View Single Post
  #1 (permalink)  
Old October 14, 2007, 03:10 PM
Lithotech's Avatar
Lithotech Lithotech is offline
MVP
 
Join Date: Apr 2007
Location: North Vancouver Rain Belt
Posts: 409
Default In Progress: AMD X2-5000 Black (unlocked multiplier)

Testing in Progress: AMD X2-5000 Black Edition (unlocked multiplier)

Preamble

This is not a formal review in any sense, and rather is a journal of my experience and results while exprimenting with one of AMD's new Black Edition X2-5000s with an unlocked multiplier. This is the first unlocked CPU I have ever owned, so it is as much a learning experience for me as an exploration of it's capabilities in various situations, settings and motherboards.

Introduction

Does it seem that AMD slashing prices isn't enough to keep sales rolling? AMD are now shipping mainstream CPUs with unlocked multipliers! They are called Black Edition, come in a BOX (PIB, not OEM), but do not include a heatsink fan.

So far, there are two such parts, the X2-6400 Black Edition @ 3.2ghz Windsor core, and the X2-5000 Black Edition 2.6ghz Brisbane core.

AMD Athlon 64 X2 6400+ Dual Core Processor Socket AM2 3.2GHZ Unlocked Black Box No HSF Windsor
http://www.ncix.com:80/products/inde...D&promoid=1058

AMD Athlon 64 X2 5000+ Dual Core Processor Socket AM2 2.6GHZ Unlocked Black Box No HSF Brisbane
http://www.ncix.com:80/products/inde...D&promoid=1058

Not many AMD CPUs will clock much farther than 3.2ghz without a performance motherboard, good ram and extreme or exotic cooling. So the X2-6400 would be no fun at all! Unlocked multi that you can't move up anyway? Nup.

The X2-5000 Black Edition looks to be much more interesting, with more headroom for increasing the multi, and damn near half the price of the X2-6400. I was initially sceptical of the X2-5000 Black, assuming they were neutered 6400's that somehow didn't make the cut, and with half of their L2 cache disabled. But after learning the X2-5000 Blacks are Brisbane cores, I feel confident they are cherry picked just like the 6400's. These are 65nm process, and 65W thermal design just like some of the other recent Brisbanes.

Here's a good link to a review/press-release info:

X-bit labs - AMD Unveils Second “Black Edition” Microprocessor for Enthusiasts.

An excerpt from above:

Quote:
The AMD Athlon 64 X2 processor 5000+ Black Edition is an affordably-priced, dual-core processor targeted to the consumer channel, boasting a customizable clock multiplier for tunable performance, enabling mainstream end-users to reap the benefits of setting their own speed limits,” a statement by the world’s largest maker of x86 central processing units maker reads.
The description on the packaging says exactly: Clock Multiplier Control.

My X2-5000 and Setup

I picked up an X2-5000 Black Edition shortly after they began shipping, when I noticed NCIX's real-time stock showing these as quickly disappearing off their shelves. I didn't want to miss out on this first shipment, as the potential for experimentation and testing was too alluring and I wanted one of these babies for this weekend.

I always take photos before starting work on anything like this, so here's the eye candy in all it's glory. I was particularly suprised to see some stickers included with the case badge and certification papers with this processor, they are attractive designs and I am sure will look very nice on some cases. I am a Scorpio, so I am especially attracted to the scorpion sticker. Anyway, it was a pleasant suprise, did the FX chips come with these too?

X2-5000 Black Edition Pics













I decided to first test it in one of the cheaper AM2 boards I have on hand, and then move on to a slightly better mATX board, and finally a performance full size ATX board. I have another mATX nForce 6150/430, an ABit NF-M2 nView, which has more CPU and RAM tuning BIOS options than many full size motherboards. If I like this chip and hang onto it, the ABit will end up being it's permenant home as an HTPC for my living room. The last AM2 board I have is a full-size ATX cheapo-but-really-decent board by BioStar, a TForce 550. It has an overwhelming number of options in the BIOS for memory alone, and I am still very much learning it. I also have an Asus M2NPV-VM which doesn't overclock at all, but I'm unsure if I'll find time to test this CPU in it.


First up:

Quote:
ASRock AM2NF6G-VSTA mATX w/ onboard nForce 6100 (onboard vid) and nF405 chipset.
These aren't bad boards at all, with plenty of settings to tune an overclock and the CPU and RAM. While the RAM voltage is limited to Low, Normal, and High; the BIOS version P2.30 I'm using has a complete set of CPU voltage (VCore) with plenty of granularity. Before cranking up any juice, lets just see how far she goes on 1.3 volts. The point of this excersise (or experiment, and the cheap mobos), is to see how far the CPU can go without any voltage adjustments, and using only the multiplier. Since one of the biggest benefits of an unlocked multi is the ability to speed the cpu up without overclocking the system -- perfect for all those cheap boards that don't overclock at all (or very poorly) -- this is what I'll shoot for first.

The ram I'm using (in this mobo at least), is some mainstream Mushkin EM5400 DDR2-667 5-5-5-15. They are quite stable at DDR2-800, but won't go much farther. It'll be interesting to see how the various multipliers affect the actual memory speed once the AMD mem controller gets finished with it. I'm not trying to smash any records here, the point is to see how she does with mainstream and relaticely tame parts.

I also picked up a Scythe Samurai Z Rev. B heatsink fan which was on sale at NCIX recently. Initially purchased to test it's performance and also see how it fits in my small form factor HTPC (an Ultra Micro-Fly), this is a perfect opportunity to get some comparative results for it along side an AMD stock HSF and the X2-5000 Black!

The system is setup open air on a motherboard box with no extra fans actively cooling anything.

Full System Specifications

Quote:
ASRock AM2NF6G-VSTA bios P2.30
X2-5000 Black Brisbane Stepping: CAA9G 0737GAC
Stock AMD HSF & Scythe Samurai Z Rev B
1GB (2 x 512MB) Mushkin EM5400 P/N: 991379A
Onboard video
Western Digital Caviar 80GB PATA
Antec SP2-450watt PSU
XP Pro SP2
After some initial temperature readings using a stock AMD heatsink fan, and then with a Scythe Samurai, my first impressions are that the Samurai is pretty decent, quiet, and small enough to fit in many SFF (small form factor) cases. However, when I removed the stock HSF, I found the goop untouched on most of the IHS. It obviously wasn't making proper contact with the CPU at all. The Scythe has a different footprint, so if the IHS has high corners it won't affect the Scythe as it doesn't make full contact. I'll check the IHS flatness later when I have to clean the goop off next, and will revisit the stock HSF on the first mobo change.







Mini Review - Zalman ZM-STG1 Thermal Paste w/ Applicator Brush

Incidentally, I use Zalman's STG1 thermal goop almost exclusively for testing like this. In some situations, it's possible to even re-use the application because you can smooth it back out with the little brush -- I've done this many times with no ill effects. The application brush makes it extremely easy and fast to apply a thin even film on any core or IHS. The brush leaves tiny groove marks in the application which I believe helps the goop smooth out after a bit of use. Takes about as long as it would to paint a fingernail with nailpolish (yeah, like I know how long that takes!). I used to (still do sometimes) use a dab of AS5 or Ceramic and spread it out by hand using a corner I cut off a credit card as a spatchula. While this is also an excellent method, it is much more time consuming in comparrison. I've also gone with the dab in the middle and squish it down with the heatsink method, and while the goop does eventually spread out, I've found it never completely covers the surface and often spills out in one area or another. I firmly believe one thin even coat on the entire surface is best, and Zalman's product makes this too fricken easy! If you've never tried this goop, I highly recommend it.






Here are the temperatures I've got so far, these are all at stock speed and pretty much auto settings in the AssRock.

Quote:
Stock HSF - 1.3v VCore
Idle in BIOS: 48cpu 33mobo
Idle in OS: 40cpu 25core1 15core2 32mobo
Load in OS: 57cpu 47core1 41core2 34mobo
Quote:
Scythe Samurai HSF - 1.3v VCore
Idle in BIOS: 34cpu 29mobo
Idle in OS: 33cpu 18core1 4core2 28mobo
Load in OS: 40cpu 26core1 13core2 29mobo
Huge difference, the Samurai helping keep the mobo cooler as well. The disparity between the two CPU temperatures isn't too suprising considering the stock HSF wasn't making proper contact. From what I've read on these Samurai, is they are about par with the stock HSF at idle (only much quieter), but really show their worth when under load. The above temperatures verify this to be true, even considering the stock HSF test is pretty much borked. At idle, there is a 7-14 degree drop in temps with the Samurai, but under load the difference jumps to 17 degrees!

Be advised, the Samurai will block the closest ram slot in many motherboards, so keep this in mind if your ram has problems running in far slots. Otherwise, it's a pretty good cooler for the money when you can't fit a Freezer Pro in the case.


Moving the Multiplier

At one point early on, I accidentally set the multi to 16X and saved... thinking I had it on default 13X! I was just looking at how high it would let me go, probably set it to 16X because that's what I'd like out of it, and then saved/exited the BIOS before I realised what I did! The voltages were at defaults! Sheesh!



She booted alrighty! But locked up mid-POST. Ah well, at least I know she'll maybe do a 16X multi with some volts to the CPU, and should get a 15X for sure stable with maybe a bit of extra voltage.

13X Multi @ 2.6ghz

I finished the stock speed benches and started ratcheting up the multi. Already knowing 16X was going to need some voltage, I really only had to test 14X and 15X. I started off with the VCore set to 1.3 volts.

14X Multi @ 2.8ghz

14X @ 2.8ghz booted to memtest without problem, with the ram at DDR667 or DDR800. Got some nice bandwidth numbers at DDR800 because at the 14X multi you'll get the full speed of the ram at 400/800. I'll come back and bench this speed later, for now I wanted to continue increasing the multi as I have owned this chip far too long already without allowing myself to touch the speed yet!

15X Multi @ 3.0ghz

15X @ 3.0ghz also booted to memtest without a burp, again with the ram at DDR667 or DDR800. Numbers were quite different, as the 15X multi gives full ram speed at 333/667, but only 375/750 at 400/800. This can help as a bit of a built in divider, giving a little more headroom to increase the HTT/FSB when they ram is set to 800 speed but really can't do much more than that -- in other words, I can increase the HTT/FSB 25mhz to 225mhz before the ram is actually working at DDR2-800.

Benchmarks

Already knowing the 16X multi will be a voltage challenge, I decided to go ahead and boot Windows with the 15X multi and begin testing. Wondering if it will bluescreen, I watched the loading screen nervously. It hit the desktop just fine, so I popped open memset and cpuz and was just dragging the cpuz window when it happened... BSOD.

Not willing to pull back and bench (/yawn) the 14X 2.8ghz setting just yet, I instead went into the BIOS and increased the vcore to 1.35 in hopes that it'd at least let me look a bit longer. It may not be totally stock VCore, but is still low enough that it isn't technically overvolted.

She passed 32m Super-Pi and currently chugging through dual prime with Orthos... so far error free! Off to bed I went, letting it prime overnight and I'll bench it the next day. In the morning, I found Orthos errored so I again elected to up the VCore, this time to 1.4volts. It's priming again, so the benchmarks will have to wait till later. I promise this time, if it still isn't stable I'll simply concede that for the ASRock AM2NF6G-VSTA simply cannot run this Black Brisbane with a 15X multi at stock voltage, and move on to other tests.

So for now, the benchmarks will have to wait. As things progress, I will edit and add to this post and upload the benchmark screens hopefully tonight.

Summary - ASRock AM2NF6G-VSTA

With 1.3 VCore
13X Multi @ 2.6ghz - no problemo
14X Multi @ 2.8ghz - no problemo
15X Multi @ 3.0ghz - boots to desktop, BSOD while moving cpuz window
16X Multi @ 3.2ghz - only partially POSTs

With 1.35 VCore
15X Multi @ 2.8ghz - passes Super-Pi 32m, fails Orthos dual Prime after 2.5 hours

With 1.40 VCore (maximum in BIOS)
15X Multi @ 2.8ghz - Passes 8+ hours Orthos dual Prime
16X Multi @ 3.0ghz - boots Memtest, boots to desktop, fails Super-Pi and Orthos almost instantly





Scythe Samurai Z Rev. B Pics

Here are some pics of the Scythe Samurai, including some shots of how close it comes to the RAM on the ASRock motherboard, which has plenty of clearance in the first slot or for ram with tall heatspreaders like OCZ Reaper HPC.
























Stock Speed Screenshots
13X Multi
200mhz HTT
2.6ghz @ 1.3v VCore









































Overclock Screenshots
15X Multi
200mhz HTT
3.0ghz @ 1.4v VCore











































.
__________________
eVGA 680i SLI (A1) P31
Intel Q6700 Kentsfield | Scythe Ninja-B | Antec Tri-Cool 80cfm
4GB (4x1024MB) Corsair Dominator 8500
boot: 4x WD Caviar SE16 250gb SATA2 Raid-0
backup: Seagate 500gb 32mb SATA2
eye: eVGA 9800GTX+ SC 512mb 770/2.25ghz
candy: Samsung 940B-HAS | X-Fi XMusic | Pioneer DVR-212D SATA
juice: PC Power & Cooling 750Quad | Cooler Master HAF932
steering: Logitech G15 & G5 | E3D Pro/Momo Wheel | X-530's

Last edited by Lithotech; October 23, 2007 at 01:06 AM. Reason: Added benchmark screenshots
Reply With Quote