Athlon II (Regor) In the house! - UPDATED -Underclking+Power consump
** UPDATED AGAN with UNderclocking an Power consumption.- see below**
Picked one up a Athlong II X2 250 today.
Doing a full review on the chip atm that i'm going to post on another site,
I'll post direct Windsor K8 CLOCK to CLOCK comparisons for us curious folk, as well (will be the FIRST time we can actually do an apples to apples comparo of the two Uarchs [K8 vs K10] without differing cache structures) so i'm looking fwd to getting into that (nerd)
Anyway thought I'd post some stocko performance, and info for now :
Specs.
TDP: 65w (vs 80w for the PHII X2's)
Stk clock 3.0Ghz
Stk Vcore 1.35v
Cache: 2x 1MB L2 no L3
NB clock: 2000Mhz
Thoughts:
* New 65w Heatsink is TINY.. pic's coming when I find my god damn card reader.
* Despite, runs very cool @ stock. with Fan speed increase minimal at full load
*Super pi performance appears about on par with or slightly faster than old Agena.. not sure about the others.
***CPU-Z needs an update Mr CPU-Z man (can't remember his nick )
EDIT.. Added Die Shot's
Estimated Die Size ~110mm2
UPDATED:
out of box stock/stock comparison with its predecassesors..
Note, NOT ane exact uARCH comparison, that will be RE-Run at 2.8ghz +2.8ghz sync NB
OK, so the purpose of this is to finally take a look at what core level improvments were made in the evolution from K8 to K10. Since barcelona we've had a different Cache hierarchy (wiht 2MB + of L3 cache), and independtently clocked northbridge.
Fot this test I've matched the two CPU's clock speeds, and memory speeds to (finally) bring a comparison between the two.
Of note though is the following:
this uARCh is unofficially known as K10.5. In this revamp of the core, certain applications, most evident in POV RAY gained substantial extra performance since Phenom 1 /Barcelona, so this is obviously included in the results.
secondly. the K8 runs only Ganged memory mode. This CPU or even the MOBO for whatever reason is unable to run in ganged mode, so all tests had to be run in Unganged.
apart from that though, i've done my very best to match everything else. CPU's were swapped over and all RAM timings and clock speeds verified 100%. As follows:
* Clockspeed 200 x 14 (2800Mhz
* NB speed (K10) 200x14 (2800Mhz)
* DDR2: 400Mhz (standard divider for the K8, 1:2 multi for K10)
- timings 4-4-4-15 2T. With identical subtimings.
Hypertransport: stock for K8. downclocked to 5x (1000Mhz) for Regor.
All settings verified in CPU-Z (i forgot to take regor's ss, will post later)
RESULTS:
Here I've split the % advantage Averages into catagories.. my reasoning as follows:
1st Average: no weighting of results. A plain average.
2nd Average: Mandrelbot removed from the avg..
Mandelbot benchmarks have recieved a massive increase in performance with K10. It may be just SSE enhancments, but it's a huge improvment. As its no indicative of overall performance i've classed it as a very large bottleneck improvment, or optimisation.
3rd average:
There are several other benchmarks that gained a huge boost from K10 . It's perfectly valid to keep these results in the average, however the majoirty of benchmarks show lesser improvements. I've made the cutoff at 30%. this is purely to segregate legacy/ commonplace code improvements.
Please Discuss!
UPDATE: UNDERCLOCKING AND POWER CONSUMPTION
Notes etc:
Power measured @ the wall. PSU is a 550w S12+ with PFC, so power readings should be reasonably accurate..
NO CPU reading taken as a reference, obviously it's not to be taken TOO seriously as the memory, chipset at the least will probably be in a lower power state with no CPU, but it at least gives a worst case idea.
Last edited by mAJORD; 06-08-2009 at 04:29 AM.
Reason: updated with k8 vs k10 apples to apples comparison
Doing a full review on the chip atm that i'm going to post on another site,
I'll post direct Windsor K8 CLOCK to CLOCK comparisons for us curious folk, as well (will be the FIRST time we can actually do an apples to apples comparo of the two Uarchs [K8 vs K10] without differing cache structures) so i'm looking fwd to getting into that (nerd)
Looking forward to your report.
Finally an energy-efficient offering from AMD, hopefully it'll match-up nicely with E5x00's in performance-per-watt department.
Saw the RAM timings and knew it was a gigabyte board, didn't even have to read anything
what about some overclocking? does it do 4 Ghz easily?
LOL Indeed.. typical wtf timings I get at default with my Ballistix.. go go gigabyte.
I dunno about 4 gig, but with the miniscule stock heatsink (it's seriously tiny) it's doing 3.4 on stk volts benchmark stable, reported (belivable) 40c load temp.
Wow, that cooler is tiny like the dual core Xeon HSFs. I bet if you put that massive Phenom II stock HSF on there it would make a big difference for overclocking. I see 4+Ghz on air prime stable with that HSF.
Hail fellow warrior albeit a surat Mercenary. I Hail to you from the Clans, Ghost Bear that is (Yes freebirth we still do and shall always view mercenaries with great disdain!) I have long been an honorable warrior of the mighty Warden Clan Ghost Bear the honorable Bekker surname. I salute your tenacity to show your freebirth sibkin their ignorance!
Wow, that cooler is tiny like the dual core Xeon HSFs. I bet if you put that massive Phenom II stock HSF on there it would make a big difference for overclocking. I see 4+Ghz on air prime stable with that HSF.
The Phenom I/Phenom II coolers are godly on 65-89W TDP CPUs.
This time when I am getting both a Athlon II x2 and Phenom II x2 I will even bother to try out the stock heatsink, something I've never done before
__________________ Nvidia GT300 yields are under 2% CPU: Phenom II X2 550BE @ X4 B50 Clock: 3600MHz 1.4v 24/7 crunching stable Memory: 2x2GB 1333MHz 6-6-6-18-24-1T Motherboard: MSI 790FX-GD70 GPU: ASUS HD5850 1GB AMD Forum
Quick & dirty. Opteron 170 @ 2.8GHz on Win'7. 31.625s.
EDIT: 3GHz = 30.593s. As stated, it's quick & dirty. Thought I'd do it for those who didn't have the patiance for the "Extreme" users to do a proper job.
__________________ PC1 AMD Phenom II x6 1090T BE @4300 MHz 2800 MHz NB 1.479V 24/7, ASUS Crosshair IV Formula, 2x2GB A-DAta 2000x 1600MHz+ cl7, Xigmatek 1283DK+ultra Kaze, ASUS HD5770 1GB DDR5, Seasonic 650W, CM HAF PC2: AMD Phenom II x4 965 BE C3 3.4GHz@3950MHz with 1.44V or at 4032MHZ+2828 MHz NB with 1.52-1.533V stable, CPU3: AMD Phenom II X4 955 BE 3.2@3.895GHz+2665 MHz NB 1.485V with new BIOS, or 3.915GHz 2678MHz NB 1.5V = 24/7 and lower voltage (max 4264Hz with 1.485V and 4386 MHz 1.55V) CM Hyper 212, Gigabyte MA790X-UD4 F3 bios, 2x2GB Kingston Hyper X 1066MHz cl 5, 2x320GB HDD,ASUS HD5770 1024MB, saphire HD3870 512MB (1950pro 256Mb 610/1480MHz),Seasonic 500W Aspire X Cruiser, A4tech keyboard+mouse Razer deathadder ~ 4386MHz AIR x4 955
4570 MHz AIR x4 965
AMD X6 1090T BE on the way...next chip x4 960T Zosma
Me as similar nerd to you mAJORD can't wait for full review!
Don't forget to look at the memory timings/speed influence on performance! I have a feeling it might be quite important without L3 cache for some applications!
When you get bored with OCing try also some undervolting/underclocking of this chip. I really want to get it for my HTPC, because yesterday I found a matroska Hi-Def movie my 4450 stock have problem with
__________________
RiG1: Phenom II 1090T BE 1012APMW (45nm@4000MHz+Turbo 4200MHz 64bit), ASUS Crosshair IV Formula (RD890FX), 2x2GB 2x2 G.Skill Ripjaws PC16000 GBRH, HD5870 1GB, 2x1TB Samsung F1 RAID0 + 750GB Samsung F1 + 400GB Samsung T133 + 30GB OCZ Solid SSD, Hiper 4M880
RiG2: HTPC AMD Athlon II 250, 2GB Crucial Ballistix D9, ASUS M3A78-EM, ASUS HD4670 512MB, 640GB Samsung 7200, LG Blu-ray Recorder, Thermaltake BACH
Certainly, I have that lined up.. I've completed an initial comparsion aimed at looking at difference between 6000+ and the 250 from a consumer pt of view. What I can't decide on though is NB frequency for an arcitectural comparison. With K8 there is no seperate NB clock domain. What frequency does it really run at? I've confirmed that despite NO L3 cache, performance does increase a small amount with NB overclocking. What is happening here?
These are the questions, So I'm not 100% sure on how to tackle that.
Anyway, I've Popped the lid on this beast.. so enjoy die size comparisons for now, and confirmation it is a brand new mask.
Brisbane: measured 8.5x 14.8 =125.8. Actual die size 126mm2
the nb clock is the IMC clock, so I think it's expected to affect the performance.. in the k8 I think it runs on "full speed" same clock as the cores!?
Perhaps the issue with high NB clocks on Deneb/Agena is related to the L3 itself, and the 2Ghz clock on this is to keep TDP's down, like with griffin, since performance delta is pretty insignificant over most benchmarks.
I have tried 2.9ghz clocks on this and they were fine at higher CPU-NB Voltage.. which may just be no higher than stock Vcore since by defauly NB voltage is significantly lower (on deneb)
Need the Datasheet for this CPU to confirm, as there's no indication in my Mobo's bios (voltage settings are an "offeset" from stock.. damn gigabyte)
Saw the RAM timings and knew it was a gigabyte board, didn't even have to read anything
lets see some tests against some of the K8 chips!
what about some overclocking? does it do 4 Ghz easily?
I finally threw my Ph II 940 in my machine yesterday (replacing an Windsor X2 6000+) and on the DFI board the 1066 Ram setting defaulted to similar 5-7-7-24 2T timings. heh.
mAJORD, keep up the good work; do you happen to have any Windsor X2 6000+ results at 250 x 12, DDR800 or 300 x 10, DDR667? Or are you going to run all the comparison benches for the K8 fresh?
)
Samsung LCD TV 
Linear Mode
