anyone recommend a fast mother brd AMD

low latency will then be bottle necked by the lesser bandwidth of the memory.

according to sisoft sandra you always get better memory performance with dual channel DDR, why else would they soon move the Athlon64 to use dual channel if there was no benefit.?

Again i may be missing somthing
 
the bandwidth is only usable IF the cpu takes advantage of it. It's all very well talking about theoretical mamimum bandwidths but that is a world apart from the real-world bandwidth we are talking about here.

The impact of Dual channel on an a64/fx core is minimal. just look at the benchmarks. You are basing this entirly on your P4 EE experiance, well AMD cpu's just don't work like that.

AMD a64/fx's performance is hinged on their memory interfaces. By building the interface directly on to the core itself, they have drastically reduced latencies already. The fact is though, they aren't built for dual channel use so adding dual channel brings virtually no gains to the platform, whilst it DOES induce extra latencies.


So, what do you choose? dual channel + higer latency, or single channel + lower latency? it's a personal choice, but the differences are so small it would again come down to the price of the platforms. An fx +2 sticks of EEC ram is a hell of a lot more expensive than an a64 +1 stick of non-EEC ram.

for all the extra numbers you can achieve with the fx's dual channel, is it really worth it? the difference between the fx-51 and a64 3200+ is less then 500points in 3dmark2001. When your looking at possibly double the price, can you justify that?
 
just got this from a well respected online review site

That leaves us with a clear description of the Athlon64 3400+: it's a top quality CPU that's especially suitable for games and that also lives up to its model name - albeit only in this category. At the end of the day, it still lags slightly behind the Pentium 4, a deficit that the 64-bit architecture could compensate for in the medium term, however. In the short term, Cool & Quiet could do the job, as Intel doesn't offer this type of energy management for desktop processors yet. We can only hope that the motherboard makers take note.


This was pitched agaist a P4 3.2ghz. while i agree the onboard memory controller (hyper transport) is very quick. although nvidia boards messed this up as they dont run at full bits.

while you are quite right in what you say to a point i think your zeal for amd is way over zealous. from the online benchmarks there dosent seem to be much between the two.

i gess i have been basing my results on my overclock which isnt fair as the athlon 64 dosent even come close to it
i think i already conceded i need to brush up on my home work.
 
Originally posted by gandley
while you are quite right in what you say to a point i think your zeal for amd is way over zealous. from the online benchmarks there dosent seem to be much between the two.

Not at all, i just know that the 64/fx's are the better cpu's, and when they move socket and sort out their pci and agp locks, they are only going to get faster, whilst still being half the price.

http://www.nordichardware.com/reviews/cpu/2004/Athlon64_2.2GHz/index.php?ez=6

A fantastic review on a few processors. One thing you have to take not of, is the increase of performance as the a64's ramp up their speed. They actually noted that a 10% increase in mhz gained them closer to 15% in performance in some tests, all down to the on-board controller. They even go as far as to include some beta winxp-64bit benchmarks which show the AMD's to be around 5% faster still on average.
 
but the same is true of the P4, the prescott may have lengthened
the pipe line which gives no improvement now, but when moved to the tejas core and the LGA package socket the P4 will get incredable faster due to the increased pipeline taking advantage of the extra GHZ. then add the almost certain 64bit extensions and amb will have a real fight on its hands.
Many saying on the forums amds little floruish is about to end soon.
just as north wood took the desktop lead after amd got ahead.

this was exactly the case with the north wood core which relaced
willamatte(or what ever is was called)

history seems to be repeating it self as usual. Amd always looses out. which is a crying shame. ive built about 6 amd systems with the Athlon XP chips but have had no requests for the AMD64s.
Most people in the know seem to want to bypass the presscott core at present as well which is no suprise.

Thanks for putting me straught thou, ive just been swotting up now as i have some spare time.

Ill wait to see what the FX-53 can do and what support the socket 939 gets.

otherwise ill guess ill get an alderwood chipset with pci express and LGA socket for the tejas core
 
the difference between the current situation and what happend with the willimette's and northies was that the move to the 0.13 northies brought a clear reduction in tempreature to the table which allow the core's to scale so well.

This time however, it isn't so clear cut. you can only reduce the component size so many times without having to move to an entirly different process to combat heat output. We all ready know that the presscott's are running stupidly hot - a move to a smaller core will only worsen the problem.

Amd got round this by moving the process over to SOI (silicon-on-insulator). Just look at the difference in thermal output between the a64/opteron/fx and the prescotts. Intel will HAVE to address these issue's before the move.
 
but intel already have a working tejas at other 4 ghz.

while speculation, im sure they will sort it for the LGA package.
i think the move to the tejas core as soon as poss is all about the temp issue, as said prescote is nothing short of a disaster for intel. there 90nano process has realy kicked them in the nuts.

but some top info came out of IDF this year, and if i was amd id be very worried, intel has slipped but it does appear to have recoverd its footing.

Still all this is quite exciting stuff. I`ll buy the best thats around at the time, but it does seem obvious that a wait and see approach has emerged if your buying at the bleeding edge.
if your after a mid range system and on a buget amd does seem theonly choice, unless you want rock solid stability over performance then intel is the only choice.

Does anyone know the clockspeed for the 53-fx. im expecting a 3ghz chip here but not sure if amd can deliver this yet?
I mean true 3ghz not PR rating
 
the fx-53 is a 2.4ghz part.
 
The FX and Opteron chips don't have a published PR rating, just a rating relative to the other processors in the range. For comparison the PR ratings for the current Athlon 64s are: 3000+ (2GHz, 1MB cache), 3200+ (2.2GHz, 512KB cache), 3400+ (2.2GHz, 1MB cache). The FX-51 runs at 2.2GHz and has 1MB cache, the FX-53 and A64 3600+ are expected to run at 2.4GHz and have 1MB cache.

Your P4EE couldn't really be considered a normal consumer chip, it's performance (and price) match would be the FX-51 or Opteron x48. Looking at the normal P4s the Althon 64 compares very favourably.

In order to bring some objectivity to this discussion I ran a few quick 3D mark 2001 SE bench marks on my A64 3400+ with Radeon 9800XT. The DDR400 benchmarks are with 1GB and the DDR333 benchmarks are with 1.5GB - the Athlon 64 does not support 3 memory modules at DDR400. The processor is running at 212x11=2332MHz and the graphics card is not overclocked.

DDR400, Radeon on high performance settings: approx 21,200
DDR333, Radeon on high performance settings: approx 21,100
DDR333. Radeon on normal settings: approx 17,500

Memory bandwidth doesn't seem to matter much in this benchmark.
 
i hit about 26000 with an geforce 5950

this with an overclock

fair point on memory. the on chip cache probaly makes more a difference here.

i gess its safe to say as long as it plays half life 2 who bloody cares.

i use ocz gold edition ddr as well. brillant memory. very overclocking friendly. with abit ic7 max3 MB
 
Intel's higher heat output on the Prescott is partly due to
the adoption of "Strained-Silicon" - which whilst being
beneficial for future Ghz scaling, suffers high leakage.

Vis., leakage = high idle wattage = 40W doing nothing.

Agreed, the Prescott is a very hot chip - so much so that
boards should really be operated under stress in a chassis,
so the board itself can benefit from forced air cooling also.


As we move thro 2004 into BTX, and hotter CPU, hotter VRM,
hotter RAM (DDR2 1GB 667Mhz being very hot) so the board
components itself will require heatsinks, heatspreader etc.

Voltage-drop & die-shrink are underperforming frequency as
a rule, and so thermal output continues to rise - and will rise.
Already VRM specs are 103A (!) and that will rise further.

If you are buying a new motherboard or case, might be worth
biting your lip / chewing the sofa and holding off for BTX. AMD &
Intel will release boards for it - I'm not sure that case makers
will create a hybrid ATX/BTX compatible case as with AT/ATX.
BTX flips everything around quite radically to give a fix to the
heat problem - as well as obsoleting everything in sight :)


Intel's P-M is a good comparison of how Intel can get it right,
good performance with very low thermal output (22W). Seems
the old adage of 1/2 performance is possible for <1/5 the power.
Admittedly I/O isn't great - but for blades it could be made good.

So I wouldn't extrapolate too much on the Prescott, but I would
also wait for the socket changes & some clock ramping too. Could
well be the leakage on scaling down to a smaller die was too
much with the Northwood architecture, so Intel created Prescott.
Prescott's deeper pipeline comprehensively eats the cache doubling.

Chip area / Transistor count on Prescott is odd - could well be
64-bit extensions in there, or eventually in there. P4-EE is pricey.

AMD still have a good offering in the Opteron/Athlon64, with the
P4-SMP (Xeon) being really quite jurassic in comparison. Opteron
is the Xeon Intel should have released by now, Tajas is indeed a
good chip - but again heat output thus far is somewhat shocking.

It's also not just the heat output - but the leakage, where the
"sitting idle" CPU itself is kicking out a considerable load of heat.

A case of "Human Inside, Intel Outside" perhaps :)
 
your quite right about the prescott it is a 32/64 bit chip. it was always meant to be. its just the 64bit side is disabled just like intel used to do with multi-threading.
so well spotted

I think the GHZ race will have to end soon or at least slow some.
a plateau has been reached. the present race is unsustainable.
 
Agreed - altho we might see a splitting in the market:
o Centrino was a (huge) step back in clock, small in performance
---- so the mobile mkt isn't competing so much on Ghz
o Desktop/Server has competition for Intel
---- Opteron is basically better than Xeon re 64-bit/RAM/price
---- 3 key areas for multi-cpu dbase apps re datamining/serving
---- Athlon64 is no slouch and lower heat output than P4-Prescott
---- this is a clock-based area

I think Intel may regret it's "clock is faster" marketing.
Laptop Centrino 1.7Ghz hold their own vs Laptop P4 3.0Ghz

Even more so with the efficiency of Itanium2 which despite a
very low clock performs extremely well. New sockets on the way,
which makes the current pricing models look not too great. That
itself may explain the P4-Prescott - smaller die, more yield and
so more room for aggressive price cuts before Q4-04.

Heat output on Prescotts isn't going down well in Racks,
which are already well above original design concepts. It
is known that later chips could well touch 150W, too much.
 

The latest video from AVForums

Is 4K Blu-ray Worth It?
Subscribe to our YouTube channel
Back
Top Bottom