[Renamed] AMD Athlon 64E thread
Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee
[Renamed] AMD Athlon 64E thread
AMD will launch the new revision E on April 4.
E3 = 512K L2 Rev E desktop = "BP" in OPN, like ADA3500BPBOX
E4 = 1MB L2 Rev E desktop = "BN"
The 4000+ will start shipping later on April 15.
E3 = 512K L2 Rev E desktop = "BP" in OPN, like ADA3500BPBOX
E4 = 1MB L2 Rev E desktop = "BN"
The 4000+ will start shipping later on April 15.
Last edited by Mats on Sun Apr 03, 2005 6:38 pm, edited 4 times in total.
Seriously, yeah I really think so. Either this or T64 but I think we could call it the same thing when it comes to cooling, although T64 will undervolt better in general since it's cherry picked.
Speaking of cherry picked, which CPU's needs to be cherry picked? And even though AMD uses different codenames for Turion and Athlon, aren't they still the same? Two cores instead of four:
1 MB cache: FX, 4200+, 4000+, ML, MT (exept the 512 kB ones)
512 kB cache: 3000+, 3200+, 3500+, 3800+, and those 1.6 and 1.8 GHz MT
If you're into the 1 MB group you either got money or are prepared to buy anoter CPU/mobo because the Turion didn't work with x = got money.
But if you're into the 512 kB group there's nothing to cherry pick. To do it for the 3800+? No, 2.4 GHz is not hard to reach with todays 90 nm, but most certainly it was when they showed up. The same story goes for the mobile ones, they couldn't have done that with Oakville back then, so they gave it 1.35 V. Todays D stepping is capable of lower Vcore.
And if there are four different cores the better! Just desktop 1.8-2.4 GHz, no speed binning.
So what does this mean? Probably that it will continue to be easy finding a good CPU... unless:
They will disable half of the cache, which I doubt, this is something they come up with afterwards, not with a new core. The 3000+-3500+ line is so popular that they would probably lose money in selling those big sized dies for that low price.
or
They have come up with a way to control wafer quality (this sounds like pure evil ), tested heavily under December with week 48 as good and week 51 as bad. I doubt this would happen, but the overclockers will be our best friends as usual in this issue and tell what's good and what's bad.
For a while I was into PM just because of AMD's horrible Opteron E TDP's. But now I've realized that the newer A64 ones are probably a bit more cooler than the older ones, and PM is a bit outdated (It's not just the chipset). It doesn't have 64 bits extensions, hyperthreading or integrated memorycontroller. Now I just don't care about 64 bit, but I would very much like to have either of the other two in my next CPU (having both is to much to ask for... ).
Speaking of cherry picked, which CPU's needs to be cherry picked? And even though AMD uses different codenames for Turion and Athlon, aren't they still the same? Two cores instead of four:
1 MB cache: FX, 4200+, 4000+, ML, MT (exept the 512 kB ones)
512 kB cache: 3000+, 3200+, 3500+, 3800+, and those 1.6 and 1.8 GHz MT
If you're into the 1 MB group you either got money or are prepared to buy anoter CPU/mobo because the Turion didn't work with x = got money.
But if you're into the 512 kB group there's nothing to cherry pick. To do it for the 3800+? No, 2.4 GHz is not hard to reach with todays 90 nm, but most certainly it was when they showed up. The same story goes for the mobile ones, they couldn't have done that with Oakville back then, so they gave it 1.35 V. Todays D stepping is capable of lower Vcore.
And if there are four different cores the better! Just desktop 1.8-2.4 GHz, no speed binning.
So what does this mean? Probably that it will continue to be easy finding a good CPU... unless:
They will disable half of the cache, which I doubt, this is something they come up with afterwards, not with a new core. The 3000+-3500+ line is so popular that they would probably lose money in selling those big sized dies for that low price.
or
They have come up with a way to control wafer quality (this sounds like pure evil ), tested heavily under December with week 48 as good and week 51 as bad. I doubt this would happen, but the overclockers will be our best friends as usual in this issue and tell what's good and what's bad.
For a while I was into PM just because of AMD's horrible Opteron E TDP's. But now I've realized that the newer A64 ones are probably a bit more cooler than the older ones, and PM is a bit outdated (It's not just the chipset). It doesn't have 64 bits extensions, hyperthreading or integrated memorycontroller. Now I just don't care about 64 bit, but I would very much like to have either of the other two in my next CPU (having both is to much to ask for... ).
Last edited by Mats on Sat Apr 02, 2005 8:42 am, edited 1 time in total.
Yeah, when it comes to the different cores, I'm pretty sure that they'll all be pretty much created equal. Sure, as usual they might differ somewhat in maximum achievable overclocks, but I still think that most will underclock/undervolt fairly well.
I'm personally quite interested in the amount of cooling that a 1MB San Diego chip requires to make 2.6GHz... Judging by the early reports, 1.25V could be enough to keep it stable. This should mean that the chip will run pretty cool. A regular Winchester would probably not output more than 45-50W at that speed and voltage... Here's dreaming of an XP120 setup with a ~1000RPM 120mm fan and 2.6GHz San Diego.
But then again, we don't really know how hot/cool rev E runs. It pisses me off that Anand hasn't released more benchmarks and tests on their Opteron 252 (which is based on rev E). All they've managed so far is a few measly database benchmarks and yet they've had the CPU/CPUs in weeks!
I'm personally quite interested in the amount of cooling that a 1MB San Diego chip requires to make 2.6GHz... Judging by the early reports, 1.25V could be enough to keep it stable. This should mean that the chip will run pretty cool. A regular Winchester would probably not output more than 45-50W at that speed and voltage... Here's dreaming of an XP120 setup with a ~1000RPM 120mm fan and 2.6GHz San Diego.
But then again, we don't really know how hot/cool rev E runs. It pisses me off that Anand hasn't released more benchmarks and tests on their Opteron 252 (which is based on rev E). All they've managed so far is a few measly database benchmarks and yet they've had the CPU/CPUs in weeks!
Yep, I agree. All they have to do was to make tke most simple Kill-a-Watt comparisation with Opteron D. On the other hand, I think that the Opteron D is very rare, and now it's not even on AMD's website anymore.Mikael wrote:It pisses me off that Anand hasn't released more benchmarks and tests on their Opteron 252 (which is based on rev E). All they've managed so far is a few measly database benchmarks and yet they've had the CPU/CPUs in weeks!
i agree that anand blew it (well, they can do whatever they want in their reviews, but by having the first look at an e-stepping chip they could have gotten a lot more pageviews with some more interesting tests).
one chap over at xtremesystems has a 252 arriving today, maybe he'll give us some interesting insights.
one chap over at xtremesystems has a 252 arriving today, maybe he'll give us some interesting insights.
Athlon 64 revision E won't work on some Nforce 3/4 boards
Yep, that probably includes Turion as well.
Now how many S754 nForce mobos will be revised for a CPU they don't intend to support???
Yep, that probably includes Turion as well.
Now how many S754 nForce mobos will be revised for a CPU they don't intend to support???
According to MikeC:s report from IDF it seems I wasn't very far off the mark.Mats wrote:That's the best explanation I've heard so far, although I really suspect that they will give it another stepping code just because it's dual core.Tobias wrote:I was thinking, is it possible that the E-steppings high TDP is the result of that the E-stepping will be the first stepping to incorporate dualcore processors?
That statement is implying a thing or two. First, 67W (for D-stepping, how high is it for E-stepping?) leaves some headroom since faster units will likely be released.AMD wrote:Because AMD uses existing manufacturing processes to produce dual-core AMD64 processors, the power envelope for our dual-core processors is designed to fit into the current sockets and power infrastructures
Second, 67W is for TWO cores, meaning that a single core shouldn't draw much more than a bit higer than half of that.
And on top of this, You could still use your motherboard that run your current s939-cpu.
If it is VIA, of course
Xbitlabs review.
According to their report the new CPU runs hotter and uses more power than the last revision,
although not as much as the 130 nm CPU's.
I wonder if it will undervolt better, it probably does but maybe just enough to compensate for the extra power needed.
AFAIK a good overclocker is a good undervolter in general, but I simply don't know if it's the same with this new technique.
+
OVERCLOCKING! (...= better undervolter???)
Better memory controller, works with four memory modules at stock speed.
Just a little faster in some benchmarks (VERY small difference in general).
The reviewers didn't find any mobo that didn't work. I've read that 1 MB cache CPU's have more trouble.
-
2° C Hotter
Uses more power, 7.5 W more @ 2 GHz (load, measured on mobo)
According to their report the new CPU runs hotter and uses more power than the last revision,
although not as much as the 130 nm CPU's.
I wonder if it will undervolt better, it probably does but maybe just enough to compensate for the extra power needed.
AFAIK a good overclocker is a good undervolter in general, but I simply don't know if it's the same with this new technique.
+
OVERCLOCKING! (...= better undervolter???)
Better memory controller, works with four memory modules at stock speed.
Just a little faster in some benchmarks (VERY small difference in general).
The reviewers didn't find any mobo that didn't work. I've read that 1 MB cache CPU's have more trouble.
-
2° C Hotter
Uses more power, 7.5 W more @ 2 GHz (load, measured on mobo)
Just read the review too. It does consume slightly more power, but I'm guessing that this difference can be minimised by undervolting. Those chips do seem to undervolt very well, so in the end it might even be better than Winchester.
All in all, it does look like a good CPU. Looking at the benches, there's a fair amount of tests where the 3800+ Venice actually beats the 4000+ Clawhammer!
All in all, it does look like a good CPU. Looking at the benches, there's a fair amount of tests where the 3800+ Venice actually beats the 4000+ Clawhammer!
Xbitlabs review of Sempron E, slightly offtopic but...
According to XBit, A64E will run at either 1.35 or 1.4 V and there's no way of telling which is which from the exterior.
Sounds strange, since they put a Vcore code on every CPU.
Edit: No more Vcore code, the pic in the 3800+ review shows letter "A", which doesn't tell us anything ("A" should be equal to 1.65 V theoretically, but that's just wrong.)
Sounds strange, since they put a Vcore code on every CPU.
Edit: No more Vcore code, the pic in the 3800+ review shows letter "A", which doesn't tell us anything ("A" should be equal to 1.65 V theoretically, but that's just wrong.)
-
- *Lifetime Patron*
- Posts: 477
- Joined: Thu Feb 17, 2005 6:41 am
- Location: Nashville, TN
-
- *Lifetime Patron*
- Posts: 477
- Joined: Thu Feb 17, 2005 6:41 am
- Location: Nashville, TN
It is a little surprising to see such a big gap between the two, but I wonder if it's just a pre-stabilization premium; I haven't seen venice on newegg, smalldog, or monarch computer. I think just two weeks ago the new OCZ VX memory was selling for 15-20% more than it is today b/c it had just been paper launch. I'm certainly open to other opinions on the pricing, though.
IIRC, Winchester was only a $9 premium over Newcastle; not sure if it was just low b/c of the 90nm die or what.
IIRC, Winchester was only a $9 premium over Newcastle; not sure if it was just low b/c of the 90nm die or what.
PC Perspective review of the 3200+. They're comparing the new CPU with a 130 nm S939 3200+, which doesn't really exist, so they use an underclocked 3800+. However, by doing this and not comparing to the A64D it surely looks good...
Quite unfair comparing it to an unavailable product.
Quite unfair comparing it to an unavailable product.
About availability:
Unfortunately, I can confirm that the Official Street Release time frame is Mid-May for Venice Core A64s. This is when the chips will be available in large quantities. (Retail and OEM but OEM will come sooner) It appears that AMD's wholesalers still have plenty of Winchesters and Newcastle still in stock. From the responses that I got it appears that AMD is indeed waiting to get rid of the existing supply of Athlon 64s before releasing Venice. However, as you may already know, plans might and sometimes do change. I wouldn't be surprised to see AMD let a few trickle into the Market if they found out how hot these chips are.
I was told that if I needed them real badly that I should contact my AMD Sales Representative. This means that if I were to get some I would not be able to get 30+. I would probably only be able to get about 5 or so. Of course, this is no were near enough to feed the masses. Although I have bought things directly from AMD in the past (4-way stuff), it was under special circumstances. I don't know if you guys know the way the Market works but resellers (Newegg, Monarch, etc.) usually never buy chips directly from AMD. We are usually always directed to go through AMD's certified distributors. So don't blame Monarch or NewEgg.
-
- Friend of SPCR
- Posts: 2887
- Joined: Mon Feb 28, 2005 4:21 pm
- Location: New York City zzzz
- Contact:
I dont quite get the reason for the public release of these? not much more performance but um, hotter?
YEAH WAY TO GO WOOO.
I was pricing a new chip/board on monarch today. I cant really figure this revision out. I know it will help in future processor development, but why make it for the average joe?
confused.
YEAH WAY TO GO WOOO.
I was pricing a new chip/board on monarch today. I cant really figure this revision out. I know it will help in future processor development, but why make it for the average joe?
confused.
-
- *Lifetime Patron*
- Posts: 477
- Joined: Thu Feb 17, 2005 6:41 am
- Location: Nashville, TN