TR Forums

Howdy all. Seeing as the press all got the 2700x, and not the 2600/2700, I'm curious what sort of performance and frequencies the 2700 non-X can hit under various core loads. I have one on reserve at Microcenter with a B350 board and I'll post results if anyone is interested, but if anyone gets their 2700 and wants to post, feel free.

Looks like it hovers between the 1700x and 1800x which I find impressive given its fairly low power consumption.

Looking forward to TR's findings once they get the chips.

At a guess it's going to be an inferior choice to the 2700X in many ways - I'm assuming you want it for the lower TDP?

At just $30 less than the X model, you lose a lot of clockspeed and lose Wraith Prism too, which is probably worth around $30 by itself. The Wraith Spire is adequate for the 65W variants, but it's definitely a much less capable cooler.

If you're going to run at stock and want to cram it into a shoebox mITX case or just keep it cool and quiet, there's no reason to change your order, but for an extra $30 I would definitely be looking at the 2700X. It'll likely be no louder than the 2700 anyway, thanks to the superior heatpipe cooling.

Also, Anandtech were given the 2700X and analysed power usage against the 1800X at different loads 1T to 16T. What's significant is that therereally isn't much change in the power curves, so a lot of the performance you get out of the 2700X is down to the increased TDP, which is lost to the vanilla 2700; Unless you've obtained a bundle discount, I'd be tempted to look at the 1700 instead of the 2700 since you'll save around $50 for minimal difference, given the matching TDP.

I'm a bit torn between the 2700 and 2700x. The price difference is not extreme, and my reasoning is that I have a mATX case, and the B350 boards all come with 4 phase VRMs. I don't plan on overclocking, but given what I'm seeing with the 2700x's power draw, I think the 2700 will come close enough under most workloads to the 2700x while giving me the peace of mind that I'm not melting my power delivery on my motherboard.

I'll be transferring over my Noctua U9s from my current 3770 build so the cooler doesn't really matter. I'm also concerned the higher capacity cooler will boost performance and power draw and further put pressure on the VRM's.

Well, a quick search hasn't given me a review that graphs boost speeds per loaded core of the newer 2700. I believe that AMD's tweaking of the dynamic boost means you will use more of that 65W TDP than a 1700, which only has 1C, 2C, All-core states. In the 3C-7C states, I think the 2700 will be a reasonably decent improvement over its predecessor.

In saying that though, I still think a stock 2700X is a better choice for you. VRMs on the B350 will be actively cooled by the air cooler, it's not as if you're using an AIO with no airflow around the VRMs, and the boards are officially rated to handle more than 95W, otherwise they'd be A320 boards.

Thanks Chrispy. I appreciate the thoughtful responses.

Does the ryzen platform have a tdp limit? Like the intel TDP limiter? I could see myself buying the 2700x and limiting its power consumption to 95 watts or something as a nice compromise.

computerbase.de shows the 2700 holding 3.9ghz on a 4 thread load and 3.4ghz on a 16 thread load. I'm cool with letting compiles run 10% longer than they would with the 2700X. Gaming performance is within 10% (and I'm sure I'd see little to no benefit with my GTX 960).

turtlepwr281 wrote:

Thanks Chrispy. I appreciate the thoughtful responses.

Does the ryzen platform have a tdp limit? Like the intel TDP limiter? I could see myself buying the 2700x and limiting its power consumption to 95 watts or something as a nice compromise.

computerbase.de shows the 2700 holding 3.9ghz on a 4 thread load and 3.4ghz on a 16 thread load. I'm cool with letting compiles run 10% longer than they would with the 2700X. Gaming performance is within 10% (and I'm sure I'd see little to no benefit with my GTX 960).

I haven't seen boards with TDP limiters, if there is one it's very likely to be like Intel implementations and entirely down to the board vendor.

I did pick up a R7 2700 this weekend, along with an Asus B350 Tuf mATX Board and 3000mhz CL15 Corsair Ram. Microcenter's stock of B350 Tuf's had Ryzen 2000 ready stickers on them and everything worked.

I don't have much to say about performance,coming from an i7-3770 I'm not entirely sure the system feels any snappier (and I wasn't anticipating it would). I'm only feeding a GTX 960 and most processors from the last 6 years can saturate one of those. I was mainly looking for moar coars to throw at linux so I can start goofing around with kernel compiles and whatnot. Everything is supported out of the box on Mint 18.3 and things are very fast there.

It does appear to hold 3.4ghz on an all-core load and and lightly threaded workloads hover between 3.7 and 4.1ghz. HWINFO says a total package power consumption of 70-72 watts during nT thread Cinebench runs in Windows.

The Wraith Whatever cooler had an LED ring and a copper slug, as well as a fan with a noisy bearing. It seemed to keep the processor around 65ish with Noctua paste. Once my Noctua kit came from Amazon I swapped over to my U9S and have enjoyed the silence.

I'm happy with the performance and I feel it was a worthwhile upgrade, especially having to sell the soul of my firstborn to afford the RAM.

3.4 all-core isn't too bad for the same power usage as a 1700.

The 1700s here are render nodes, where performance/Watt matters and they mostly run at 3.2GHz on all cores, so that's an efficiency improvement from the new process node, by the looks of it.

If you're using faster RAM, that's something the old 1700 isn't good with either, it tops out at 2933 with loose timings, or 2666 with tight timings.

Chrispy_ wrote:

If you're using faster RAM, that's something the old 1700 isn't good with either, it tops out at 2933 with loose timings, or 2666 with tight timings.

?

I haven't heard of the 1700 non-X being any worse at that than other SKUs, and mine certainly isn't. It was a bit rocky out of the gate like the rest of the SKUs, but there's nothing wrong with it on 1.0.0.6b.

At this point the XMP specs on my 1.35V DDR4-3000 CL15 are well within bounds. I'm running it at 1.30V 2933C16 because reliability still matters more than speed for this rig, but many would feel comfortable running it at 1.35V 3200C16, and that's with some of the memory that took manual tweaking to get much of anywhere on early AGESAs.

Hmm, I might have just been unlucky on the one I played with. We have 8 R7 1700 nodes in a farm stack, but I only played with overclocking and RAM compatibility with one of them back at launch of the 1700.

It overclocked reasonably well but DDR4-2933 required voodoo: I had to put 2133 in, manually offset timings and then swap to 2933 just to get it to POST. I actually haven't tried it since those early AGESAs on the 1700.

A bit of some voodoo or other seems the norm when pushing the limits, unfortunately. The one I keep running into is that after pushing it too far and needing a CMOS clear, it becomes unstable at values that were working right before that (and sometimes even much more relaxed ones). It seems to need a few power cycles (or time running, or something like that) under low stress to get its bearings again. At least the limits seem to be comfortably high at this point.

(Ninja edit: If anyone has an explanation for that behavior, I'd love to hear it. I've been giving it a half hour or more for CMOS clears lately, and with that out of the way have no idea.)

Chrispy_ wrote:

3.4 all-core isn't too bad for the same power usage as a 1700.

The 1700s here are render nodes, where performance/Watt matters and they mostly run at 3.2GHz on all cores, so that's an efficiency improvement from the new process node, by the looks of it.

If you're using faster RAM, that's something the old 1700 isn't good with either, it tops out at 2933 with loose timings, or 2666 with tight timings.

The clock speeds are much higher for lightly threaded loads than the 1700 I'd imagine. Now that I've gotten my Noctua U9S on here, I've taken a look at HWINFO and one of my cores does hit 4.15 and the rest hit 4.11 fairly regularly. Also total package consumption has risen to a max of 81.63 watts. Looks like whatever frequency voodoo is pushing me harder in lightly threaded games (World of Warships).

Chrispy_ wrote:

3.4 all-core isn't too bad for the same power usage as a 1700.

The 1700s here are render nodes, where performance/Watt matters and they mostly run at 3.2GHz on all cores, so that's an efficiency improvement from the new process node, by the looks of it.

If you're using faster RAM, that's something the old 1700 isn't good with either, it tops out at 2933 with loose timings, or 2666 with tight timings.

I'm curious why some people keep stating this... One of the absolute first systems I built on Ryzen was the 1700 on an Asus x370-a Prime. The only issue I had was when that 1700 was fresh out and the BIOS versions were really fresh (like v801). It's currently using D.O.C.P to run G.Skill DIMMS at 3200 at 14-14-14-34. That was on the 3803 BIOS and I recently did the 4004 (IIRC) update and did a full Memtest on it that ran for probably about 6 hours. Not a single hiccup. I'm also running 3200 16-16-16-36 on my 1600x and my wifes with no issue, along with about a dozen lower end 1500x systems also running on 18-18-18-38.

I just haven't ran into a single Ryzen system that wouldn't boot and run stable at 3200 at the advertised timings as per about the v3401 BIOS era for Asus. These are all G.Skill DIMMS and running across different boards to include Asus, MSI and Gigabyte.

Either I've been the luckiest bastard in the world when it comes to this issue, or I'm missing something.

The 1700 that I was playing with here last summer used some basic DDR4-3000 Vengeance LPX (4x16GB) using an MSI B350 mATX board and it absolutely refused to boot at anything above 2666 without severely slacking off the timings. Although we have many identical nodes running that config, I only messed around with CPU/RAM clocking on one of them, so it's a sample size of 1. In saying that, the initial Ryzen reviews last summer also indicated issues with DDR4 speeds at launch.

The 1800X I use isn't running at 3200 either, and that's had BIOS updates in the last 3 months. It's definitely stable at 2933 with tight(ish) timings though so it's better than the 1700 of the same vintage.

Maybe you're all talking about running minimal 2-DIMM RAM configs or something but Ryzen's strengths are large, heavily-threaded workloads so we always fill them up with RAM. Either way, I don't think the 1st-gen Ryzen's are guaranteed to run fast RAM like an Intel system. It's probably a mix of both you being lucky and me not having messed with the Ryzen7 1700 for almost ten months.

turtlepwr281 wrote:

Chrispy_ wrote:

3.4 all-core isn't too bad for the same power usage as a 1700.

The 1700s here are render nodes, where performance/Watt matters and they mostly run at 3.2GHz on all cores, so that's an efficiency improvement from the new process node, by the looks of it.

If you're using faster RAM, that's something the old 1700 isn't good with either, it tops out at 2933 with loose timings, or 2666 with tight timings.

The clock speeds are much higher for lightly threaded loads than the 1700 I'd imagine. Now that I've gotten my Noctua U9S on here, I've taken a look at HWINFO and one of my cores does hit 4.15 and the rest hit 4.11 fairly regularly. Also total package consumption has risen to a max of 81.63 watts. Looks like whatever frequency voodoo is pushing me harder in lightly threaded games (World of Warships).

Yep, that sounds like a decent increase over the 1700.
I was quite tempted by the 2700 for my ageing HTPC (which does a lot of encoding) because of the 65W TDP but sadly it looks like the higher all-core boost speeds are limited to the X models, with the 2700 losing a HUGE amount of ground to the 2700X once more than two cores are loaded:

Chrispy,

cool graph! I enjoyed the techpowerup review of the 2700. I appreciate that the 720p gaming results were within 5-10% of the 2700X, and a quick look at the task energy shows the 2700 being 25% more efficient for an all-thread load.

I compiled the linux kernel on mine and it took about 11 minutes. Does anyone know if there's a flag you can add that will time it?

My i7-3770 took about 25 minutes, so I'm pretty happy with the 2700 upgrade.

I've seen a few conclusions that I disagree about the 2700 though. I honestly don't think I'd notice if the kernel compiled in 9:30 or games ran 4 fps higher had I purchased the 2700X. I'm happy knowing my VRMs will run cool no matter what board I put this chip in in the future. I honestly don't think most B350's 4 phase power delivery can run the 2700X long term. 140 watts+ over 4 phases would be a very large current pull. Perhaps B450's will come with a higher capacity power delivery system. Or they'll come out with X470 mATX boards.

You can time any shell statement by using the time command with it.

https://linux.die.net/man/1/time

I'm jealous heh My 1700 has been offline most of the last six months because I didn't care enough to troubleshoot it for various reasons.

Now I'm finally trying to get it fixed o.o blargh

With my previous computers I probably would've just replaced suspect parts but things are a bit expensive for that right now.