AMD Ryzen

[Ganrich]

 

Intel would have to remove their IGPs in order to make room for more cores on their chips. They can't just slap more cores on an architecture without removing things. They could have had separate skews for apus and cpus a long time ago, but that wouldnt let them price gouge people nearly as well.

 

Intel has chips without IGPs.  Some of the Xeon server chips are pure CPUs, no GPU bolted on.  There are Xeon E3 variants that are more or less the exact same chip as the same generation of i3/i5/i7, minus the IGP.

Higher prices, more cores, and low clock speeds tend to accompany those Xeons. They aren't mainstream processors. Intel will have to increase core count while maintaining clock speed and be close in price, or they could pull their strong arm tactics (or something similar) from the early 2000s. These need to be included in their mainstream line, or a new line to replace that line. I can't remember what their current socket is... 1151? That's where they need to have this stuff. They've avoided it to gouge the mainstream, and only offer more variety in the server cpu line because it isn't even close to needed and AMD hasn't released anything there in forever.

 

Either way, in a month I plan to build a Ryzen machine. I'd preorder, but Deadfire lowered my bank account a little too much to build a computer this week.

[AwesomeOcelot]

My 5820K isn't a Xeon and doesn't have a iGPU, it had a reasonable price/performance for a 6 core 12 thread CPU 3 1/2 years before Ryzen. Problem is Intel hasn't progressed much from there.

[Zoraptor]

The current Broadwell E varieties don't have an iGPU either. But like that 5820k was they're also comparatively expensive and decidedly workstation/ top tier enthusiast orientated.

 

I've always found Intel's semi obsession with integrated graphics to be strange, especially so on anything not aimed at corporate McBoxes or laptops. Suppose they're useful if your main graphics card packs up or if you want to buy a dedicated GPU later, but I don't imagine that fits into Intel's thoughts too much.

[mkreku]

Time to rein in that hype a little:

 

http://wccftech.com/amd-ryzen-7-1700x-review-leak-gaming-overclock-benchmarks/

 

Ryzen 1700x beaten by Intel in almost all gaming scenarios (although they use 2133 MHz RAM on the Ryzen vs 3200 MHz on Intel because early hardware lacked XMP profile support).

[Keyrock]

(although they use 2133 MHz RAM on the Ryzen vs 3200 MHz on Intel because early hardware lacked XMP profile support).

Not just significantly slower RAM, but also higher latency.  The RAM used for the Ryzen setup is 2133 MHz CL15 while the RAM used in the Intel setup is 3200 MHz CL14.  Usually higher speed RAM has that advantage offset somewhat by having higher CAS latency, the fact that the RAM used in the Intel variant is both significantly higher speed AND slightly lower latency gives it a fairly significant advantage.  I'll wait for benchmarks done on a truly even playingfield.

[Zoraptor]

Plus engineering sample MB and CPU, which means they may differ significantly from release models. Or may not, no real way to tell until 12 hours time or whatever. Since some of the slides have Farsi on them I'd guess they're older ES, but that's a guess.

 

I really wouldn't be expecting an R7 to beat a 7700k in many current games and whatever the set up, though it probably will beat it for most future games that take advantage of more cores. The R5s ought to be better than an R7 for most current games as well.

[Bartimaeus]

 

(although they use 2133 MHz RAM on the Ryzen vs 3200 MHz on Intel because early hardware lacked XMP profile support).

Not just significantly slower RAM, but also higher latency.  The RAM used for the Ryzen setup is 2133 MHz CL15 while the RAM used in the Intel setup is 3200 MHz CL14.  Usually higher speed RAM has that advantage offset somewhat by having higher CAS latency, the fact that the RAM used in the Intel variant is both significantly higher speed AND slightly lower latency gives it a fairly significant advantage.  I'll wait for benchmarks done on a truly even playingfield.

 

 

...Why the hell wouldn't they just set both to the same settings so it was on an even playing ground, ESPECIALLY when they know it's a temporary issue that the AMD board can't use the higher speeds... That's just silly.

[Sarex]

Guess they talked to Intel before reviewing. :D

[teknoman2]

reviews have started popping up on youtube

[Sarex]

Similar power for half the money.

[teknoman2]

one thing most have in common is that results from these early versions are a bit luck based. one may get a very smooth testing experience, another may have to jump through some hoops to get the job done

[Zoraptor]

Yeah, benchmarks are all over the place. 50% variance over what seems to be the same test from two different sites seems... extreme, and presumably (presumably) the higher value is more indicative of what can be expected with stable/ optimised hardware. Some benchmarks even put the 1800x worse than Vishera which seems highly unlikely.

 

The R5 should be far better value for money in games anyway (still won't beat a 7700k often though). For productivity it looks very competitive vs the 6900K so far as price performance goes since it has close performance at less than half the 6900k's price. Plenty of ammunition for both Intel and AMD fanboys there.

[teknoman2]

most of the time it seems to be a motherboard issue though

[Sarex]

Even with all of that, the Ryzen is a no brainer.

[Keyrock]

Overall, Ryzen definitely is a massive step in the right direction for AMD and puts them right back in the game in the enthusiast desktop sector, a place they've been pretty much completely absent from for well over half a decade.  I'm happy to see Intel again have someone to compete with in that sector other than themselves.  That said, nothing out right now makes me interested in upgrading my current i7-4790K/Titan X (Maxwell) setup, neither on the CPU, nor on the GPU front.  When Zen2 and Coffee Lake come out, likely in the latter half of 2018, I'll see what the situation looks like then.  By then Volta and Navi GPUs should also be out.  As with any brand new architecture, the second iteration tends to be much better than the first, not just because of a refined manufacturing process and getting the kinks out on the hardware side, but also because compilers will have had time to get more optimizations put in (which would also benefit the first gen).  I'm expecting big things from Zen2.  Hopefully Intel steps up their game to match also.

Edited March 12, 2017 by Keyrock

[teknoman2]

one complaint about the reviews that i've seen raised is that the testing methodology is all about absolute numerical values and make no effort to consider a realistic application of the hardware. the test method is not representative of the use people will have for a $2000+ PC like those used in the tests. content creation benchmarks aside, if someone made a pc like that for gaming, he would certainly not use it with a 1080p screen but with a 1440p (possibly even dual) or 4k. while the reason the testing is done like that is valid for the purpose of testing the raw numbers, it disregards the practical side of things

[Zoraptor]

I would have thought that both 1080p (and some did 720p) and maybe 4k would be fairer, with some acknowledgement if there were differences in min frames etc. The whole benchmarking thing has been a bit inconsistent though, since most reviewers got ASUS MBs that were not ready for primetime. Not really ASUS' fault as it seems the release was rushed by AMD and they've suffered a bit for it, but fact is that reviewers with Gigabyte MBs had a better experience but they were the minority. Add to that the whole windows scheduler not playing nice with AMD's core set up and the RAM speeds being a bit of a crap shoot due to the MB/ BIOS issues and there's plenty of potential for improvement there even before Ryzen 2.0.

 

I'm still pretty much definitely doing a Ryzen system since I still think it fits my needs best- primarily having better longevity and price performance than an equivalent Intel. I'll see if I can wait for the R5s and especially Vega/ 5x0s to see how they turn out and since they must be close (heh), but if I can't I'll be pretty happy with a 1700 and maybe pick up a cheap interim card 2nd hand since anything has to be better than a 5770.

 

(I've been checking hardware for a potential build and most stuff is fine, but looking at cases... it must be like how old people feel when they go into a cafe and have to do a questionnaire to get a plain coffee instead a non skinny non chai latte free trade Ethopian medium roasted in a counter clockwise rotating kiln coarse grind double shot- I just want a case with good cooling and expansion thanks, rather than endless irrelevant LED and case window variants from a dozen different makers which each seem to have their own model number. At least most sites have some sort of sensible sorting options...)

[Zoraptor]

R5 release April 11.

 

 

1600X looks pretty good value, 100USD less than the 1700 and should be (here at least) about the same price as the 7600k, only with 3 times the threads, a decent stock cooler and no need (well, 'need' at least) for a premier motherboard.

Edited March 15, 2017 by Zoraptor

[Azdeus]

I'm a bit torn between the 1600x and the 1700x myself, I'll be having to look at some benchmarks when that time comes.

[teknoman2]

a problem has been found that explains the gaming performance of Ryzen so far. the chip is divided in 2 nodes with 4 cores each and there is a significant latency in the communication between the nodes. the problem seems also to relate to the inability of current motherboards to accept fast RAM, so with more mature motherboards it may disappear.

i also heard that AMD has come up with a software solution that they handed over to microsoft to implement on a windows update.

i don't know if that solution was part of yesterday's update for win10 but if it was, it seems to also affect the older FX series because after installing said update i gained a significant boost in performance in games that were very heavy on my 8350

[Zoraptor]

The chip structure is not really a problem per se, it's pretty much an inevitable consequence of how they are manufactured, and communication within a 'node' (ccx) will be faster than between different ones. It is able to be improved by better scheduling from OS and/or per program basis, there certainly seems to be an issue where it is detected as having 16 physical cores rather than 8/8. Though AMD has rather equivocated on whether it is/ isn't working as intended it certainly isn't working optimally; the equivocation may be to stop rabid fanboys annoying MS to 'fix' their scheduler to take Ryzen into account.

 

Base RAM speed determines the speed of the inter core communication so yeah, going from 2400 to 3200Mhz will make a huge difference. If you were unlucky enough to have a MB/ RAM combo that didn't like each other or wanted to use 4 sticks of RAM then there will be a large performance increase at some point when the MB/ BIOS revisions are released. Most of the ASRock and Gigabyte MBs seem to already be capable of getting decent quality RAM to 3200Mhz (in 2 stick set ups, 4 are still at 2400 max so far as I know), the ASUS and MSI ones were lagging.

Edited March 18, 2017 by Zoraptor

[Azdeus]

I liked the Eurogamer/Digital Foundry article on Ryzen 7, there apparently are some real headscratchers with the chip aswell.

 

Also this could be promising, might be worth picking up a Ryzen 5 later on.

 

 

If there is an issue about how gaming workloads are distributed across cores, the most interesting evidence we uncovered concerned gaming performance results when we used the MSI BIOS to disable one core in each CCX, giving games six cores and 12 threads to work with. This is the '3+3' configuration that's being deployed on the Ryzen 5 1600/1600X, arriving next month. The 1600X actually has access to the same amount of cache memory as the 1800X and runs at the same base and boost frequencies. The results of the BIOS tweak are quite remarkable. Shorn of two cores and four threads, games only lose between three to seven per cent of the performance of the fully enabled eight-core chip. Assuming this is indeed representative of the upcoming Ryzen 5 1600X's turnout (and this can only be confirmed with actual hands-on time with the product itself), AMD may well have a highly compelling couple of mainstream products waiting in the wings that could make you think twice about a prospective Core i5 or indeed a Ryzen 7 purchase - but that's a discussion for another time.

[Sarex]

http://www.guru3d.com/news-story/amd-x390-and-x399-chipsets-diagrams-reveal-hedt-information.html

[Ganrich]

Ran across this vid on AMD's Reddit page. It is interesting, and adds to what Zorapter was saying about Ryzen and base ram speeds. I'm dragging my feet for bios and windows updates. All I really need to update soon is my GPU (waiting on Vega). Everything else is fine for a good while longer.

 

[Zoraptor]

Shame that fast RAM is so expensive (and here pretty unavailable, I can't source any 3200MHz CL14 from reputable sources at all) and anything 3200+ isn't officially supported, and it's still a bit random as to whether it will actually work at rated frequency. Being faster than a 7700k at games, albeit marginally overall and dependent on the game, is one of the last tick box Ryzen needs to fill.