Question Zen2 vs Zen3 improvements for DC?

Assimilator1 · Jul 8, 2023

Yea I know that's nearly old skool now! . But I don't keep my rig bang up to date anymore (if I ever did).

After noticing that my electricity charges have slightly more than doubled in 3yrs! From 13.44p/kwhr in 6/2020 to 30.3p/kwhr last month!
And that's with a subsidy, otherwise it'd be 48p/kwhr!! [edit] (Just discovered the subsidy is being phased out from this month! I hope the prices drop soon!).
Which means with a wall draw of 112w (CPU crunching only), running 24/7 the monthly cost has gone from ~£11/month > ~£25/month > ~£39/month! (£470/yr!).
[edit] 11-11-24 now 23.44p/kwhr

I'm finally look at upgrading my Ryzen 3600 to maybe a 5600 or 5600X, IIRC they use a little less power (as well as being a little faster).
Anyone remember the difference? Looking at AT's review of the Zen3 the 5600X drew a little over 10% less power than the 3600, and Zen 3 has ~19% faster IPC than Zen2 (plus a little clock boost).
A post by biodoc showed his 5950X was ~11% faster than his 3950X in wanless (Mersenne+2), is this typical for DC?

If so I think I'll start hunting for a Zen3! I see the 5800 [edit, seems to be OEM only, but the 5700X is the same thing?] has the same TDP as the 5600/5600X, but will it draw more power overall under full load anyway? (I'm guessing that the 5800 all core load clocks will be less than the 5600X, but that it's extra 2 cores more than make up for that for ppd?).
I need to see if my mbrd BIOS is upto date for these newer CPUs (rig spec in sig btw).

StefanR5R · Jul 8, 2023

For several of PrimeGrid's CPU subprojects, Zen 3's larger core complexes with 32 MB last-level cache per CCX (in desktop Zen 3 flavors at least) is a big improvement over Zen 2's 16 MB sized last-level cache segments. E.g. from the projects of the 2023 PrimeGrid challenge season, PPS-MEGA, GFN-17-MEGA, SGS-LLR, 321-LLR, GFN-18, GFN-19 are still small enough to fit one or more of such tasks into a Zen 2 CCX. But PSP-LLR, SoB-LLR, ESP-LLR, and GFN-20 are too big for a Zen 2 CCX, and the resulting cross-CCX communications starves the vector arithmetic units. Zen 3 CCXes however can still go full throttle at these bigger projects.

Assimilator1 · Jul 8, 2023

Interesting, and I'm guessing a very large difference in performance!

Looking on ebay for 2nd hand Zen 3s it seems the plain 5600s are rare (5800s non existent), most are the 5600Gs (any disadvantage to having one with onboard grx?)
The few 5600s and 5600Xs seem to go for about £85-95 (some of the latter going for £120! wth??[edit] actually, most are £120-130!)
5700Xs prices some where in between at £130-150.
5800Xs for about £150+....,

cellarnoise · Jul 8, 2023

Zen 4 is even more efficient and has avx-512 . Most of the desktop zen family you can set your own power use limits.

The desktop apus have less level 3 cache per core than the non-apu models.

Assimilator1 · Jul 8, 2023

Yea but Zen 4 needs a new mbrd and RAM right? I can't afford to spend a load more atm.

APUs? Ah I guess the 5600G is an APU, I see it's L3 cache is 16MB vs 32MB for the 5600, good to know!

cellarnoise · Jul 8, 2023

Zen 4 does require new MB and ddr5.

We have found the entry level ASRock works well though and has been most affordable.

5950x works well power restricted all the way down to 65 watts even, though a bit better at 100w or so. Double your output!

Assimilator1 · Jul 8, 2023

The 5950 would be great , but what would happen to 3-4 core clock speeds? I do occasionally game.
Actually, nm, having seen the prices for 5800Xs I wouldn't want to pay for a 5950!

cellarnoise · Jul 8, 2023

Maybe 5900 then? Better price per-core !

Looks like 5700x non-apu used might be best price per core?

StefanR5R · Jul 8, 2023

cellarnoise said:
The desktop apus have less level 3 cache per core than the non-apu models.

Indeed; the 5600G in particular only has got 16 MB L3$. As do the 5700G and even the 5500 which is based on the same die but has got the iGPU disabled.

Edit,

Assimilator1 said:
I'm guessing that the 5800 all core load clocks will be less than the 5600X, but that it's extra 2 cores more than make up for that for ppd?

Yes, within the (by default) 65 W TDP/ 88 W PPT limits, the 8 core CPUs will have higher throughput than 6 core CPUs of the same generation, but I don't know exactly how much higher.

Assimilator1 · Jul 8, 2023

cellarnoise said:
Maybe 5900 then? Better price per-core !

Looks like 5700x non-apu used might be best price per core?

Lol, I'm sure it is, but probably not enough to make it cheaper

And yea the 5700X/5800 does look like the best deal, and looking at this review of the 5800 by GN it used less power than my 3600 running their Blender benchmark!

drnickriviera · Jul 11, 2023

Heck i'm still wondering what improvement I would get from Zen to Zen 2 in WCG? May find another 3700 to replace my 1600

StefanR5R · Jul 12, 2023

6c/12t Zen+ @ 65/88 W --> 8c/16t Zen 2 @ 65/88 W should bring an improvement in many all-core workloads at an order of magnitude of +100%, shouldn't it? (More cores, much wider cores especially WRT floating point units and cache, faster clocks at same power thanks to GloFo 12nm --> TSMC 7nm.)

Assimilator1 · Jul 12, 2023

There's a big jump in performance from Zen+ to Zen 2, I think the AnandTech article I linked in my op shows something on that.

Assimilator1 · Jul 15, 2023

Actually, it's not as much as I thought going from Zen+ to Zen2, AMD's slide in the AT article I linked in the op shows an IPC increase of 15%, and showing about a 8% increase in max boost clocks. I looked at AT's article on the Zen 2, it sees the 3700X using 32% less power than the 2700X!

StefanR5R · Jul 15, 2023

For all-core workloads¹ like in distributed computing, and especially for floating-point heavy and vector arithmetic heavy workloads², the improvement from Zen+ to Zen 2 is a lot larger than what IPC increase and max boost clock increase tell us.

¹) mainly due to the GloFo 12nm --> TSMC 7nm move
²) mainly due to the ≈doubled FP bandwidth/ ≈doubled AVX bandwidth per core

I never had a Zen or Zen+ CPU myself. My own main comparison is between Broadwell-EP and Rome, and specifically the 22c/44t SKU vs. a 32c/64t SKU of these. Machine throughputs are near the order of 1:2, and power efficiency at the order of 1:2.5 in many distributed computing workloads. (Sometimes a bit more, sometimes a bit less.) — As far as I recall, Zen and Zen+ perform somewhat comparable with Broadwell(-X,-EP) when driven in similar clock ranges.

The Zen 2 to Zen 3 step is quite different: The core was made quite a bit wider, but the process node stood the same. The result was that light workloads get a pretty good uplift in Zen 3 vs. Zen 2, while power-hungry workloads, notably all-core workloads, see only mild improvements (from what I remember reading; I don't have Zen 3 CPUs myself). Special exceptions are those which profit from the different CCX organization, as mentioned in post #2.

mmonnin03 · Jul 16, 2023

StefanR5R said:
For all-core workloads¹ like in distributed computing, and especially for floating-point heavy and vector arithmetic heavy workloads², the improvement from Zen+ to Zen 2 is a lot larger than what IPC increase and max boost clock increase tell us.

¹) mainly due to the GloFo 12nm --> TSMC 7nm move
²) mainly due to the ≈doubled FP bandwidth/ ≈doubled AVX bandwidth per core

I never had a Zen or Zen+ CPU myself. My own main comparison is between Broadwell-EP and Rome, and specifically the 22c/44t SKU vs. a 32c/64t SKU of these. Machine throughputs are near the order of 1:2, and power efficiency at the order of 1:2.5 in many distributed computing workloads. (Sometimes a bit more, sometimes a bit less.) — As far as I recall, Zen and Zen+ perform somewhat comparable with Broadwell(-X,-EP) when driven in similar clock ranges.

The Zen 2 to Zen 3 step is quite different: The core was made quite a bit wider, but the process node stood the same. The result was that light workloads get a pretty good uplift in Zen 3 vs. Zen 2, while power-hungry workloads, notably all-core workloads, see only mild improvements (from what I remember reading; I don't have Zen 3 CPUs myself). Special exceptions are those which profit from the different CCX organization, as mentioned in post #2.

I've got 1950x and 2700x CPU/mobo/RAM combo's not in use if you'd want any to try.

Assimilator1 · Oct 19, 2024

I'd totally forgotten about this, lol.
Now having plenty of WUs again, but lacking CPU throughput , I thought I'd re-visit it.
Btw, I don't want to change platform.

Ebay prices again for used CPUs, most sold are in this range now.
5600Xs (6C/12T) .... £80-90
5700Xs (8C/16T) .... £95-115
5800Xs (8C/16T) .... £115-130
5900Xs (12C/24T) .. £155-175

And it seems I can sell my 3600 for £40-50.
A thought, if I were to restrict a 5800X to 65w PPT, would it then perform pretty much like a 5700X and make it pointless to get the former?

Assimilator1 · Oct 21, 2024

Assimilator1 said:
A thought, if I were to restrict a 5800X to 65w PPT, would it then perform pretty much like a 5700X and make it pointless to get the former?

StefanR5R · Oct 21, 2024

I suspect they would be indistinguishable then. My *guess* is that it is reasonably easy to get a 5800X to operate at similar Voltage:frequency as a 5700X, and vice versa. But I don't own Ryzens myself, nor have I been looking at Ryzen related discussions much.

For what it's worth, 5800X was first produced at the B0 stepping, later at B2 stepping. 5700X which was introduced later than 5800X was produced at B2 stepping right away. I don't recall whether there was any end-user relevant difference between the steppings.

In2Photos · Oct 21, 2024

Assimilator1 said:
A thought, if I were to restrict a 5800X to 65w PPT, would it then perform pretty much like a 5700X and make it pointless to get the former?

Restricting the 5800X to 65W would likely lower the max frequency to something similar to the 5700X. And even in gaming there doesn't seem to be too much improvement between the two.

In the Gamer's Nexus review of the 5700X the blender benchmark had the 5800X at 11% faster than the 5700X (15.7 minutes vs 17.7 minutes), but it consumed nearly twice the power (127.2W vs 63.6W).

Assimilator1 · Oct 21, 2024

Nice one, thanks guys 👍

StefanR5R · Oct 26, 2024

BTW, the default power limits are actually twofold:

The 65 W tier of Ryzens has got 65 W TDP (thermal design power, basically the absolute minimum of cooling capacity of CPU cooler and computer chassis, according to AMD) and 88 W PPT limit (package power tracking limit, the maximum sustained power delivery to the CPU as long as it is not thermally throttling or constrained by maximum clocks or simply not needing it because the computational workload isn't as heavy).
The 105 W tier of Ryzens has got 105 W TDP and 142 W PPT limit. For an eight-cores CPU, the latter is almost synonymous with "unlimited", as it is unlikely to sustain this much power use unless highly overclocked.

Now, if you feel a lack of computing throughput but don't want to change platforms, how about more than eight cores? You could configure their TDP and PPT limit down to server levels.

EPYC Milan 7513 and 7543, that is, 32-cores SKUs, are basically four 8-cores Ryzens baked into a single package, and dialed down a lot WRT power limits and clocks. (They have got 4 instead of 1 cpu chiplet, and an I/O chip which is four times as big and capable as Ryzen's I/O chiplet.)
7513: TDP = PPT limit = 200 W = 4 × 50 W
7543: TDP = PPT limit = 225 W = 4 × 56 W
7543's cTDP High is 240 W = 4 × 60 W.
EPYC 7643, the 48-cores Milan ≈ four down-clocked 12-cores Ryzens:
TDP = PPT limit = 225 W = 4 × 56 W
7643's cTDP High is 240 W = 4 × 60 W.
EPYC 7713 and 7763, the 64-cores Milans ≈ four down-clocked 16-cores Ryzens:
7713: TDP = PPT limit = 225 W = 4 × 56 W
7763: TDP = PPT limit = 280 W = 4 × 70 W
7763's cTDP High is the same as its default, 280 W = 4 × 70 W.

(Source for the server CPU power limits: servethehome's table with Milan and Rome.)

So, how about replacing the 65/88 W Ryzen 3600 by a Ryzen 5900X or 5950X, configured down to the same 65 W TDP and 88 w PPT limit (or some more than these if cooling and electricity budget allow for it)? Clock frequencies will suffer, but computing throughput will still be higher with more but slower cores, at same power draw. Though during situations when you would like high clocks, e.g. while playing a game, you would need to suspend BOINC so that all power can go to the few cores which the game needs.

Also, if you were to go the route of a higher core count CPU, for some BOINC projects which need much RAM per task, a RAM capacity upgrade may be called for in order to get all CPU threads to participate rather than being suspended by BOINC due to main memory all used up. On the other hand, if you used e.g. only 50 % of all CPU threads of a 16-cores CPU, that would still be faster than using 100 % of all CPU threads of an 8-cores CPU, obviously, as the former would still utilize all of its actual cores (most of the time if you let the operating system do its thing, or all of the time if CPU affinity is set for BOINC) whereas the latter crams two processes onto each core by means of SMT.

Assimilator1 · Oct 27, 2024

I mentioned in post #7 I'm not willing to spend more than ~£150, I listed the 5900 in my more recent post out of curiosity, it's still more than I would want to spend unfortunately. And I'm not buying anymore RAM (did you see I have 32 GB already?).

Good point about 5900X & 5950X and BOINC during gaming, I sort of do that anyway as I restrict BOINC to 50% usage whilst playing Elite Dangerous. Originally I only left 4 threads spare, thinking that would be enough for Elite, but I would occasionally get stuttering. I thought perhaps it used more than 2 cores (despite info to the contrary), but perhaps it's down to the clock speed differences? With BOINC at 50% clock speeds go from ~3.5-3.6 GHz to 3.8-4 GHz (depending upon which project is running. Asteroids brings the clocks down to below base @100%! Despite temps remaining low. Talking of temps, CPU temps go up 3-4C when going from 100% BOINC to 50%, quite a step up).
Anyway, worth bearing in mind if I were to get a 5900X or 5950X in the future (very unlikely).

In the meantime, I won an auction for a 5800X for the price of a 5700X (~£110) , so although I'll be restricting it so that it (likely) performs like a 5700X, it hasn't cost me anymore. The only thing I didn't take into account is the later CPU stepping the 5700Xs have, that was mentioned by you, I meant to look into that but simply forgot! Oh well, too late now, I will look into it anyway out of interest.

On another note, I hope my current cooler is upto the task! I think it will be, but I'm not sure, lol.

StefanR5R · Oct 27, 2024

Assimilator1 said:
I hope my current cooler is upto the task! I think it will be, but I'm not sure, lol.

While it is more compact than some other tower coolers, it does have six heatpipes, which seems decent for your purposes. If you restrict the new CPU to similar power levels as your former, you should be good anyway. I don't know though whether Ryzen 5000 has got worse thermal hotspots than Ryzen 3000 (at same power level) — manufacturing process node is the same, but core layout differs. In case that you do get uncomfortably high peak temperatures, you could cap the maximum turbo clock frequency alongside with the power limit.

Assimilator1 said:
And I'm not buying anymore RAM (did you see I have 32 GB already?).

That's of course alright for almost all DC projects in combination with an 8c/16t CPU. As I said, more RAM would only be a consideration if you exceeded your budget and went with 5900X/5950X _and_ wanted to run certain RAM hungry projects on as many CPU threads as possible.

BTW, my own main PC still is a 4c/8t Haswell Xeon E3 with 32 GB RAM, which is the maximum amount supported by this platform. I do want to replace it with something newer eventually, but not so much because of the old CPU, but more so because of the RAM limit. And that's not because of the DC projects which I run on it, but because of the X11 display service leaking memory when having long uptimes (current OS uptime is 205 days, X11 uptime is probably less) plus one or another bloated desktop software.

Assimilator1 · Oct 27, 2024

Good to know about my cooler , I'll keep an eye on temps whilst running Asteroids (still seems to be the hardest on CPUs!) and after gaming.
I'll probably report back here.

Re RAM, ah ok, so with 12 Cores or more, even 32 GB isn't enough on some projects, that's nuts! lol

Question Zen2 vs Zen3 improvements for DC?

Elite Member

Elite Member

Elite Member

Senior member

Elite Member

Senior member

Elite Member

Senior member

Elite Member

Elite Member

Platinum Member

Elite Member

Elite Member

Elite Member

Elite Member

Senior member

Elite Member

Elite Member

Elite Member

Platinum Member

Elite Member

Elite Member

Elite Member

Elite Member

Elite Member