What about Socket 754 w/ Sempron 64

ronrem · Post by **ronrem** » Sun Sep 11, 2005 5:07 pm

Since AMD has not offered a 2600 econo-Venice 939 the Semprons do have a niche-being relatively cool in the price-power bracket. The midrange Semps don't get some of the cutting edge goodies,the high end Semps do have many of the Venice's features-but without much price gap. The 3400 has a faster clock-less cache than a Venice 3000-cost is real close

I see the Venice/939/NF4 combo as a helluva sweet mix of power,coolness,quiet-plus all that cutting edge stuff built into most NF4 boards-at a price diff that really is pretty mild. IF someone has a high end AGP vid card,a tight budget, no real need for a bit more horsepower,SATA2 etc,saving a few bucks makes sense

Hifriday · Post by **Hifriday** » Mon Sep 12, 2005 6:54 am

I setup a DFI K8M800-MLVF (s754 matx with onboard video+firewire) and it does have FID, VID settings in the BIOS with voltages down to 0.8v. Unfortunately I am unable to get any voltage or temperature readings off this board (BIOS or OS). According to DFI forums there are no sensors on this board!! Also fans are not controllable by BIOS nor Speedfan. Unfortunately this does not make a good budget quiet mb.

The MSI K8MM-V is a better option IMO although it lacks firewire and voltage cannot be set via BIOS, it does have fan control either by BIOS or Speedfan, and comes with temp and voltage sensors (shouldn't all mbs have this?). Also at $60 it is slightly cheaper than the DFI at $70, overall quality of parts, manual, etc. also goes in favor of the MSI board (Winguy - BTW my BIOS is v1.2).
Too bad MSI K8MM-ILSR (with firewire) is not available in my area.

For the DFI, I did undervolt the VID in BIOS, but as there are no sensors I had no way to verify this. However the new VID setting, did show up in CrystalCPUID as the current setting. Oddly though I noticed the minimum VID in CrystalCPUID was 1.1v regardless if I set the VID lower in BIOS. This same minimum also occurred on the MSI board. Maybe this is related to the Via K8M800 chipset as others have posted getting the Semprons down to 0.8v?

winguy · Post by **winguy** » Mon Sep 12, 2005 7:19 am

There's still MSI K8MM3-V. There're minor differences, the most noticeable being firewire.

=assassin= · Post by **=assassin=** » Thu Sep 15, 2005 1:47 am

Just for reference, there are quite a few NF4x chipsets for S754, so Socket 754 an still provide users with features such as PCI-Express along with Cool&Quiet on the Semprons 3000+ and higher, and of course 64 bit functionality. For the ultra bargain buyers who don't need the power but do want the quiet and coolness, this is a good solution. I am planning to build a Sempron 64 system, and from the information gathered from this forum and elsewhere, a Sempron 64 3000+ system with an NF4x S754 motherboard gives you all the features you'd need (SSE3 too!), and in my case would work out around over £60 cheaper than a Venice 3000+ S939 system, which on a sempron system I'd priced up at just over £400 is quite a large amount - the extra cost coming from around £45 for the CPU, £12 for the motherboard (the S754 NF4x boards work out about this much cheaper than the S939 NF4x boards) and around £3 for RAM (as two sticks are needed for Dual-Channel on S939, as opposed to one for the single channel S754). As I don't have a great deal of cash, over £60 is alot to me, and so getting the same features, as well as still an opportunity for some upgrade in the future seems like a good deal to me as a budget buyer.

All I've got to do now is save up for for my planned budget system now... :/

ratchetcat · Post by **ratchetcat** » Thu Sep 15, 2005 5:28 am

Greetings, all -- first time poster, long time lurker checking in here...

I just picked up a Sempron 64 2600+ and Abit KV-80 for a low-cost, low-heat Debian-based box. I based my choices pretty heavily on the comments in this thread. Thanks to everyone for the insightful discussion. The Sempron is running cooler than I anticipated (32C idle, according to the chipset) and the absence of a hot chipset, video card, and their associated noise make a huge difference in my desire to use the computer. (I have a feeling my 939/NF4 system is going to sit idle until winter now...when I can use the heat!)

Since it's been touched on a few times here: The Sempron is a really nice solution if you're on a tight budget, and especially if you're coming from socket A. I'm using the same PC2700 memory that was in my old system, and the board has integrated video (which is adequate for most stuff). The chipset also appears to be very well supported in the Kubuntu distro (and others, I'm sure). The video, SATA, and network interface all worked from the install.

Major drawbacks with the KV-80: The BIOS lacks any undervolting or multiplier adjustment capability that I can discover. It looks like those are entirely dependent on the CPU -- a CnQ-enabled Sempron would be allowed to fall back to lower speeds/voltages, but I can't tell whether it would be controllable from the BIOS.

Hifriday · Post by **Hifriday** » Thu Sep 15, 2005 8:13 am

Ratchetcat

WELCOME TO SPCR ! ! !
(to posting anyways...)

Not familiar with the Abit mb, however if it supports CnQ (plus you have a CnQ CPU) you should be able to control voltage/multiplier via software with CrystalCPUID even if you don't have these settings in BIOS.

However if you are running a 2600+, even at stock 1.6ghz 1.4v there probably will not be a significant temp or power consumption difference compared to 1.0ghz 1.1v at idle. I say 1.1v as seems with these Rev E cores, that is the lowest voltage I can set, either via BIOS or using CrystalCPUID. Oddly even with a S939 Venice core (also Rev E) on a NF4 mb, it seems I can't set VID lower than 1.1v.

A while back I did a rough comparison with a Sempron 64 3000+ at 1.8ghz 1.4v compared to 0.8ghz 1.1v and the difference at idle was negligible. However during load, maybe the difference from undervolting would be more significant, but for light CPU use it shouldn't be a concern.

winguy · Post by **winguy** » Thu Sep 15, 2005 9:14 am

Hifriday, can a Sempron E6 2500+/2600+/2800+ be undervolted using CrystalCPUID on the MSI K8MM-V ?

ratchetcat · Post by **ratchetcat** » Fri Sep 16, 2005 5:30 am

Thanks for the welcome, Hifriday!

I appreciate the summary on the effects of the voltages on the temps for these Semprons. That makes me feel better about CnQ/undervolting not being available with this chip and motherboard.

Like you, I tried undervolting a 3000+ Venice on an NF4. While 1.1V worked well (I guess that's where it hits the 22W TDP?) and the chip temp did fall a bit, the overall heat production of the system remained about the same (which is a little too much in a well-insulated, but not well air-conditioned room in the summer...). The Via/Sempron combo is MUCH better in this regard.

Hifriday · Post by **Hifriday** » Fri Sep 16, 2005 7:14 pm

Winguy
As 2800+ and lower Semprons do not support CnQ, VID/FID is locked. The MSI board will not allow you to change VID via CrystalCPUID, and I think that applies to any mb. However undervolting via BIOS maybe possible. When I get a chance, I will try to hookup a watt meter to verify if my DFI K8M800-MLVF VID settings in BIOS actually affect the vcore or not.

ratchetcat
I suspect if you clocked the 3000+ Venice down to Sempron speeds that the CPU heat/power consumption should be pretter close as the cores are very similar (Venice has the extra cache and dual-channel memory support). However regarding the mb, NF4 is another matter. You would definitely need very good airflow to run it in a silent rig.

Hifriday · Post by **Hifriday** » Tue Sep 20, 2005 3:28 am

The DFI mb BIOS VID settings do in fact control the vcore for the Sempron. As there are no sensors on this mb, I indirectly verified this by reading the AC power draw. However again the lowest VID seems to be 1.1volts even though BIOS will allow settings down to 0.8v. The power draw between 0.8 to 1.1v remains exactly the same (suggesting VID is not going lower than 1.1v). It seems my Sempron 64 3000+ chip is stable at stock speed undervolted to 1.1v.

I compared various FID/VID settings and came up with the following approx AC power draws measured using a clamp meter. This was using an Antec Aria 300W PSU, so I guess roughly at 65-70% efficiency for the actual DC draw.

IDLE IN WINDOWS
4x 1.1v = 47w
9x 1.1v = 50w
9x 1.4v = 54w

LOAD UNDER PRIME95
4x 1.1v = 54w
4x 1.4v = 63w
7x 1.1v = 61w
7x 1.4v = 72w
9x 1.1v = 63w
9x 1.4v = 77w

At idle, the difference between lowest FID/VID settings and the 3000+ stock settings is only 7w. For lower clocked Semprons the difference will be even less. Under load, the difference is larger at 23w. However if we compare the effects of just undervolting, the difference is no more than 14w. Not sure how significant this extra power is for a fanless rig, but I suspect for a fanned system the effects would be minor.

One thing odd about the DFI board in suspend mode and power off mode, the system was drawing around 10watts more than the MSI board?!

dukla2000 · Post by **dukla2000** » Mon Oct 17, 2005 12:47 pm

I blame this thread for my latest upgrade. Went out and bought an E6 stepping Sempron 3000+ and an Abit KV8-Pro (cause I found one cheap

). Have been installing and overclocking today, currently running
267 * 9 = 2403 MHz
Stock VCore (CPU-Z, BIOS & Speedfan reporting 1.42V. Dont want to go higher as I want to use CnQ.)
Have been running short bits of Prime95 while stepping through the overclocks as well as MemTest86+ so am pretty confident in my stability: a 24 hour run has just started. And then on to 64-bit Linux.

The punchline: total AC power draw under Prime95 is 86W. My previous XP2700 was running 122W (AC) under Prime95! (Same hsf, memory, disk, optical, vidcard etc.) Glad I ignored the folklore that says you need more than 20A on the +12V (or 500W psu) for socket 754

ceraf · Post by **ceraf** » Mon Oct 17, 2005 12:53 pm

^wow, nice setup....I'm hoping to have a similar setup for my next system.

boy, 86W on load =)

ronrem · Post by **ronrem** » Fri Oct 21, 2005 7:05 pm

dukla2000 wrote:I blame this thread for my latest upgrade. Went out and bought an E6 stepping Sempron 3000+ and an Abit KV8-Pro (cause I found one cheap ). Have been installing and overclocking today, currently running
267 * 9 = 2403 MHz
Stock VCore (CPU-Z, BIOS & Speedfan reporting 1.42V. Dont want to go higher as I want to use CnQ.)
Have been running short bits of Prime95 while stepping through the overclocks as well as MemTest86+ so am pretty confident in my stability: a 24 hour run has just started. And then on to 64-bit Linux.

The punchline: total AC power draw under Prime95 is 86W. My previous XP2700 was running 122W (AC) under Prime95! (Same hsf, memory, disk, optical, vidcard etc.) Glad I ignored the folklore that says you need more than 20A on the +12V (or 500W psu) for socket 754

VERY interesting,a pretty heavy OC and she's stable with only a 5v PSU fan for case airflow,plus and undervolted Nexus on the CPU. Anyone have numbers on stock-clock Semp vs Venice as to power draw? Sounds like these "stripped down" Venices have less heat out,less power in.

I have had a general idea of an ideal system based on Venice 3000 and NF4,but I think I will take a fresh look at a Sempy-754 alternative.

dukla2000 · Post by **dukla2000** » Sat Oct 22, 2005 12:52 am

ronrem wrote:VERY interesting,a pretty heavy OC and she's stable with only a 5v PSU fan for case airflow,plus and undervolted Nexus on the CPU. Anyone have numbers on stock-clock Semp vs Venice as to power draw? Sounds like these "stripped down" Venices have less heat out,less power in.

I have had a general idea of an ideal system based on Venice 3000 and NF4,but I think I will take a fresh look at a Sempy-754 alternative.

I also have some 'extra' caseflow as the hsf is a tower/heatpipe job, and the Nexus is aimed out the back where there are 2*80mm fan cutouts (above the I/O panel). But no proper ducting. Also the front of my case has a 120mm fan cutout, so the case static pressure must be pretty close to 0.

But at such low power case airflow is almost a non-issue. While folding I am running 80W AC. Assuming 75% psu efficiency I am dissipating 20W inside the psu itself which a 5V Papst should cope with even if 5 of the blades fall off. 60W around the cpu and the rest of the system is also getting close to noise-level rather than a problem. BTW: under CnQ the system AC draw is around 50W. So IMHO this Sempron is running nowhere close to the 62W quoted TDP. 40W peak would be my guess. What is the VCore for the Venice? Also note there are NF4 s754 boards out already.

However OC results for s754 Semprons seem to be very variable: I haven't seen any good data collections that may try help indicate which batches or what is causing the variations: some 2500+ and 2600+ are hitting 2.4GHz comfortably on air (with a mobo that has v high stable HTT), whereas some 3000+/3100+ seem to be struggling much past 2.2GHz. My gut feel is some issues are related to the BIOS implementations of CnQ: some mobos OC better with CnQ disabled.

I am still very happy. 267HTT remains stable (passed Prime for 24 hours, have installed Linux and been folding for 60 hours now). Still early days but indications are folding performance is twice that of my other box that is running an Athlon XP2400+, at only 2/3 of the AC Wattage. The one hiccup is Linux has a problem with CnQ @ 267 HTT (that I posted in another thread) - pretty confident that is a parameter though. As I am likely to leave this box folding CnQ is less relevant.

KorruptioN · Post by **KorruptioN** » Sat Oct 22, 2005 2:52 pm

I built a Sempron64 machine for a friend of mine about two months ago.

Antec SLK3000B + SmartPower 2.0 400W (rear 120mm Tricool set to medium)
Asus K8V-X (K8N wasn't in stock)
Sempron 2800+ (RevE, 256KB L2.. the higher-rated RevE 3000+ only has 128KB L2, not good)
2x 256MB Corsair ValueSelect PC3200 (2.5-3-3-7)
Gigabyte Radeon 9600 Pro 256MB, passively-cooled
Seagate 7200.7 80GB ATA-100

Decently quiet machine, the loudest component is the AMD thermal solution. Runs very cool too, from what AsusProbe was telling me, the CPU was idling in the high-20°C range, which is fantastic. Cool & Quiet did not work... the complained about the CPU not being compatible with the driver software or something along those lines. I was pretty impressed with the temperature output... the rear fan in the SmartPower 2.0 PSU never activated.

dukla2000 · Post by **dukla2000** » Sat Oct 22, 2005 3:23 pm

Nice system, and inexpensive I guess.

KorruptioN wrote:Sempron 2800+ (RevE, 256KB L2.. the higher-rated RevE 3000+ only has 128KB L2, not good)

Well it depends what you are doing with it: apart from gaming X-bit labs struggled to find a benchmark where L2 made a meaningful difference.

KorruptioN wrote:Cool & Quiet did not work...

Cool n Quiet is not enabled on the Sempron 2800+ and below - only on the 3000+ and up

wainwra · Post by **wainwra** » Sun Oct 23, 2005 1:44 am

cute baby!

winguy · Post by **winguy** » Sun Oct 23, 2005 2:25 am

dukla2000 wrote:Nice system, and inexpensive I guess.
KorruptioN wrote:Sempron 2800+ (RevE, 256KB L2.. the higher-rated RevE 3000+ only has 128KB L2, not good)
Well it depends what you are doing with it: apart from gaming X-bit labs struggled to find a benchmark where L2 made a meaningful difference.
KorruptioN wrote:Cool & Quiet did not work...
Cool n Quiet is not enabled on the Sempron 2800+ and below - only on the 3000+ and up

Conclusion: should have gotten the 3000+.

KorruptioN · Post by **KorruptioN** » Sun Oct 23, 2005 6:23 am

winguy wrote:
dukla2000 wrote:Nice system, and inexpensive I guess.
KorruptioN wrote:Sempron 2800+ (RevE, 256KB L2.. the higher-rated RevE 3000+ only has 128KB L2, not good)
Well it depends what you are doing with it: apart from gaming X-bit labs struggled to find a benchmark where L2 made a meaningful difference.
KorruptioN wrote:Cool & Quiet did not work...
Cool n Quiet is not enabled on the Sempron 2800+ and below - only on the 3000+ and up
Conclusion: should have gotten the 3000+.

I would prefer more cache over more clockspeed. Might be a placebo effect, but you can gain clockspeed, but not cache.

mb2 · Post by **mb2** » Sun Jan 29, 2006 8:22 am

i'd just like to highlight a bit of an issue in the (otherwise very good) xbit labs review..

www.xbitlabs.com/articles/cpu/display/sempron-2600_5.html wrote:Note the curious fact that the Sempron with 128KB of cache is faster in Quake 3 and Unreal Tournament 2004 than the analogous Sempron with 256KB L2 cache! This paradox is easily explained: the sample of the processor with 128KB of cache memory was based on the 90nm stepping D0 core, while the rest of the participating CPUs were based on 130nm CG stepping cores. The D0 core features an improved integrated memory controller which sometimes provides certain performance gains. This gain compensates for the reduced cache with interest in Quake 3 and Unreal Tournament 2004!

so, comparing simliar 90nm processors would (may) put the 128kb cached cpus relatively slower.
they seem to have ignored this when the 128kb cached 2600+ doesn't loose.
of course the improvements they made may not effect the other situations very much.. who knows?
still not a huge difference anyway i guess

=assassin= · Post by **=assassin=** » Mon Jan 30, 2006 3:43 am

Ok - I inputted some data from a tomshardware CPU review:

The following are average game performance in frames per second for Sempron 64 models:

higher is better (fps):

2500+ 109.72
2600+ 116.73
2800+ 120.53
3000+ 125.98
3100+ 130.12
3300+ 133.97
3400+ 139.07

So for the processors with the same clock speed, L2 cache increase improves performance by 4-6fps. Now for general applications and encoding, the average time of each of the tests they constructed:

lower is better (time in seconds):

2500+ 456
2600+ 404.3
2800+ 400.5
3000+ 361.5
3100+ 356.42
3300+ 334.67
3400+ 323.5

Overall, for gaming, L2 becomes more important the higher the frequency gets, as the lead of a higher frequency part with 128kb cache compared to a lower frequency part with 256kb cache lessens. For example - 2500+ to 2600+ gap is 7 fps, 2800+ to 3000+ gap is 5.4 and 3100+ to 3300+ gap is only 3.8fps - contrast that from raising L2 cache: 2600+ to 2800+ is 3.8fps, 3000+ to 3100+ is 4.1fps, and then 3300+ to 3400+ is 5.1 fps, so L2 cache makes more of a difference when clock speeds are higher, as expected really.

In terms of general applications and encoding etc, the frequencies are very important, with very little difference between models with 128kb and 256kb of cache at the same frequency. Again, it's not until the 3300+ to 3400+ move (2.0Ghz) that L2 starts to make more of a difference.

I'll try and find the actual link of their review a little later - just saw this thread and I already had a spreadsheet of data I'd created, so thought I'd post it.

stupid · Post by **stupid** » Mon Jan 30, 2006 7:55 am

I was comparing some video and gaming benchmarks of the Sempron 64 3400+ for someone recently. Based on those benchmarks, the S64 performs similarly to the old Athlon XP 2600+ when converting video to DivX. On the gaming side of things the S64 seems to perform marginally better than the old Athlon XP 3200+.

These are broad stroke observations, but in the end the Sempron is crap in my opinion. It's only worth it to upgrade to the fastest Sempron 64 if the current system is slower than an Athlon XP 2200+.

The Sempron 64 strong points over Athlon XP:
1) Lower power consumption
2) Support for 64 bit OS

=assassin= · Post by **=assassin=** » Mon Jan 30, 2006 9:33 am

I disagree, most Sempron 64's are much better than the Athlon XP's in games - a 2800+ S64 will give a 3200+ XP a run for it's money in games, although granted, in general apps the 3000+/3100+ Semprons are pretty close to an XP3200+, so all in all, considering the less cache, and that the XP3200+ is 2.2Ghz - 400mhz higher than the Sempron 3000/3100+, the Semprons are pretty efficient and quick. Of course it wouldn't be wise to start swapping to an equivalent rated CPU or only a little higher as it's not worth the money.