Ryzen 2nd Gen

[Varvara Spiros Gelashvili]

anyway its a good refresh with the only dissapointing thing being OC - TPU only managed a 4.2 all-core OC on 2700X, which is only 100 mhz more than 1700X/1800X typical max OC


https://www.techpowerup.com/reviews/..._2600X/17.html

and 4.1 max OC on 2600X

Golden samples can possibly hit 4.35 but this give me hope for ryzen 2.

[Shakadam]

Ram prices arent actually that crazy, only for a short period of time were 16gb's of ram under 100 dollars, now they are ~150. Thing is for most people 8gb's is still adequate, 16gb only became the recommended go to when ram prices dipped below 100 for those few months.

Lowest pricing of any 16GB kit in Europe is ~160€, and that's low speed crap memory. Any decent kit is 200€ and above =/. So, ~250$ and above. It suuuuuuuuuuucks.

[Vash The Stampede]

My fault, when I did that multi quote I confused you with the other poster who was saying ram was around $600. You're running the 8350 on water cooling? If you don't mind my asking which setup/brand did you go with? I've considered trying water cooling again for a while now but ended up going with Noctua.

My water cooling setup is mostly from Ebay, from China. Though my radiators are copper from known brands, I think XSPC, but I bought whatever was the cheapest. CPU water block is from China, this one looks like it. The pump is a Delphi Water cooling pump ddc28 that I got off Ebay for $20. I can't find it now but it's basically a Laing DCC pump. And the tubing is regular flexible kind and not rigid, cause I'm too lazy to convert to rigid. Maybe if I switch to Ryzen I'll go rigid.

My HTPC has a similar setup and that runs a AMD A10 5800k 24/7 except with cheap Chinese Ebay parts. That too has been running for years with no maintenance done. I got sick and tired of it and just put 50/50 green car coolant. No erosion and no algae growth.

[Epoch]

My water cooling setup is mostly from Ebay, from China. Though my radiators are copper from known brands, I think XSPC, but I bought whatever was the cheapest. CPU water block is from China, this one looks like it. The pump is a Delphi Water cooling pump ddc28 that I got off Ebay for $20. I can't find it now but it's basically a Laing DCC pump. And the tubing is regular flexible kind and not rigid, cause I'm too lazy to convert to rigid. Maybe if I switch to Ryzen I'll go rigid.

My HTPC has a similar setup and that runs a AMD A10 5800k 24/7 except with cheap Chinese Ebay parts. That too has been running for years with no maintenance done. I got sick and tired of it and just put 50/50 green car coolant. No erosion and no algae growth.

Awesome! My last foray into water cooling was around 2005, AMD dual core & a radeon x800 xt platinum edition. Thought I had done something wrong, tried quite a few water cooling and air cooling options - it ran above what I considered reasonable temps. Got tired of it after sending it back twice for replacements that did the same thing and went with (iirc) the geforce 6800. Built a system around a ryzen 1600x & the noctua nh d15 (my god that beige fan color ) a few weeks back, debating windows 10 + if it would be worth to sli nvidia 980 gtx's and considering picking up another samsung 850 evo.

[Vash The Stampede]

Awesome! My last foray into water cooling was around 2005, AMD dual core & a radeon x800 xt platinum edition. Thought I had done something wrong, tried quite a few water cooling and air cooling options - it ran above what I considered reasonable temps. Got tired of it after sending it back twice for replacements that did the same thing and went with (iirc) the geforce 6800. Built a system around a ryzen 1600x & the noctua nh d15 (my god that beige fan color ) a few weeks back, debating windows 10 + if it would be worth to sli nvidia 980 gtx's and considering picking up another samsung 850 evo.

Sounds like one of those Water Cooling kits. I've bought one of those like in 2003 or something and that thing failed really quickly. The pump failed and the tubing was clogged from algae. Years later I tried to see if I could build one better, and honestly I also ran into the algae problems, despite using additives that were suppose to prevent that. Erosion was also a problem, due to mixing aluminum and copper, which turns your water loop into a battery. Then I thought, why don't cars have this problem? That's when I used car coolant. No problems since then.

[Deleted]

FWIIW Anandtech seems to have updated their benchmarks and they now more or less match the other reviews, i.e. i7-8700k is beating the 2700x by a good margin in gaming (10-20%).

So while Ryzen+ has slightly improved at gaming the i7-8700k is still king there and at certain other workloads (AVX, Adobe) - with Ryzen+ being even further ahead at workloads that can be parallelized well (rendering) and both are trading blows on most other stuff.


No explanation what the problem with the first set of results was yet, though.

[Fascinate]

Heh and people in this thread were giving me crap about not trusting anands results. I swear some websites just simulate benchmarks without actually running them, which is of course bad and incredibly lazy.

[Deleted]

If 16gb ram prices drop to 100-120 range i would order a ryzen 2 build right now.

[mrgreenthump]

With overclocking being all but dead on Zen+ (aside from BLCK), I wonder if Intel follows suit with their own precision boost type system and we basically have no need for overclocking in the future.

FWIIW Anandtech seems to have updated their benchmarks and they now more or less match the other reviews, i.e. i7-8700k is beating the 2700x by a good margin in gaming (10-20%).

The overall average over early reviews was around 5% difference, so they went from being an outlier to being an outlier in the other direction.

- - - Updated - - -

Heh and people in this thread were giving me crap about not trusting anands results. I swear some websites just simulate benchmarks without actually running them, which is of course bad and incredibly lazy.

Anandtech has been somewhat falling for a long time now, their tests hasn't been to the level we were used to before.

[Shakadam]

FWIIW Anandtech seems to have updated their benchmarks and they now more or less match the other reviews, i.e. i7-8700k is beating the 2700x by a good margin in gaming (10-20%).

So while Ryzen+ has slightly improved at gaming the i7-8700k is still king there and at certain other workloads (AVX, Adobe) - with Ryzen+ being even further ahead at workloads that can be parallelized well (rendering) and both are trading blows on most other stuff.


No explanation what the problem with the first set of results was yet, though.

Just wanted to say that the difference is smaller if you're using higher quality memory (meaning higher speed, lower timings).
https://www.computerbase.de/2018-04/...harfen_timings

With optimized memory, the gap between the Core i7-8700K and the Ryzen 7 2700X is reduced from eleven percent to five percent.

Note: Both Intel and AMD running optimized memory that is.

[sil3ntearth]

DDR4 prices ruin the chance of me upgrading to anything from Haswell refresh, I refuse that scam. ~5 USD / GB or GTFO

I agree. That's why I haven't upgraded one of the old FX 6300 machines I have sitting around.

[Deleted]

The overall average over early reviews was around 5% difference, so they went from being an outlier to being an outlier in the other direction.

Well I suppose that depends on what you include in that average - also that 10-20% number was my quick estimate of their results.
And if you include the high resolution benchmarks then the average is going to be lower.

Also it depends somewhat on what games you bench and anandtech's selection differs a bit from the usual choice.

Edit:

Update 04/25: After publishing this review last week, it quickly became clear that some of our results were not aligned with those from other media. Initially we were under the impression that this was as a result of the Spectre and Meltdown (or Smeltdown) updates, as we were one of the few media outlets to go back and perform retesting under the new standard.

However in order to ensure that our results were accurate and completely reproducible, we decided to undertake an extensive internal audit of our test results. As a result of that process we believe we have found the reason for our test results being so different from the results of others, which we have documented in detail in this follow-up article. In short, an issue with a non-standard use timers on Windows was causing the performance of both AMD and Intel processors to dip, particularly impacting the latter.

Now that we know what the issue is and how to correct for it, we are going back and re-running all of our benchmarks for this review to generate updated results. Some of those results have already been added to this article, and the rest are expected next week. In the meantime I have decided to retracting our conclusion for this article until the rest of our results are in, and we can update our conclusion to account for any differences. -Ryan Smith

- - - Updated - - -

https://www.anandtech.com/show/12678...ryzen-results/

Interesting read.

There's a bunch on timers in a PC and one of them is called High Precision Event Timer and you can force all other timers to be derived from it.
This however appears to have an enormous impact on their gaming benchmarks especially on Intel systems (up to 76%) compared to AMD (up to 12%).

The old data was taken with forcing all timers being derived from the HPET and apparently causing the i7s bad performance.

Now I don't recall the old data exactly anymore but that effect seems stronger than the new vs old difference - but maybe that's just my memory playing tricks.

However for me this begs the question - how exactly is performance (i.e. fps) on any benchmark measured? Does it depend on any timer in the system or is it derived from some independent clock?

[durenas]

The TLDR is that HPET should not be forced on if you're running an intel chip, or spectre/meltdown changes will impact you even more, and overclocking utilities often need this to be on in order to work. So once you've got your overclock stable, disable force HPET in the bios.

[Life-Binder]

disabled HPET in BIOS years ago, it does absolutely nothing on Win10 as far as I can tell (Win10 has better more modern timers I think) and may or may not hamper performance (at least on Intel)

lol Anand


dont know about OC utilities, but I OC CPU from BIOS and the GPU from Afterburner and HPET wasnt needed in either case




in other news - https://www.techpowerup.com/243673/r...er-joins-intel

Zen creator Jim Keller joins Intel


I wonder if he actually finished his Tesla work or if he dumped them for Intel

- - - Updated - - -

according to the Anand article - apparently disabling HPET in BIOS can be a bad thing, but I have never encountered such a thing on my system in any circumstances

and Ive seen ppl recommend to disable HPET in BIOS on Win10

[Blackmist]

The reality is, this is the future. Multithreaded games will become more and more common. And Ryzen does really really well.

Still hampered by lacklustre single thread performance.

The reality is that although games can take advantage of 8+ cores running many threads simultaneously, most of the time you'll be held back by just one of them. Weakest link in the chain and all that. DirectX12 and Vulkan were supposed to prevent this, meaning you'd see flat performance across the board as everything would be GPU limited, and while that seems to be common at 4K, it's just not the case at 1080p.

But when you're talking about the difference between 120 and 140 fps on a GTX1080ti is that even relevant? Who is running 1080p on a card like that? A GTX1060 would be a more relevant comparison, and I suspect they'll all run much closer together at that. I went for a i5 8400 because it runs within a few fps of far more expensive CPUs. I suspect if you spend more than £150 on a CPU you'll find it perfectly adequate for mainstream gaming.

[Sorshen]

Honestly, I'd still get the R5 2600... but I don't want to blow 175€ on 2*8Gb of 3000MHz DDR4 atm.

[Deleted]

Still hampered by lacklustre single thread performance.

The reality is that although games can take advantage of 8+ cores running many threads simultaneously, most of the time you'll be held back by just one of them. Weakest link in the chain and all that. DirectX12 and Vulkan were supposed to prevent this, meaning you'd see flat performance across the board as everything would be GPU limited, and while that seems to be common at 4K, it's just not the case at 1080p.

But when you're talking about the difference between 120 and 140 fps on a GTX1080ti is that even relevant? Who is running 1080p on a card like that? A GTX1060 would be a more relevant comparison, and I suspect they'll all run much closer together at that. I went for a i5 8400 because it runs within a few fps of far more expensive CPUs. I suspect if you spend more than £150 on a CPU you'll find it perfectly adequate for mainstream gaming.

Question, do you know the definition of the word lacklustre?

[Zenny]

Ryzen doesn't have lacklustre single threaded performance, it has lacklustre overclocking potential. Hopefully 7nm can put them on par or ahead of Intel in that regard.

[Afrospinach]

it has lacklustre overclocking potential.

So basically what you are saying is it has lacklustre single threaded performance?

Single threaded performance is not IPC.

[Deleted]

So basically what you are saying is it has lacklustre single threaded performance?

Single threaded performance is not IPC.

No it doesn't have lackluster single threaded performance... I has lackluster potential beyond what AMD specified them for. Doesn't change FACT that they have very good single and multi threaded performance - being slower than competition does not necessarily make them lackluster.

Also I wonder what Jim is going to work on at Intels. Surely he signed Non compete for x86 platform when worked for AMD..