -- looks promising so far

 

[iPanda]

I really hope they figure out the gaming part... I would love to upgrade my i5-2500k but that would require new mobo and ram as well... if ryzen can figure out the performance issues, then i'd be great to save on that whole set up. then again, if i do plan on streaming... then I guess ryzen it is anyways. IIRC, weren't they behind over 30fps (no streaming) in some games? that is what would irk me (again, non-streaming scenario).

 

[Fulgurant]

As long as R3/R5 offer game performance similar to R7's, clock-for-clock, they will be an astoundingly good value, turning the entire low-to-mid-range market on its head. Certain internet posters may dismiss this news, but Intel won't.

And that's a very good thing for all of us. Welcome back AMD.

 

[InvalidError]

It'll be interesting to compare 2+2 Ryzen against APUs with a single quad-core CCX when those become available later assuming AMD doesn't do anything to sabotage the APUs' CPU performance. If the APUs end up outperforming similarly clocked 2+2 "enthusiast" CPUs in most cases as I expect them to, then we'll have proof beyond reasonable doubt that 2+2 was a bad idea performance-wise.

 

[rhysiam]

It'll be interesting to compare 2+2 Ryzen against APUs with a single quad-core CCX when those become available later assuming AMD doesn't do anything to sabotage the APUs' CPU performance. If the APUs end up outperforming similarly clocked 2+2 "enthusiast" CPUs in most cases as I expect them to, then we'll have proof beyond reasonable doubt that 2+2 was a bad idea performance-wise.

Well that's the big question isn't it. Many people in this thread and elsewhere seem fairly convinced that the dual CCX layout is the primary (only?) reason why Ryzen's IPC is so much lower in gaming compared to other CPU intensive tasks (in non-gaming tasks, Ryzen IPC seems similar to Broadwell - which is very decent).

Do people actually know that it is the additional latency and prefetch issues associated with the dual CCX layout that's causing the issues? That's just speculation at this stage isn't it? If we had a fully enabled single CCX CPU we could answer that very good question definitely, but I don't think we can right now. If you or others have sources which explore this I'd welcome a link. But as I understand it, this is just theorising/speculation at this stage is it not?

 

[InvalidError]

Do people actually know that it is the additional latency and prefetch issues associated with the dual CCX layout that's causing the issues?

It should be possible to find out which CPU core numbers are associated with threads on each CCX and disable all of those on a given CCX to simulate a CPU with all threads on only one CCX. The leftover cache from the ignored CCX may still be an issue though as any hits on that extra cache would incur the fabric latency hit.

 

[cryoburner]

Do people actually know that it is the additional latency and prefetch issues associated with the dual CCX layout that's causing the issues?

PC Perspective did some tests when they were investigating the rumors about incompatibility with the Windows thread scheduler, and found that the scheduler was behaving as designed, but also found that there's significantly greater latency when a core on one CCX communicates with one on another...
https://www.youtube.com/watch?v=6laL-_hiAK0

Jump to around the 9:00 mark if you want to skip to the part where they show a graph depicting the core to core latencies of an i7 5960X, and then compare those to an R7 1800X. In short, the core to core latencies of the 1800X within a single CCX were nearly twice as fast as those of the Intel chip, but the latency between cores on separate CCX were almost half as fast. It can take over three times as long for cores to communicate with one another if they're not on the same CCX, so if threads on separate CCX are in constant communication, presumably that could cause a hit to performance. However, while this is not necessarily a flaw in the way Windows was designed to assign threads to cores, it is something that could likely be addressed. If either Windows, or the games themselves, made sure to group frequently-communicating threads on the same CCX, this performance hit could likely be avoided, potentially making the issue "fixable".

And sure, it's possible there might be some other issue at play, but this seems like a plausible explanation for why Ryzen under-performs in many games, and perhaps also explains how it managed to turn the tables and beat Intel's chips by a similar margin in Deus Ex: Mankind Divided. If frequent core to core communications are staying within the same CCX, performance could actually be getting a boost over Intel's chips. And that could potentially speak well for Ryzen's future performance when games are optimized with it in mind.

Back on the topic of the article, this actually follows along fairly closely with the lineup and pricing information that was leaked over a month ago, only the prices are $10 lower, and some of the product names and clock rates are a bit different. The $219 R5 1600 could be a really good deal for a 6 core processor, particularly if it can be overclocked to similar levels as the R7 1700.

 

[calken]

I think everyone is getting their knickers in a twist over the CCX config. Everyone knows that multicore is the future and scalability is a fundamental design consideration. With AMD locking the requirement to use the infinity fabric, code will evolve to overcome their limitations which makes future CPU revisions automatically optimised for more core complexes.

 

[madmatt30]

Ryzen 5 1600X vs Intel i5 7500
Ryzen 5 1600 vs Intel i5 7500T
Ryzen 5 1500X vs Intel i5 7400
Ryzen 5 1400 vs Intel i5 7400T

Jeez i wonder which one going win lol despite the Cores and Threads

99% sure that 1600X can`t beat a Intel i5 7500.

well such a seriously fanboy comment & so far off the mark its not even funny.
a 1600 at $219 if its capable of hitting 3.8ghz (which I would fully expect) is going to be enough to make pretty much any locked i5 a complete waste of money.
2 extra physical cores,6 extra threads,very very similar ipc to a 7500 at 3.8ghz - there is absolutely nothing not to like about that.
The core configuration gaming performance (if it is indeed partly down to windows) will be sorted in double quick time imo - remember that although MS & AMD are hardly bedfellows theyre certainly well past the cuddling stage with the time & money invested in their console platform.

 

[OfficialG3]

more interested to being able to unlock the hidden cores, you know, like back in the phenom era ?

 

[gggplaya]

Ryzen 5 1600X vs Intel i5 7500
Ryzen 5 1600 vs Intel i5 7500T
Ryzen 5 1500X vs Intel i5 7400
Ryzen 5 1400 vs Intel i5 7400T

Jeez i wonder which one going win lol despite the Cores and Threads

99% sure that 1600X can`t beat a Intel i5 7500.

well such a seriously fanboy comment & so far off the mark its not even funny.
a 1600 at $219 if its capable of hitting 3.8ghz (which I would fully expect) is going to be enough to make pretty much any locked i5 a complete waste of money.
2 extra physical cores,6 extra threads,very very similar ipc to a 7500 at 3.8ghz - there is absolutely nothing not to like about that.
The core configuration gaming performance (if it is indeed partly down to windows) will be sorted in double quick time imo - remember that although MS & AMD are hardly bedfellows theyre certainly well past the cuddling stage with the time & money invested in their console platform.

Not to mention a much better factory cooler, saving even more money. Plus the B350 motherboard for $90 allows you to overclock. The savings are real.

 

[TMTOWTSAC]

more interested to being able to unlock the hidden cores, you know, like back in the phenom era ?

Seems doubtful as long as demand for the higher core count products are high. Anything that could validate as a higher binned part would likely be sold as such, rather than disabled and downclocked. Assuming they're actually full chips with cores disabled and not a cut down.

 

[InvalidError]

Seems doubtful as long as demand for the higher core count products are high.

If demand for higher-end parts was that high, then there would be insufficient parts left to sell as lower-end parts to launch the lower-end models. If the defect rate was so high that AMD didn't need to artificially cripple CPUs to meet demand for R5 parts, that would mean AMD has horrible yields.

In all likelihood, the bulk of R5 chips will be fully functional and artificially crippled as needed to meet demand, just like the Phenoms were.

 

[cryoburner]

2 extra physical cores,6 extra threads,very very similar ipc to a 7500 at 3.8ghz - there is absolutely nothing not to like about that.

Actually, that works out to 8 extra threads. The i5 7500 has 4 cores with 4 threads, while the R5 1600 will have 6 cores with 12 threads. Just thought I'd point that out. : P

 

[gggplaya]

Seems doubtful as long as demand for the higher core count products are high.

If demand for higher-end parts was that high, then there would be insufficient parts left to sell as lower-end parts to launch the lower-end models. If the defect rate was so high that AMD didn't need to artificially cripple CPUs to meet demand for R5 parts, that would mean AMD has horrible yields.

In all likelihood, the bulk of R5 chips will be fully functional and artificially crippled as needed to meet demand, just like the Phenoms were.

There's nothing wrong with bad yields as long as you have enough good yields to sustain your higher product line. Going smaller and smaller on lithography, I imagine yields will get worse. That's good news for the consumer because AMD can offer cheaper chips with broken cores. There's nothing wrong with that.

Actually I think it's bad for a company to have to artifically cripple cores to offer a cheaper product. Seems like a messed up way of doing business.

 

[InvalidError]

Actually I think it's bad for a company to have to artifically cripple cores to offer a cheaper product. Seems like a messed up way of doing business.

The L3 cache is bigger and more densely packed with wiring and transistors than the CPU cores. If AMD can offer 16MB of cache on the 1500, then it means that the defect rate in the L3 cache must be very low and in all likelihood, that the defect rate in the CPU cores themselves can only be even lower. If 5/6 of currently announced Ryzen SKUs based on the same 8C16T die have the full 16MB cache, then the vast majority of those should also have eight fully functional cores.

And artificially crippling products does make sense: it enables you to maximize margins on a product by selling practically the same product price points to suit your target markets' wallets. If AMD has an excess inventory of R7 chips, it can disable cores and pass those off as R5 along with the salvageable defects. The alternatives are an increasing supply of unsold R7 chips wasting space in warehouses while waiting to have enough defects to fill orders of R5 chips or lowering the price of R7 chips to keep inventory rolling. As the fabrication process improves, filling R5 orders using defects alone will become impossible assuming it isn't already.

 

[dE_logics]

Still stuck on Intel since this only runs on Windows.