There was a Ryzen 5 3600 review posted on reddit, it’s using an x470 motherboard and ddr4 3200. It scored 192 in cinebench single core and 3509 on the multi core. The i9 9900k got 204 single core ad 4835 in the multi core. Keep in mind the r5 3600 is 6 cores and 12 threads not like the 8 cores 16 threads of the i9 9900k. So this 4.2ghz single core turbo r5 3600 is on the heels of the 5.0ghz i9 9900k in cinebench single core.
Here is what’s interesting though, latency on the r5 3600 is 80ns (which is worse than first gen and second generation ryzen), and it still games way better than first/second gen ryzen because of the 16mb of L3 cache per ccx over the 8mb of L3 cache per ccx of Zen 1 & Zen+.
The i9 9900k wins the gaming bench marks but the r5 3600 with it’s 4.2ghz single core turbo still beats the r7 2700x with it’s 4.35ghz single core turbo by a reasonable amount. While it matched the i7 6700k (4.2ghz single core turbo) in FarCry 5.
Wouldn’t count it as a very good test. I mean, it’s ok to test against higher parts to see, but where the main point is is to test against similar parts.
1 - The BIOS used to perform the test is BETA and doesn’t fully support Ryzen 3000 chips and has memory support problems. For example memory write speeds are 50% slower than previous Ryzen chips.
2 - The Ryzen CPU had overclocked RAM. Ryzen scales with RAM very well while the Intel systems had no mention of OC whatsoever.
So yes, a $200 might destroy Intel’s lineup but this review doesn’t provide enough information for us to support their conclusion one way or another.
200 dollars isn’t low end, it’s mid range, and it doesn’t really match it, the 9900k still beats it.
For your avg. consumer comparing a 9900k to a 3600 means basically nothing, it’s not a great test. That, and their testing methodology is a little suspect, will wait until GN has results since they are pretty spot on with their tests.
Once reviewers really start to get their hands on it we’ll see tests at the same price point and if ryzen is beating out intel at that price point we’ll see what kind of money you have to drop on intel to get equivalent performance.
Yea, it’s going to be a wait and see. When people are looking for the best part for the money they have they want to see things for that amount, we know the 3600 will probably stomp the 9600k but the question is by how much. How much more performance am I getting out the 200 dollar part vs the 250 dollar part matters a lot at that price point.
Same thing with the 9900k vs whatever ryzen 9 pulls out, that’s what you want to see.
I will say this, i thought 7nm Ryzen was going to clock higher. Seem like they decided to go wider instead and bring 16 cores 32 threads to main stream.
Personally I feel like (yes this is an opinion and just my own opnion) if the r9 3950x can hit 4.7ghz single core boost; then i don’t seen why the r5 3600 can’t do 4.5ghz.
It’s about cannibalising sales of the higher-margin chips. I touched on it in one of the earlier threads on the Zen 2.
Basically it boils down to people wanting processors for (arguably) one of three reasons: cost, single-threaded performance (gamers), or multi-threaded performance (creators). Most everyone will naturally be limited by the former, of course. But anyone who wants either of the latter has no choice but to opt for a chip which has both, which introduces much higher profit margins for a larger market.
So while a 4-core, 8-thread (or 6/12 - it’s mostly academic at this point) chip which runs into the high 4GHz range or even low 5GHz would be a hugely desirable chip, it would take away sales of their higher-clocked, wider offerings. Since these narrow chips would have to be priced lower (they have less cores, so fewer/smaller parts and higher yields) there would be a large number of people opting for them over the ultra-wide variants and the bottom line would take a hit.
Don’t think it would happen? Cast your mind back, if you’re old enough, to the Pentium II era. A Celeron 333A could often reach 500MHz, giving it a situational performance lead over the top-end P-II 450 for substantially less money. Sales of the P-II gave way to the Celeron, taking profits with them; even if it was a crapshoot as to whether a given Celeron chip could hit those speeds or not.
_Ironically, the cynic in me believes AMD’s modular approach is a hidden way of avoiding the same situation. Nothing says you can’t OC a Ryzen 5 to the same speeds or more than a Ryzen 7. But if they can more easily cherry-pick their yields and deliberately use less desirable silicon on the lower chips, it removes a lot of that potential. Increase yields and profits in one go.
But that’s starting to get into conspiracy theory territory, so take with a liberal amount of salt._
The thing is aster, if you overclock the r5 3600 using what AMD uses for their turbo boost ramping per core charts; you should absolutely hit 4.5~4.6ghz. It would be 4.6ghz single core (2 because of the design). At that point though, why would you purchase an r7 3800x when all you do is gaming?
Hence why they don’t release it [the R5 3600] at those speeds out of the box. Not everyone OCs, nor does every chip necessarily OC all that well (my 1700 requires over 1.3v to hit 3.8GHz stable on all cores, yet only 1.225v for 3.7GHz), and many of those that do want to shoot for the highest speeds possible so they want to start with a better binned chip at a higher speed.
There will always be enthusiasts who push the limits - the ones who try to find the best “bang for buck” after you tweak it to the 9s (the Celeron 333A crowd, the i3 7350K or 9350K crowds - even the R7 1700 crowds, to an extent). But they’re a minority.
By and large, people who buy a computer will leave it absolutely stock. Hence you will find gamers who get a 3800X, or even a 3950X, simply because it’s a faster gaming chip without doing anything to it aside from turning it on. And these chips command a more substantial profit margin than a high-clocked 3600 would (and certainly more than one an end user OCs themselves).
*ed: Someone mentioned in the other thread that there are services which will OC your chips for you (and even de-lid them, likely only for Intel chips). They’re probably cheaper than buying a higher-end, unnecessarily wide chip. But I didn’t even know such a service existed until they mentioned it (I’d assumed, naturally), and that would only be cheaper for people nearby to such a facility.
Much like Cardinalsins, I’m from Aus. I dread to think what it would cost to ship a CPU, motherboard, and RAM (the latter two have a significant bearing on OC results) both ways to such a place, never mind the cost of the service itself.
AOI cooling is mainly to help reduce throttling on the 9900K if the environment’s temps are bad. Overall both CPUs are well within air cool-able territory.
We don’t know that. Elchapuzasinformatico’s 2700X pre-release gaming benchmarks were completely false. Also the 9900K on their site is underperforming (most likely not using a power-guzzling mobo that violates the 9900k specs).
Supposedly the silicon can hit 5Ghz from Chinese sources but the performance gains mostly stop at 4.8Ghz per non-final silicon. Either way with a 10-15% ipc lift a 4.6Ghz Zen2 will be equal to a 5Ghz Zen+.
Also we can’t expect insane high clock if we want to keep compatibility with garbage tier A320 boards.
I’m thinking they’re segmenting based core count also. The single threaded performance isn’g going to be head and shoulders difference between the 3600 and 3950X considering their main ST perf increase is from tweaking RAM timings.
AMD is limiting max clocks based on BETA BIOS settings. Not sure if the final BIOSes will keep that though.
Ryzen 2 would be 4.6Ghz all core. There is no ‘single-core boost’ on that class of CPUs.
Really? I hadn’t heard that, and it honestly doesn’t make a lot of sense. Even with the IPC improvements there is still going to be times where power and thermal restrictions will kick in much sooner with more cores than with fewer, and artificially limiting the fewer-core top speed is just throwing away potential gains (both real-world and benchmarked).
I’m of course open to being wrong, but it doesn’t seem right to me at all. Unless those speeds are a hard limit for the silicon - but even that seems weird given the existence of PBO (which reportedly is a flat +200MHz overclock switch).
![:rofl:](https://d38bqls1q93fod.cloudfront.net/en/wow/images/emoji/twitter/rofl.png?v=6" ":rofl:")