AMD Ryzen 7 2700 and Ryzen 5 2600 (12nm)

[HEXUS]

Worth going for non-X this time around?

Read more.

[CAT-THE-FIFTH]

The Ryzen 5 2600X is not worth the extra cost over the Ryzen 5 2600 since the price difference is enough to get a better cooler anyway. However,the Ryzen 7 2700X is worth it over the Ryzen 7 2700 for a normal sized system as the cooler is better and probably worth the extra money.

[kalniel]

Good for the supposed TDP cap, but doesn't seem to be that much more efficient in the real world. I was expecting much better things from precision boost 2 - manual overclocking shouldn't give a noticeable increase, but perhaps the TDP cap is limiting it. Is there a way to increase the TDP cap and then letting PB2 do its thing?

[MLyons]

Good for the supposed TDP cap, but doesn't seem to be that much more efficient in the real world. I was expecting much better things from precision boost 2 - manual overclocking shouldn't give a noticeable increase, but perhaps the TDP cap is limiting it. Is there a way to increase the TDP cap and then letting PB2 do its thing?

should happen in AMDs new overdrive but I don't think it's currently implemented.

[Tarinder]

Yup, didn't see it in the options. Waiting for support.

[CAT-THE-FIFTH]

Good for the supposed TDP cap, but doesn't seem to be that much more efficient in the real world. I was expecting much better things from precision boost 2 - manual overclocking shouldn't give a noticeable increase, but perhaps the TDP cap is limiting it. Is there a way to increase the TDP cap and then letting PB2 do its thing?

Most reviews found manual overclocking didn't do much for the X series parts and sometimes peformance was worse,so its definitely fine for the X series parts to the extent overclocking currently is not worth it:

https://www.hardocp.com/article/2018...rism_deep_dive

The main issue is apparently Precision Boost Overdrive is not properly implemented,so there might be additional gains to be made.

However,most reviews for the non-X parts show decent gains from manual overclocking,so I assume the CPUs are hitting the TDP wall as the 2700X has a 105W TDP for example - power consumption seems to be around 20W to 30W better for the non-X parts looking at multiple reviews.

[Fizzl]

Only 10% more for the X and half the fun of a SFF is making the cooling work.

[CAT-THE-FIFTH]

Part of the reason why you are not seeing as much a difference with the Ryzen X and non-X parts is down to the node used.




The 14NM had some noticeable voltage increases once you started going past the 3.3GHZ mark.

12NM OTH,seems much flatter overall,and requires less voltage.

[Firejack]

....

https://www.hardocp.com/article/2018...rism_deep_dive ....

Kyle from HardOCP posted a follow up to that article explaining that Precision Boost Overdrive and XFR Enhanced are future features, currently not implemented: https://www.hardocp.com/article/2018...nced_confusion

[lumireleon]

After the benchmarks please explain why a friend still ops to get the i7 8700K or any other intel chip? is intel still the best? I don't get it. He speculates that when he is compiling his complex programs on AMD since FX days they have a frequent bad habit of hanging (freezing) during compile but when he does the same with intel (any +ivy bridge) gives minimum issues. So he said he aint ever getting AMD again! How do you convince such a programmer?

[afiretruck]

After the benchmarks please explain why a friend still ops to get the i7 8700K or any other intel chip? is intel still the best? I don't get it. He speculates that when he is compiling his complex programs on AMD since FX days they have a frequent bad habit of hanging (freezing) during compile but when he does the same with intel (any +ivy bridge) gives minimum issues. So he said he aint ever getting AMD again! How do you convince such a programmer?

With difficulty, I'm afraid. If he's like most other programmers out there (myself included) there's going to be a certain element of pride in play here. I.e. He just won't like being told his hardware choice was the wrong one and will actively defend it. I'm speculating on pride here as his reasoning is based on speculation and observations instead of hard facts (such as the 8700K having lower memory latencies and typically higher single-threaded performance).

I think the only way you could really convince him is to show him how much better your similarly-priced AMD Ryzen/Threadripper is at tearing through the exact same workloads.

Finally, if his 8700K *is* faster than, say, a Threadripper 1920X he really needs to sort his compiling jobs out. Most "complex" programs are made of 10s, maybe even 100s of embarrassingly-parallelisable compilation units, which the Threadrippers and Ryzen are fantastic at.

For full disclosure, I went with a Threadripper 1900X (planning on upgrading to the full-fat 3rd-gen Threadripper when it comes out) mostly because my compiling workloads are highly parallelised.

[Irien]

Lumireleon - it seems likely that the FX-era AMD problems were more with chipset/motherboard than the actual CPU core. One of the things Intel have executed *really* well on since the core2-era is producing a really solid package (cpu and supporting chipset). Widespread Intel adoption also made RAM compatibility extremely solid on Intel cpu/motherboard combos. It was (almost) impossible to produce a really bad intel implementation, which has made Intel-ecosystem really attractive to PC builders in the last 10 years.

The reality is that unless the customer is pro-AMD to begin with, or at least ambivalent, this isn't going to change overnight. The early Ryzen boards did nothing to help - RAM compatibility was spotty, and many aspects seemed immature at launch. Realistically, this dogged AMD through the whole of Ryzen "year one", and is the single biggest reason to recommend the 400-series mobos, and 2000-series CPUs. The reviews of both have been very solid. My suspicion is that they'll need to keep executing this solidity across a couple more generations to turn the tide in ecosystem credibility. I just hope (*really* hope) that architecture refinements in Zen2 don't trigger similar issues, otherwise it is back to square one.

So, how to convince your friend? Well, I'd start by pointing at the various reviews of 400-series boards and 2000-series CPUs. Many of these comment on just how well AMD executed this launch, with little-to-no compatibility issues. Secondly, point him at Threadripper which also had an excellent launch (thanks to work done by AMD in the intervening time). Look at the platform "bang for buck" (eg. Threadripper compared to Intel HEDT solutions, or motherboard+cpu costs on Ryzen 2000-series).

Beyond that, he may feel he needs the Intel single-threaded grunt. If that's the case, he's genuinely better off sticking to Intel. I have quite a few customers in this boat, and would dearly love to build them AMD systems, but the honest appraisal is that Intel will serve them better for the software they need.

And, going back to where we started, that Intel Ecosystem thing is still a big deal. If the customer doesn't need AMD's strengths, it is simply hard to build a bad Intel PC by combining a respectable-brand mainstream mobo, ram and PSU etc. Going AMD means doing some homework to ensure best compatibility (eg. ram). Not everyone wants to do that.

[Zak33]

that is SO much grunt for 65W

If cars had improved on fuel use like CPU's have, we'd have no emmission worries at all and be driving V6 and V8's all over the place

[3dcandy]

that is SO much grunt for 65W

If cars had improved on fuel use like CPU's have, we'd have no emmission worries at all and be driving V6 and V8's all over the place

8c/16t in 65w is amazing. And at a decent speed too

[DanceswithUnix]

that is SO much grunt for 65W

If cars had improved on fuel use like CPU's have, we'd have no emmission worries at all and be driving V6 and V8's all over the place

Nah, the cars would just weight 40 tonnes rather than the current 2 tonnes to keep them slow

[sweatyBallacks]

The 2600 achieving an all-core overclock of 4.25Ghz would be a little monster for the price. That's worth it over the X-version.

The 2700 not so much. But 8-core/16-thread in a 65W package is one hell of a CPU for SFF builds. Small but very powerful. Normal desktop I have so I will be upgrading to a 2700X.