Can someone please explain the effects of raising vcore?
I'm a relatively newbie to overclocking but I remember in the past (with Athlon XPs) raising vcore was almost certain death unless you had water cooling or your pc running in a freezer :)
Just got my winchester 3000+ and my MSI K8N MSI Neo2 Plat. Since it's an x-mas gift can put it together until then. Putting it with Corsair XMS DDR 400 memory. I'll let you guys know how it goes.
I dont get it...
1)IS 3500 64 voltage 1.5 Volts?
2)Looks like the decreased the CPu multiplier on 3500 coz by default it is 11? Why? Does it affect ur performance?
Cheaper Ram Altnernatives:
A few other readers have mentioned this, but it seems there hasn't been much of an answer (though a few worthy attempts, notably that the 3200+ is probably a better choice).
The Question: What ram would let us run at the highest FSB for the least $.
Obviously you can just buy the faster stuff (DDR500, or whatever), but there must be sticks that, with looser timings(small cost), can let your processor scream(huge gains) while staying relatively stable.
Anyway, I haven't found any articles about it, but if anyone has, or has some personal experience here, i'm sure we're all eager to hear it.
Err.....i'm really a noob....didn't see my post so i thought that there was a problem, but apparently it doesn't reload page 4, but page one. Feel free to delete the other responses.
great article. what was the core of the 3500+ chip? was it newcastle or winchester for the 90nm? and is there any difference between these two cores? on sites such as newegg and zipzoomfly, the 90nm 3500+ are advertised as winchester on zipzoomfly and neweggs are advertised as newcastles. thanks
#73 -
You normally have to drop HT at higher overclocks, to keep the aggregate somewhere in the 1000HT range. Some boards handle higher HT than others. Since x3 HT was used for the 290x9 benchmarks in the review, I think it should be clear that the lower HT ratio does not adversely affect performance as long as the HT is somewhere around 800 or greater. 290x3 is an HT of 870.
4X HT usually stops working around 260 to 275 (1040 to 1100) on most boards that support 1000HT (5X) and you need to drop to x3. As a side note, none of the 1000HT boards we have tested work well at 2x HT.
One thing I've noticed when overclocking the MSI board with the 3000+ A64 (same as in the review) is that you have to lower the Hyper Transport (HT) from x5 to x3 to get the overclock of 290 FSB, giving an HT speed of 870 mhz instead of the stock 1000 mhz. My cpu/board refuses to run at a HT speed of over 1070.
Is this true of the overclock in the article? Will the decreased HT speed negate the performance benefits of the overclock in any practical areas?
actually, disregard the above. The PQI 1024DP has a higher latency rating, the one you'd need for DDR570 is the 1024DBU, which is $245... stupid dealtime and it's incorrect linking ^^ (yeah, I'll blame it on dealtime)
This ram was reviewed here at anandtech and while not as insanely good as the OCZ EL Platinum, the performance compared to other ram was not too bad. As mentioned in the review, though it has lifetime warranty, the manufacturer is a new name, so the support service is a big unknown.
Furthermore, running the FSB at 285MHz instead of 290MHz will give you an overclock that is 45Mhz lower (2.565GHz) but the ram being 100 dollars cheaper is worth it, in my opinion.
anyone notice the price for these things is going UP after AMD just lowered their price? newegg had them for $199, up from $189 a few days ago, now its $215!!! HELLO, the prices are supposed to go DOWN after AMD loweres the price!
I looks like I am finally getting close to being able to upgrade my athlon xp setup without breaking the bank and get a real performance boost I feel the earlier A64 cpus and mobos lacked. To bad the PCIe problem is still around. I dont want an upgrade that requires a new motherboard with my next GPU purchase.
I currently have an A64 3200 Clawhammer on a VIA K8T800, so OCing is a no go without a working AGP/PCI lock. I have a Coolmax Taurus 400w PSU and OCZ EL Platinum 1GB (2x512) memory (2-3-2-5).
If I upgraded my mobo to the Neo2 and my CPU to the 300+, do you think i could hit 2.6GHz with my current memory and PSU?
Yea, I'm beginning to think I should get a 3200, a conservative OC would be 2.5GHz at 10x250. And DDR500 isnt that hard to get a hold of. Plus it fits in nicely with a 1GHz hypertransport bus.
People who want the best performance and value ought to get the 3200+ instead of the 3000+ as it not only stands a better chance of reaching something like 2610MHz than the 3000+, but it should do so at a slightly lower voltage.
More importantly, the memory with the 3200+ could run 1:1 at 261x10 instead of the 290x9 required by the 3000+. Even though the 3200+ will be a little more expensive than the 3000+, theres a much better chance of finding reasonably priced memory which can run at DDR522 with good timings, than there is of memory that can manage DDR580. At least with the 3200+ you've got the choice of a 9x or 10x multiplier to get the best out of your memory.
I know you used OCZ with Samsung TCCD for this review, but you have said in past reviews that any TCCD will come close to these memory results. That means you can choose Samsung TCCD from Corsair, Mushkin, Kingston, OCZ, Geil, PQI, and probably others. Crucial Ballistix is another choice from your Athlon 64 memory tests. Those concerned about Value ought to shop all the TCCD brands and Crucial Ballistix, because some are cheaper than others.
#62 - The MSI K8N Neo2 can be set to Auto, 100, 133, 166, 200 (DDR200,266,333,400 or 1:2.2:3,5:6,1:1). With these ratios you could run 290 FSB with a 2:3 ratio (266) and memory would be at DDR387, at 5:6 (333) mem would be running at DDR482.
#58- tha major selling point of nForce 4 will be PCIe support, if you don't want PCIe then get an nForce3 instead. The Inquirer claim to have a piccy of MSI's new nForce4 S939 board at
That is presumably the successor of the popular K8N Neo2 Platinum. Its a shame it only has two PCIe x1 sockets and still has four old PCI sockets, I'd have liked to see at least three PCIe x1 and just a couple of PCI for legacy support.
#57 - There were NO tests on Socket 754 processors in this review, since 90nm is only available as 939. This is stated in the review. To see the impact of the new Winchester core and die-shrink on performance we downclocked a 939 .13 CPU to 1.8GHz - the same specs as the 90nm 939. This is clearly stated in bold in the review "We also ran benchmarks of the 130nm processor at Socket 939 3000+ speeds, but these results are theoretical. There is no production 130nm Socket 939 3000+, so these results were just to compare the impact of the die-shrink and Winchester core on performance."
#56 can you post an official Nvidia link stating that. I can't seem to find the official word anywhere. I just figured since you stated that they will not support AGP as a fact you have seen something official.
The 1.8GHz Winchester is faster than the Sempron 3100+ for a few reasons:
1) dual channel support;
2) 512KB L2 cache (versus 256KB on the Sempron);
3) small improvements on the Winchester core.
Not to mention the lack of 64-bit support on the Sempron. But that's not so bad for a chip that's half the price (they cost $100 and $199 on newegg). The price difference is too big, the suggested price for the A64 3000+ is $165, so I expect to pay 150-165 for the OEM version. Newegg's logistics is too good. ;)
#50 - All Athlon 64 processors can be set to lower CPU ratios. Only the FX can be set either lower or higher multipliers. I was testing and had set the 3800+ to a 9 multiplier in BIOS. The 3800+ is at stock a 12 multiplier and runs at 2.4GHz.
To minimize confusion we replaced the 9x capture with a 12x 3800+ screen capture in the review.
#52 is correct - the Sempron PR is based on Celeron and not A64. The Sempron 3100+ is actually a bit slower than a 2800+ Athlon 64. The Sempron is also 32-bit only and does not support 64-bit operation even though it will fit in Socket 754.
#47 - the 3400+ is for Socket 754. The only 90nm chips are for Socket 939 at present. 3500+ is the correct name and you can buy the chip from Monarch Computers for one at http://www.monarchcomputer.com/Merchant2/merchant.... New Egg does have the 3200+ and 3000+ 90nm in stock - but not the 3500+. Look for Core: Winchester and Process: 90nm in the description. I don't have any idea what the 3400+ Socket 939 chip is that is advertised at NewEgg, but it's not Winchester core or 90nm process.
#48 - All Athlon 64 processors can be set to lower CPU ratios. Only the FX can be set either lower or higher multipliers. I was testing and had set the 3800+ to a 9 multiplier in BIOS. The 3800+ is at stock a 12 multiplier and runs at 2.4GHz.
This comment is a response to a lot of post i've seen across different forums.
People are really getting confused on which chip can OC better. I'm hearing people say "the 3000+ can OC just as far as the 3500+ so what's the point of getting a 3500+" they say...
The point is the 3500+ made it to 290 FSB on stock voltages, while the 3000+ had to get an 8.3% voltage increase, which means the 3500+ has a lot more headroom to OC, as compared to the 3000+ which already will be running at higher temps due to the 8.3% voltage increase.
But either way I'm still stoked that the OC'd 3000+ is beating a $600-700 FX chip!!!
The whole bang-for-buck idea doesn't quite work when you have to spend a lot of money on memory, though. Is there a good bang-for-buck memory that can be paired with this cpu?
I have corrected the CPU voltages for overclocking in the review. The MSI K8N Neo2 increments voltages by % - default, +3.3%, +5%, +8.3%, +10.0%. After talking with AMD the correct default voltage is 1.4V. Reviewing my notes and screen captures, the correct overclock voltages for 290x9 were Default for the 90nm 3500+ and +8.3% for the 90nm 3000+. I had incorrectly assumed 1.5V as the default voltage instead of the correct 1.4V.
The charts have now been corrected. Thank you for helping me get the voltages corrected. If you notice the screen capture for the 3000+ at 290x9 on the Overclock page it is showing just over 1.5V, which is consistent with a +8.3% CPU voltage of 1.52V.
I'm wondering if the board you guys used can set the memory ratios to keep memory around 400fsb? Just wondering if we all have to buy this expensive ram, or can we use our current Corsair C2 DDR400 stuff? Can you do something like this:
http://www.xbitlabs.com/articles/cpu/display/sempr...
Which allows everyone to use older memory (decent stuff anyway) and still get a LOT out of these cpus. Let's face it, A64 isn't memory starved (or the 754's wouldn't be so good, including the sempron) so could we get a chart similar to the one on xbit labs (for the Epox 8KDA3+ I think) for the k8n neo2 you used? Or is the Epox just better for people that don't want to rush out to buy $281 memory?
Wow...I guess I take back what I said about getting a S754 AMD64 Mobile w/a DFI LanParty NF3-Gb and overclocking it. From the looks of things a I'm going to wait for the 90nm core revision w/SSE3 and pair it with a NF4 mobo...DFI should have their S939 LanParty NF4 board out by then ^^ Looks like the memory controllers on these new chips are a lot better then their predecessors.
290 FSB on just 1.6V is awesome! 1.75V and a decent vatercooling system should yield some nice FSB speeds...325 FSB anyone?
#36 - Athlon 64 processors are unlocked downward, so Wesley was able to go from 11x on the 3500 down to 9x. This downward unlocking allows overclockers to max-out their motherboards and RAM while keeping the CPU running at the same speed. Wesley could have easily achieved the same 2610MHz CPU speed with the 11x multiplier at FSB of 237, but the FSB would have slowed things down.
Athlon 64FX processors are unlocked both up and down.
#34- SSE3 support along with other improvements are planned to be added with the E0 revision of the core. The current part is the D0 revision and the only performance improving features are the improved DRAM page closing policy, and second write-combining buffer. They explain why the 90nm parts were slightly faster than equivalent 130nm parts.
#35- thanks for pointing out where it says they used a different HSF to the standard retail one. I guess that explains why he was able to put 1.6V in the 3000+ and take it up to 2610MHz without frying it :)
Great article, I'm liking what I see and it's definiately on my hot purchase list.
However, I noticed you didn't discuss SSE3 support - I read somewhere else this isn't implemented yet - is that true, and if so when is SSE3 pencilled in for A64s?
I would love for an overclocked SEMPRON 3200+ (939) to be included in this comparison. Initial reviews give this chip high expectations, but I want to see apples-to-apples in a well done review such as this one.
What about the FX line? Might (have) asked them when they will transition to 90nm? This is important due to the FX having unlocked multipliers (I use Phase-Change and watercooling together so I'm looking to increase multipliers as well as FSB).
Was the retail HSF used for these overclocks?
Also, I dont get why temperatures were not documented... maybe you could list the temps that these CPUs were running under prime95.
This information on the overclocking capabilities of the new Atlon 64s is very helpful. The article is much appreciated.
It looks like it makes even more important the question of what memories provide the best performance at close to DDR 600 speeds on the Athlon 64 platform. I know this has been addressed looking at the latest PC3200 memories, by Anandtech, but are there other memories with higher speed ratings that might also be considered. For example, what about the Ballistix PC4000; does it pick up where their PC3200 leaves off? Or, is the latest OCZ PC3200 high performance memory that you used the best memory for DDR 600 speeds (or thereabouts) on the NForce3 platform?
If you already have overclock results of a p4 from another article, how difficult would it be to include in the graphs? Or were those results using a different enough configuration that it is not an applicable comparison?(In which case, as a reader that loves Anandtech for your thoroughness, I would like to see an applicable comparison.)
All in all, good review. Not as overly wordy as some have been recently(Though I won't name names. ;P).
With the 3500+ showing a 20% overclock and the 3000+ hitting a 45% overclock, it would be great to know how the 3200+ would overclock in this comparison.
Wesley, thank you so much. Once again, another fine job.
Further to my earlier comment, the default core voltage of all the Winchester-core 90nm A64 parts currently available is 1.4V, not 1.5V as indicated in the review. Its important this is corrected on the Overclocking page of the review as it is very relevant to the obtained results.
I now see that you didn't actually measure the temperature under full-load conditions. Other reports suggest that the 90nm parts do run cooler when idle than the equivalent 130nm parts, but are hotter under full-load conditions due to the higher thermal density. They have been measured as using less power under full-load than the 130nm parts, but run hotter because that power is concentrated in a smaller core.
I'd be very interested to know just how hot that 3000+ got under full-load conditions (eg. running Prime95) when you were feeding it 1.6V instead of 1.4V, and had it clocked at the maximum of 2610MHz. If you were using the standard retail HSF, it may have been rather hot :)
----
As for why the 90nm parts run a little faster than the 130nm parts, I found this post on the AMD forum. I don't know if the info is accurate, but it sounds reasonable:
Whether the 90nm process for the 3000+ to 3500+ runs cooler is still up for speculation to a degree. What will eventually be shown is that the TDP for these processors is lower than the current 130nm. (currently it is 89W TDP, the TDP for these three - when the information is released - is 67W).
In addition the 90nm A64 (DH8-D0) has these improvements over the 130nm (DH7-CG):
- improved DRAM page closing policy
- improved memory addressing with graphics cards using main memory (eg. integrated cards) as frame buffer
- memory controller power reductions (DDR receivers go off in default)
- memory power consumption reductions (CKE pins disconnect)
- second write combining buffer
- SAHF and LAHF instructions are now supported in 64bit mode
The highest stable overclock we could achieve with the P4 on air cooling was 3.92GHz (280x14) on the best overclocking 925X board. Others have achieved higher overclocks with water and phase-change cooling, and higher overclocks will also likely be achieved with those methods on the new 90nm Athlon 64 processors.
We will be looking at Pentium 4 overclocking again in the upcoming launch of some new and improved P4 processors.
I am an AMD freak, and Im happy they launched the winchester. You should have, however overclocked the Pentium-4 also, just to compare the scalability of both the CPUs.I had read an article on somebody overclocking the pentium-4 to 6 Ghz. Though this was an unstable overclock, what this indirectly implies is that despite of have a 30-odd stage pipeline, intel may find it difficult to reach speeds in excess of 5Ghz using the 0.09u process...I expect a more thorough comparo soon.....
cheers!
All the buzz in this article is about the O/C'ing capabilities of the new .90 die...personally im just as impressed or maybe even more so with the performance of the memory used in this testing. Having made that statement it is clear that the O/C'ing capability of the 3000+ version of this Proc. takes us back to the good old days of the Celery 300.
#11 & #16 - The memory brand is identified in the "Performace Test Configuration" on p.4 and the timings are in Overclocking table on p.5.
The OCZ PC3200 Platinum Rev. 2 and other top performing memory is tested on the Athlon 64 in "Athlon 64 Memory: Rewriting the Rules" at http://www.anandtech.com/memory/showdoc.aspx?i=222... Some memory in that review made it to DDR618 on A64, but DDR580 at 1T was the fastest 1T performance.
I haven't finished reading the article, but in the overclocking section of the review, you say the 3500+ overclocked from 2200 to 2610MHz at the default core voltage of 1.5V, and that the 3000+ went from 1800 to 2610MHz just by raising the core voltage from the default of 1.5V up to 1.6V.
I was under the impression that the default core voltage for the 90nm parts is 1.4V! Was the mobo BIOS version used not correctly setting 1.4V by default, or is the default actually 1.5V?
If as I believe the default is 1.4V, both chips were overvolted to reach 2610MHz, and the 3000+ in particular had to be raised from 1.4V to a much higher 1.6V. Its good that it still seemed to be running at a normal temperature!
wesley, One more thing -- as a future idea for an article -- a comparison of typical systems running different memory speeds -- I was under the impression that the price / performance ratio is very poor ... eg. ddr600 is 80% more expensive but only gives 10% more performance?
also found this article for those who want to know -- no athlon xp's though , but you can guess -- I look at the ut03 botmatch table, they seem to be comparable over these three articles (correct me if im wrong)
To help in comparisons we have added test results for the fastest current CPUs from AMD and Intel. The FX53 runs at 2.4GHz with 1MB of cache compared to the 512k on the 3000+ and 3500+. The Intel 560 runs at 3.6GHz and was tested on an Intel 925X chipset motherboard.
#4 - Corrections made. It should also be pointed out that Socket 939 nForce3 uses the Ultra chipset which already supports 1000 HT. It is the Socket 754 nF3-250 that normally supports just 800HT. All VIA 939 chipsets also support 1000 HT.
Wow, I wonder what types of memory will work best with the Athlons in the .90 die size? I'd love to be able to get a 3000+ at 2.6 and have extra money in my pocket for other things! Sucks that nForce 4 won't have AGP for my 6800GT so I'll just get a Asus A8V and stay with AGP for one more generation.
Go AMD go, I can't wait to get rid of my intel setup and go back to my one and only AMD!
Thank you for the write of AnandTech. We have been waiting for some guidance and this is much appreciated. I will probably buy the rest of my AMD build (motherboard and CPU) next weeks. The rest of the parts have arrived. Was just waiting on an assessment of the 90nm chips.
Nice finally be able to afford a socket 939 just got to wait to some NForce 4 boards come out and have revision 2 so they have 1GHz HT.
Theres a couple of other typos such as saying "but they confirmed that the AMD 130nm process appears to run at least as cool as current 130nm processors."
I believe it should have said on the first 130nm (90nm).
I'm having problems finding benchies that compare these new cpus with my current rig.... how much of a performance gap are we talking about over a 3.0 northwood? 50%?
These look like great chips - I'd say my next processor will be a 90nm 3000+.
Btw, you've got the table a little screwy for overclocking - the processor speeds are labelled the wrong way around, and for the 3000+ overclock it should be 2592 (assuming 288x9).
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
89 Comments
Back to Article
gchen77 - Wednesday, March 30, 2005 - link
Can someone please explain the effects of raising vcore?I'm a relatively newbie to overclocking but I remember in the past (with Athlon XPs) raising vcore was almost certain death unless you had water cooling or your pc running in a freezer :)
jer - Wednesday, December 8, 2004 - link
Wesley Fink,could u make a screenshot of the Memory tab in CPU-Z of the 90nm A64 3000+ cpu ??
thx so much
Goomzz - Saturday, December 4, 2004 - link
Just got my winchester 3000+ and my MSI K8N MSI Neo2 Plat. Since it's an x-mas gift can put it together until then. Putting it with Corsair XMS DDR 400 memory. I'll let you guys know how it goes.Goomzz - Saturday, December 4, 2004 - link
romano25 - Wednesday, November 24, 2004 - link
I dont get it...1)IS 3500 64 voltage 1.5 Volts?
2)Looks like the decreased the CPu multiplier on 3500 coz by default it is 11? Why? Does it affect ur performance?
romano25 - Wednesday, November 24, 2004 - link
bobbozzo - Monday, November 22, 2004 - link
#82: it's been answered: get a board (MSI) that allows the Memory & FSB to run at an adjustable ratio, so the memory can run slower than the FSB.scius - Tuesday, November 16, 2004 - link
Cheaper Ram Altnernatives:A few other readers have mentioned this, but it seems there hasn't been much of an answer (though a few worthy attempts, notably that the 3200+ is probably a better choice).
The Question: What ram would let us run at the highest FSB for the least $.
Obviously you can just buy the faster stuff (DDR500, or whatever), but there must be sticks that, with looser timings(small cost), can let your processor scream(huge gains) while staying relatively stable.
Anyway, I haven't found any articles about it, but if anyone has, or has some personal experience here, i'm sure we're all eager to hear it.
VoodooGamez - Thursday, November 4, 2004 - link
Great article Wesley!cryptonomicon - Wednesday, November 3, 2004 - link
great article anand!The 90nm process sounds like a great improvement (especially for oc).
athlon2004 - Sunday, October 31, 2004 - link
Err.....i'm really a noob....didn't see my post so i thought that there was a problem, but apparently it doesn't reload page 4, but page one. Feel free to delete the other responses.athlon2004 - Sunday, October 31, 2004 - link
Are those speed prime stable?athlon2004 - Sunday, October 31, 2004 - link
Are those speeds prime stable?athlon2004 - Sunday, October 31, 2004 - link
nitenichiryu1 - Saturday, October 30, 2004 - link
great article. what was the core of the 3500+ chip? was it newcastle or winchester for the 90nm? and is there any difference between these two cores? on sites such as newegg and zipzoomfly, the 90nm 3500+ are advertised as winchester on zipzoomfly and neweggs are advertised as newcastles. thanksWesley Fink - Saturday, October 23, 2004 - link
#73 -You normally have to drop HT at higher overclocks, to keep the aggregate somewhere in the 1000HT range. Some boards handle higher HT than others. Since x3 HT was used for the 290x9 benchmarks in the review, I think it should be clear that the lower HT ratio does not adversely affect performance as long as the HT is somewhere around 800 or greater. 290x3 is an HT of 870.
4X HT usually stops working around 260 to 275 (1040 to 1100) on most boards that support 1000HT (5X) and you need to drop to x3. As a side note, none of the 1000HT boards we have tested work well at 2x HT.
DaveHull - Thursday, October 21, 2004 - link
Wesley,One thing I've noticed when overclocking the MSI board with the 3000+ A64 (same as in the review) is that you have to lower the Hyper Transport (HT) from x5 to x3 to get the overclock of 290 FSB, giving an HT speed of 870 mhz instead of the stock 1000 mhz. My cpu/board refuses to run at a HT speed of over 1070.
Is this true of the overclock in the article? Will the decreased HT speed negate the performance benefits of the overclock in any practical areas?
Thank you,
Dave
Furen - Thursday, October 21, 2004 - link
actually, disregard the above. The PQI 1024DP has a higher latency rating, the one you'd need for DDR570 is the 1024DBU, which is $245... stupid dealtime and it's incorrect linking ^^ (yeah, I'll blame it on dealtime)Furen - Thursday, October 21, 2004 - link
For all you people asking questions about ram, here's an alternative: =)You could always use the PQI 3200 Turbo, which supports speeds of up to DDR570 and goes for $172.00 at newegg, here's the link:
http://www.newegg.com/app/viewproductdesc.asp?DEPA...
This ram was reviewed here at anandtech and while not as insanely good as the OCZ EL Platinum, the performance compared to other ram was not too bad. As mentioned in the review, though it has lifetime warranty, the manufacturer is a new name, so the support service is a big unknown.
Furthermore, running the FSB at 285MHz instead of 290MHz will give you an overclock that is 45Mhz lower (2.565GHz) but the ram being 100 dollars cheaper is worth it, in my opinion.
AlphaFox - Tuesday, October 19, 2004 - link
anyone notice the price for these things is going UP after AMD just lowered their price? newegg had them for $199, up from $189 a few days ago, now its $215!!! HELLO, the prices are supposed to go DOWN after AMD loweres the price!Live - Tuesday, October 19, 2004 - link
Great read!Good work Wesley Fink.
I looks like I am finally getting close to being able to upgrade my athlon xp setup without breaking the bank and get a real performance boost I feel the earlier A64 cpus and mobos lacked. To bad the PCIe problem is still around. I dont want an upgrade that requires a new motherboard with my next GPU purchase.
toNka64 - Tuesday, October 19, 2004 - link
I currently have an A64 3200 Clawhammer on a VIA K8T800, so OCing is a no go without a working AGP/PCI lock. I have a Coolmax Taurus 400w PSU and OCZ EL Platinum 1GB (2x512) memory (2-3-2-5).If I upgraded my mobo to the Neo2 and my CPU to the 300+, do you think i could hit 2.6GHz with my current memory and PSU?
Doormat - Saturday, October 16, 2004 - link
Yea, I'm beginning to think I should get a 3200, a conservative OC would be 2.5GHz at 10x250. And DDR500 isnt that hard to get a hold of. Plus it fits in nicely with a 1GHz hypertransport bus.BBoe - Saturday, October 16, 2004 - link
Is this FSB of 290 for reals? It really sounds like a big leap from the 250 or 261.PrinceGaz - Saturday, October 16, 2004 - link
People who want the best performance and value ought to get the 3200+ instead of the 3000+ as it not only stands a better chance of reaching something like 2610MHz than the 3000+, but it should do so at a slightly lower voltage.More importantly, the memory with the 3200+ could run 1:1 at 261x10 instead of the 290x9 required by the 3000+. Even though the 3200+ will be a little more expensive than the 3000+, theres a much better chance of finding reasonably priced memory which can run at DDR522 with good timings, than there is of memory that can manage DDR580. At least with the 3200+ you've got the choice of a 9x or 10x multiplier to get the best out of your memory.
rjm55 - Saturday, October 16, 2004 - link
I know you used OCZ with Samsung TCCD for this review, but you have said in past reviews that any TCCD will come close to these memory results. That means you can choose Samsung TCCD from Corsair, Mushkin, Kingston, OCZ, Geil, PQI, and probably others. Crucial Ballistix is another choice from your Athlon 64 memory tests. Those concerned about Value ought to shop all the TCCD brands and Crucial Ballistix, because some are cheaper than others.Wesley Fink - Friday, October 15, 2004 - link
#62 - The MSI K8N Neo2 can be set to Auto, 100, 133, 166, 200 (DDR200,266,333,400 or 1:2.2:3,5:6,1:1). With these ratios you could run 290 FSB with a 2:3 ratio (266) and memory would be at DDR387, at 5:6 (333) mem would be running at DDR482.ThePlagiarmaster - Friday, October 15, 2004 - link
Wesley:Could you answer my #43 post about memory ratios on the board you tested with?
Thanks
PrinceGaz - Friday, October 15, 2004 - link
#58- tha major selling point of nForce 4 will be PCIe support, if you don't want PCIe then get an nForce3 instead. The Inquirer claim to have a piccy of MSI's new nForce4 S939 board athttp://www.theinquirer.net/?article=19026
That is presumably the successor of the popular K8N Neo2 Platinum. Its a shame it only has two PCIe x1 sockets and still has four old PCI sockets, I'd have liked to see at least three PCIe x1 and just a couple of PCI for legacy support.
ciwell - Friday, October 15, 2004 - link
Can you address the RAM question posted above?Wesley Fink - Friday, October 15, 2004 - link
#57 - There were NO tests on Socket 754 processors in this review, since 90nm is only available as 939. This is stated in the review. To see the impact of the new Winchester core and die-shrink on performance we downclocked a 939 .13 CPU to 1.8GHz - the same specs as the 90nm 939. This is clearly stated in bold in the review "We also ran benchmarks of the 130nm processor at Socket 939 3000+ speeds, but these results are theoretical. There is no production 130nm Socket 939 3000+, so these results were just to compare the impact of the die-shrink and Winchester core on performance."Akira1224 - Friday, October 15, 2004 - link
#56 can you post an official Nvidia link stating that. I can't seem to find the official word anywhere. I just figured since you stated that they will not support AGP as a fact you have seen something official.Thank you !
Cybercat - Friday, October 15, 2004 - link
AMD .09 Athlon 64 3500+AMD .13 Athlon 64 3500+
AMD .09 Athlon 64 3000+
AMD .09 Athlon 64 3000+ (downclocked .13 CPU)
AMD FX53 A64 (.13-2.4GHz-1MB Cache)
So, does this mean that the Socket 754 3000+ was downclocked to 1.8GHz?
IceWindius - Friday, October 15, 2004 - link
#44Fraid not, nForce 4 will only support PCI-E video cards from this point forward.
thebluesgnr - Friday, October 15, 2004 - link
The 1.8GHz Winchester is faster than the Sempron 3100+ for a few reasons:1) dual channel support;
2) 512KB L2 cache (versus 256KB on the Sempron);
3) small improvements on the Winchester core.
Not to mention the lack of 64-bit support on the Sempron. But that's not so bad for a chip that's half the price (they cost $100 and $199 on newegg). The price difference is too big, the suggested price for the A64 3000+ is $165, so I expect to pay 150-165 for the OEM version. Newegg's logistics is too good. ;)
Wesley Fink - Thursday, October 14, 2004 - link
#50 - All Athlon 64 processors can be set to lower CPU ratios. Only the FX can be set either lower or higher multipliers. I was testing and had set the 3800+ to a 9 multiplier in BIOS. The 3800+ is at stock a 12 multiplier and runs at 2.4GHz.To minimize confusion we replaced the 9x capture with a 12x 3800+ screen capture in the review.
#52 is correct - the Sempron PR is based on Celeron and not A64. The Sempron 3100+ is actually a bit slower than a 2800+ Athlon 64. The Sempron is also 32-bit only and does not support 64-bit operation even though it will fit in Socket 754.
Wesley Fink - Thursday, October 14, 2004 - link
#47 - the 3400+ is for Socket 754. The only 90nm chips are for Socket 939 at present. 3500+ is the correct name and you can buy the chip from Monarch Computers for one at http://www.monarchcomputer.com/Merchant2/merchant.... New Egg does have the 3200+ and 3000+ 90nm in stock - but not the 3500+. Look for Core: Winchester and Process: 90nm in the description. I don't have any idea what the 3400+ Socket 939 chip is that is advertised at NewEgg, but it's not Winchester core or 90nm process.#48 - All Athlon 64 processors can be set to lower CPU ratios. Only the FX can be set either lower or higher multipliers. I was testing and had set the 3800+ to a 9 multiplier in BIOS. The 3800+ is at stock a 12 multiplier and runs at 2.4GHz.
Keypo - Thursday, October 14, 2004 - link
The Sempron PR is based towards the Celeron Performance and the Athlon is PR is for PentiumKeypo - Thursday, October 14, 2004 - link
Araemo - Thursday, October 14, 2004 - link
Soemone tell me I'm crazy.. please?Two things: one, on page 1 of the review, look at the second cpu-z screenshot.
Why is the 3800+ running at 1.8?
Anyways..
Amd Athlon64 3000+ on S939 = 1.8 Ghz.
Amd Sempron 3100+ on S754 = 1.8ghz
Why does a sempron have a higher 'rating' than an identically clocked athlon64?
PrinceGaz - Thursday, October 14, 2004 - link
Thankyou Wesley for correcting the voltage figures. The originally posted required voltages were not at all encouarging imo.Being able to get the 3500+ to 2610MHz on default voltage, and the 3000+ to the same speed with an acceptable 1.52V is good to know.
I'm a lot happier now :)
FinalFantasy - Thursday, October 14, 2004 - link
This comment is a response to a lot of post i've seen across different forums.People are really getting confused on which chip can OC better. I'm hearing people say "the 3000+ can OC just as far as the 3500+ so what's the point of getting a 3500+" they say...
The point is the 3500+ made it to 290 FSB on stock voltages, while the 3000+ had to get an 8.3% voltage increase, which means the 3500+ has a lot more headroom to OC, as compared to the 3000+ which already will be running at higher temps due to the 8.3% voltage increase.
But either way I'm still stoked that the OC'd 3000+ is beating a $600-700 FX chip!!!
330Pilot - Thursday, October 14, 2004 - link
Looking at the Newegg site, I notice that the 90nm 2.2GHz version is known as the 3400+ and not the 3500+.Is there a difference between what Anandtech reviewed the the vesion available on Newegg, or is one of them mistaken?
kmmatney - Thursday, October 14, 2004 - link
Great review!The whole bang-for-buck idea doesn't quite work when you have to spend a lot of money on memory, though. Is there a good bang-for-buck memory that can be paired with this cpu?
Wesley Fink - Thursday, October 14, 2004 - link
I have corrected the CPU voltages for overclocking in the review. The MSI K8N Neo2 increments voltages by % - default, +3.3%, +5%, +8.3%, +10.0%. After talking with AMD the correct default voltage is 1.4V. Reviewing my notes and screen captures, the correct overclock voltages for 290x9 were Default for the 90nm 3500+ and +8.3% for the 90nm 3000+. I had incorrectly assumed 1.5V as the default voltage instead of the correct 1.4V.The charts have now been corrected. Thank you for helping me get the voltages corrected. If you notice the screen capture for the 3000+ at 290x9 on the Overclock page it is showing just over 1.5V, which is consistent with a +8.3% CPU voltage of 1.52V.
DEMO24 - Thursday, October 14, 2004 - link
#8 Nforce WILL have AGP dunno what your talking about.(hope that wasnt answered before cuase I didnt read them all)
ThePlagiarmaster - Thursday, October 14, 2004 - link
Wesley:I'm wondering if the board you guys used can set the memory ratios to keep memory around 400fsb? Just wondering if we all have to buy this expensive ram, or can we use our current Corsair C2 DDR400 stuff? Can you do something like this:
http://www.xbitlabs.com/articles/cpu/display/sempr...
Which allows everyone to use older memory (decent stuff anyway) and still get a LOT out of these cpus. Let's face it, A64 isn't memory starved (or the 754's wouldn't be so good, including the sempron) so could we get a chart similar to the one on xbit labs (for the Epox 8KDA3+ I think) for the k8n neo2 you used? Or is the Epox just better for people that don't want to rush out to buy $281 memory?
glennpratt - Thursday, October 14, 2004 - link
Better hurry, the dual cores will come out and you will have to change you mind again!FinalFantasy - Thursday, October 14, 2004 - link
Wow...I guess I take back what I said about getting a S754 AMD64 Mobile w/a DFI LanParty NF3-Gb and overclocking it. From the looks of things a I'm going to wait for the 90nm core revision w/SSE3 and pair it with a NF4 mobo...DFI should have their S939 LanParty NF4 board out by then ^^ Looks like the memory controllers on these new chips are a lot better then their predecessors.290 FSB on just 1.6V is awesome! 1.75V and a decent vatercooling system should yield some nice FSB speeds...325 FSB anyone?
Araemo - Thursday, October 14, 2004 - link
#36 -Beause of cool'n'quiet, you can lower the multiplier. They're only locked to a max. The 3000+ is 1.8ghz, 200x9.
the 3500+ is 2.2, 200x11
They just lowered the multiplier on the 3500+ to 9.
slashbinslashbash - Thursday, October 14, 2004 - link
#36 - Athlon 64 processors are unlocked downward, so Wesley was able to go from 11x on the 3500 down to 9x. This downward unlocking allows overclockers to max-out their motherboards and RAM while keeping the CPU running at the same speed. Wesley could have easily achieved the same 2610MHz CPU speed with the 11x multiplier at FSB of 237, but the FSB would have slowed things down.Athlon 64FX processors are unlocked both up and down.
PrinceGaz - Thursday, October 14, 2004 - link
#34- SSE3 support along with other improvements are planned to be added with the E0 revision of the core. The current part is the D0 revision and the only performance improving features are the improved DRAM page closing policy, and second write-combining buffer. They explain why the 90nm parts were slightly faster than equivalent 130nm parts.#35- thanks for pointing out where it says they used a different HSF to the standard retail one. I guess that explains why he was able to put 1.6V in the 3000+ and take it up to 2610MHz without frying it :)
Entropy531 - Thursday, October 14, 2004 - link
#34, look at the CPU-Z screenshot. No SSE3.pio!pio! - Thursday, October 14, 2004 - link
How did you overclock both pieces to 290x9? Dont they have different locked multipliers?fic - Thursday, October 14, 2004 - link
HS used: Thermaltake Silent Boost K8From "Front Side Bus Overclocking Testbed" table in "Overclocking Results and Heat"
IdahoB - Thursday, October 14, 2004 - link
Great article, I'm liking what I see and it's definiately on my hot purchase list.However, I noticed you didn't discuss SSE3 support - I read somewhere else this isn't implemented yet - is that true, and if so when is SSE3 pencilled in for A64s?
tr00p - Thursday, October 14, 2004 - link
I would love for an overclocked SEMPRON 3200+ (939) to be included in this comparison. Initial reviews give this chip high expectations, but I want to see apples-to-apples in a well done review such as this one.SLI - Thursday, October 14, 2004 - link
What about the FX line? Might (have) asked them when they will transition to 90nm? This is important due to the FX having unlocked multipliers (I use Phase-Change and watercooling together so I'm looking to increase multipliers as well as FSB).AlphaFox - Thursday, October 14, 2004 - link
Was the retail HSF used for these overclocks?Also, I dont get why temperatures were not documented... maybe you could list the temps that these CPUs were running under prime95.
Spacecomber - Thursday, October 14, 2004 - link
This information on the overclocking capabilities of the new Atlon 64s is very helpful. The article is much appreciated.It looks like it makes even more important the question of what memories provide the best performance at close to DDR 600 speeds on the Athlon 64 platform. I know this has been addressed looking at the latest PC3200 memories, by Anandtech, but are there other memories with higher speed ratings that might also be considered. For example, what about the Ballistix PC4000; does it pick up where their PC3200 leaves off? Or, is the latest OCZ PC3200 high performance memory that you used the best memory for DDR 600 speeds (or thereabouts) on the NForce3 platform?
Space
Bugler - Thursday, October 14, 2004 - link
Newegg Model#: OCZ4001024ELDCPER2-KItem#: N82E16820146890
OCZ EL Platinum Revision 2 Dual Channel Kit 184-Pin 1GB(512MBx2) DDR PC-3200 - Retail $281
Araemo - Thursday, October 14, 2004 - link
nevermind...It is only for sale in 1 gig packs of 2x512 right now, different part #:
http://www.newegg.com/app/viewproductdesc.asp?DEPA...
Araemo - Thursday, October 14, 2004 - link
Also.. is that ram available in the retail channel? I wanted to look up the price, and found the part number(I believe) OCZ400512ELPER2However, this isn't on newegg, or pricewatch.
Araemo - Thursday, October 14, 2004 - link
#23 - WesleyIf you already have overclock results of a p4 from another article, how difficult would it be to include in the graphs? Or were those results using a different enough configuration that it is not an applicable comparison?(In which case, as a reader that loves Anandtech for your thoroughness, I would like to see an applicable comparison.)
All in all, good review. Not as overly wordy as some have been recently(Though I won't name names. ;P).
Bugler - Thursday, October 14, 2004 - link
With the 3500+ showing a 20% overclock and the 3000+ hitting a 45% overclock, it would be great to know how the 3200+ would overclock in this comparison.Wesley, thank you so much. Once again, another fine job.
PrinceGaz - Thursday, October 14, 2004 - link
Further to my earlier comment, the default core voltage of all the Winchester-core 90nm A64 parts currently available is 1.4V, not 1.5V as indicated in the review. Its important this is corrected on the Overclocking page of the review as it is very relevant to the obtained results.I now see that you didn't actually measure the temperature under full-load conditions. Other reports suggest that the 90nm parts do run cooler when idle than the equivalent 130nm parts, but are hotter under full-load conditions due to the higher thermal density. They have been measured as using less power under full-load than the 130nm parts, but run hotter because that power is concentrated in a smaller core.
I'd be very interested to know just how hot that 3000+ got under full-load conditions (eg. running Prime95) when you were feeding it 1.6V instead of 1.4V, and had it clocked at the maximum of 2610MHz. If you were using the standard retail HSF, it may have been rather hot :)
----
As for why the 90nm parts run a little faster than the 130nm parts, I found this post on the AMD forum. I don't know if the info is accurate, but it sounds reasonable:
Whether the 90nm process for the 3000+ to 3500+ runs cooler is still up for speculation to a degree. What will eventually be shown is that the TDP for these processors is lower than the current 130nm. (currently it is 89W TDP, the TDP for these three - when the information is released - is 67W).
In addition the 90nm A64 (DH8-D0) has these improvements over the 130nm (DH7-CG):
- improved DRAM page closing policy
- improved memory addressing with graphics cards using main memory (eg. integrated cards) as frame buffer
- memory controller power reductions (DDR receivers go off in default)
- memory power consumption reductions (CKE pins disconnect)
- second write combining buffer
- SAHF and LAHF instructions are now supported in 64bit mode
Wesley Fink - Thursday, October 14, 2004 - link
#22 - I appreciate your suggestion, and we did overclock the Pentium 4 775 in our "Intel 925X Roundup: Creative Engineering 101" at http://www.anandtech.com/mb/showdoc.aspx?i=2162.The highest stable overclock we could achieve with the P4 on air cooling was 3.92GHz (280x14) on the best overclocking 925X board. Others have achieved higher overclocks with water and phase-change cooling, and higher overclocks will also likely be achieved with those methods on the new 90nm Athlon 64 processors.
We will be looking at Pentium 4 overclocking again in the upcoming launch of some new and improved P4 processors.
thermalpaste - Thursday, October 14, 2004 - link
I am an AMD freak, and Im happy they launched the winchester. You should have, however overclocked the Pentium-4 also, just to compare the scalability of both the CPUs.I had read an article on somebody overclocking the pentium-4 to 6 Ghz. Though this was an unstable overclock, what this indirectly implies is that despite of have a 30-odd stage pipeline, intel may find it difficult to reach speeds in excess of 5Ghz using the 0.09u process...I expect a more thorough comparo soon.....cheers!
deathwalker - Thursday, October 14, 2004 - link
All the buzz in this article is about the O/C'ing capabilities of the new .90 die...personally im just as impressed or maybe even more so with the performance of the memory used in this testing. Having made that statement it is clear that the O/C'ing capability of the 3000+ version of this Proc. takes us back to the good old days of the Celery 300.Wesley Fink - Thursday, October 14, 2004 - link
#11 & #16 - The memory brand is identified in the "Performace Test Configuration" on p.4 and the timings are in Overclocking table on p.5.The OCZ PC3200 Platinum Rev. 2 and other top performing memory is tested on the Athlon 64 in "Athlon 64 Memory: Rewriting the Rules" at http://www.anandtech.com/memory/showdoc.aspx?i=222... Some memory in that review made it to DDR618 on A64, but DDR580 at 1T was the fastest 1T performance.
Zar0n - Thursday, October 14, 2004 - link
Great article.Great CPU, now all we need is PCI-E bords.
PrinceGaz - Thursday, October 14, 2004 - link
I haven't finished reading the article, but in the overclocking section of the review, you say the 3500+ overclocked from 2200 to 2610MHz at the default core voltage of 1.5V, and that the 3000+ went from 1800 to 2610MHz just by raising the core voltage from the default of 1.5V up to 1.6V.I was under the impression that the default core voltage for the 90nm parts is 1.4V! Was the mobo BIOS version used not correctly setting 1.4V by default, or is the default actually 1.5V?
If as I believe the default is 1.4V, both chips were overvolted to reach 2610MHz, and the 3000+ in particular had to be raised from 1.4V to a much higher 1.6V. Its good that it still seemed to be running at a normal temperature!
xsilver - Thursday, October 14, 2004 - link
wesley, One more thing -- as a future idea for an article -- a comparison of typical systems running different memory speeds -- I was under the impression that the price / performance ratio is very poor ... eg. ddr600 is 80% more expensive but only gives 10% more performance?KHysiek - Thursday, October 14, 2004 - link
So this memory (double 512MB pack) was running at 580MHz ?! Wow.What timings then ?
xsilver - Thursday, October 14, 2004 - link
thanks wesleyalso found this article for those who want to know -- no athlon xp's though , but you can guess -- I look at the ut03 botmatch table, they seem to be comparable over these three articles (correct me if im wrong)
http://www.anandtech.com/cpuchipsets/showdoc.aspx?...
For those who cant be bothered here's an interesting stat
prescott 2.8 - ut03 botmatch 67.9 fps
3500+/3000+ OC @290x9 132.7 fps
for those who are mathmatically challenged -- THATS DOUBLE!!! time to upgrade
Zebo - Thursday, October 14, 2004 - link
I dislike overclcoking reviews that don't show a pentium equivalent in thier test.What I want to know if I buy a 3.2C or E for the same price and overclock it with similar cooling how it would compare too....does'nt everybody?
Reviewers really need to work these ideas in.
Wesley Fink - Thursday, October 14, 2004 - link
#11 - Mem:FSB was 1:1 in overclocking. At 290 we were still running 1:1.Wesley Fink - Thursday, October 14, 2004 - link
#3 and #9 -Results for the FX53 and Intel 92X/915 running the nVidia 6800 Ultra are available at http://www.anandtech.com/mb/showdoc.aspx?i=2128.
To help in comparisons we have added test results for the fastest current CPUs from AMD and Intel. The FX53 runs at 2.4GHz with 1MB of cache compared to the 512k on the 3000+ and 3500+. The Intel 560 runs at 3.6GHz and was tested on an Intel 925X chipset motherboard.
KHysiek - Thursday, October 14, 2004 - link
What was mem:fsb ratio in this overclocking (benchmarks) ?Wesley Fink - Thursday, October 14, 2004 - link
#1 - Table has been corrected.#4 - Corrections made. It should also be pointed out that Socket 939 nForce3 uses the Ultra chipset which already supports 1000 HT. It is the Socket 754 nF3-250 that normally supports just 800HT. All VIA 939 chipsets also support 1000 HT.
ViRGE - Thursday, October 14, 2004 - link
I agree with #3, some more numbers would be nice, preferably at least one Northwood, a Prescott, and a S754 3400+(2.4ghz).IceWindius - Thursday, October 14, 2004 - link
Wow, I wonder what types of memory will work best with the Athlons in the .90 die size? I'd love to be able to get a 3000+ at 2.6 and have extra money in my pocket for other things! Sucks that nForce 4 won't have AGP for my 6800GT so I'll just get a Asus A8V and stay with AGP for one more generation.Go AMD go, I can't wait to get rid of my intel setup and go back to my one and only AMD!
Myrandex - Thursday, October 14, 2004 - link
Sweet article and good results. 90nm is predicted in my near future.Bugler - Thursday, October 14, 2004 - link
Thank you for the write of AnandTech. We have been waiting for some guidance and this is much appreciated. I will probably buy the rest of my AMD build (motherboard and CPU) next weeks. The rest of the parts have arrived. Was just waiting on an assessment of the 90nm chips.ariafrost - Thursday, October 14, 2004 - link
And I thought the days of nearly 50% overclocks were long gone (I had a Celeron 300A @ 450MHz waaayy back)...My next proc will be a 939 90nm part, and mobo will be Nforce 4 :)
ModFX - Thursday, October 14, 2004 - link
Nice finally be able to afford a socket 939 just got to wait to some NForce 4 boards come out and have revision 2 so they have 1GHz HT.Theres a couple of other typos such as saying "but they confirmed that the AMD 130nm process appears to run at least as cool as current 130nm processors."
I believe it should have said on the first 130nm (90nm).
xsilver - Thursday, October 14, 2004 - link
I'm having problems finding benchies that compare these new cpus with my current rig.... how much of a performance gap are we talking about over a 3.0 northwood? 50%?Decoder - Thursday, October 14, 2004 - link
Thanks for this super article. My next upgrade will be a AMD64 3000+ on a NForce 4 board! Kudos to AMD.Degrador - Thursday, October 14, 2004 - link
These look like great chips - I'd say my next processor will be a 90nm 3000+.Btw, you've got the table a little screwy for overclocking - the processor speeds are labelled the wrong way around, and for the 3000+ overclock it should be 2592 (assuming 288x9).