At the start of 1983, SERC had 101 PERQs ordered and 97 delivered. Of the 101, 86 were allocated to projects and 15 were held in Suspense Account against future requirements.
Some of the systems had only 256K bytes of memory and could not use UNIX. The decision had been made to upgrade all systems to at least 0.5 Mbytes and preferably 1 Mbyte.
Some systems required an upgrade from the 4K writeab1e control store (4K WCS) to 16K writeab1e control store to allow additional firmware to be added to support the AI languages. ICL themselves were improving the PERQ performances by adding different floating point firmware, adding virtual memory support and basic graphics operations into the firmware. As the year progressed, the need to upgrade more systems to 16K WCS became apparent.
The problems in 1983 were mainly concerned with upgrading the memory and 16K WCS changes together with the ability to get PERQs to new customers quickly.
In the Spring of 1983, ICL announced the PERQ2 with the intention of producing about 150 systems before the Autumn. SERC had about 15 PERQs awarded on the September 1982 Grant Round that had not been ordered due to SERC's cash flow problems and it was agreed to order these as PERQ1s. We could not delay these until the Autumn.
ICL were prepared to let SERC have 50 out of the first 150 PERQ2s and were prepared to increase the discount to 35% if RAL put a bulk order in for 100 systems. At that time. 55 PERQs were requested on EB grant proposals and conservatively it looked as though a minimum of 20 systems would be awarded in the March grant round. Requiring 100 systems in the year 1983/4 did not seem unreasonable (and actually happened).
Consequently, Bob Hopgood wrote to Dr Manning suggesting that RAL accept ICL's offer.
Dr Manning was reluctant to do this due to the adverse publicity particularly from the Science Board users. By April, the failure to order PERQ systems was causing a backlash from users who could get systems off the shelf from other suppliers but could not get them from SERC because RAL had exhausted the pool around February 1983.
Bob Hopgood wrote to Dr Manning on 25 April 1983 pointing out that RAL were failing to honour their side of the Memorandum of Understanding in terms of systems ordered and people were now having to wait several months because RAL had not ordered any PERQs since 1982. Bob Hopgood urged him to get Swindon to agree a pool of at least 40 systems and preferably 60 or 70. Unless RAL did, they would be in the situation of ICL coming out with a much improved PERQ2 and SERC being the only organisation who did not receive any. Also, the lack of a pool was holding back system development. People locally were having to queue for machines as all the systems allocated to users had been sent out. Bob Hopgood made the point that RAL were not proceeding purely because four or five Science Board users, who had only been using POS and not PNX, were making bad publicity for a project that had turned the corner. Dr Manning wrote to Paul Williams at Swindon on 28 April 1983 requesting that RAL have authority to purchase a pool of 40 PERQ2 systems.
By 9 May 1983, progress on GKS was beginning to suffer as they only had one PERQ for several systems programmers and this was a borrowed system (a DoI OAS system destined for Swindon) which would soon have to be returned. During July and August, it was essential for at least three PERQs to be made available to the GKS team and yet this was almost certainly not going to happen without seriously disrupting another part of the programme.
On 10 May 1983, Dr Manning visited Edinburgh University returning with complaints concerning late delivery, the lack of 16K WCS, the need to contact ICL Dalkeith directly rather than via RAL etc. As a result, he was nervous of any presentation to the Science Board in June. He was also wanting RAL to decommit from Cambridge Ring support.
My reply dated 16 May 1983 said that the only machines delivered late were the ones awarded on grants and this was due to no progress on reinstating the pool. Due to lack of finance by SERC and lack of confidence, nothing had been done to ensure deliveries of systems in reasonable timescales. The lack of a pool of PERQs was causing problems all round. RAL was being blamed because grants involving PERQs had been awarded and yet the culprit was SERC who refused to get a pool of PERQs to allow delivery on grant announcement.
The problem with 16K WCS was that ICL were unhappy with the Three Rivers boards and were having them redone in the UK. They were concentrating on making them available on PERQ2s and were unclear as to whether they would fit them on PERQ1s at all. This was a cause of much discussion between RAL and ICL.
The problems with Local Area Network support were varied and complex. The JNT were still very much insisting that Cambridge Ring should be the UK academic standard. There were technical problems associated with the Ring interface and Z80 firmware on the PERQ. A number of products commissioned by the JNT to attach hosts to Rings had been decommitted. The UK industry was moving to ethernet and had little sympathy with the academic initiatives. The whole question of an academic strategy based on Cambridge Rings was being questioned.
The situation deteriorated during the summer. There were applications for PERQs in the April Grant Round with many awarded.
A bulk order for 20 systems had been approved by Paul Williams on June 15 but by the end of June this was being reconsidered due to yet another financial crisis in SERC. As a result of this, all the April grant announcements were frozen. This solved a short term problem but left the users extremely unhappy.
We still had 11 systems to allocate from the December grant round and the remaining 9 would be swallowed up as soon as they released the April grants.
An order for these 20 systems was delayed and it was not until the end of June that ICL received the order for 20 PERQ1s. By now the poor decision not to order a batch of PERQ2s was becoming very obvious. The ICL production was devoted to PERQ2s and PERQ1s were being imported from Three Rivers. The best RAL could hope for was 10 PERQ1s in September and 10 later. At least one grant announced in December 1982 would wait a year before it was delivered.
By the Autumn, permission had been obtained to order a further 15 systems making 35 systems in total. The only way they could achieve the timescales RAL had been pressing on them was to deliver 16K WCS systems despite the order being for 4K WCS systems (at the time of order the 16K WCS option was not in the price list). RAL were happy to keep these systems but ICL were requesting the additional funds knowing full well that RAL would have ordered 16K WCS versions if they had been available. By November 1983, 15 systems had been delivered. They were manufactured by Three Rivers and had come direct to SERC without testing. Of these, 13 proved to be faulty. They even had US plugs on them!
ICL suggested a deal whereby SERC placed a bulk order for 75 systems (PERQ2 16K WCS) and they would give SERC the 35 16K WCS boards free. This represented an increased discount from 27.5% to 31% for the bulk deal. Less than the 35% already offered but a significant increase any way representing about £70K on the deal.
By October, ICL had decided not to support GKS on 4K WCS systems. Although feasible, to do line drawing and area fill outside the microcode would be extremely inefficient and make GKS unusable. Also, it was impossible to get the various PNX microcode changes and the GKS ones into a 4K WCS system.
The PERQ Business Centre had just opened, and Charles Hughes, the new manager, was presented with the problem. SERC had 130 PERQ systems with 4K WCS and, at the current prices, this would cost SERC £400K to upgrade to 16K WCS. As the upgrade was necessary to run the software that had been produced jointly by SERC and ICL, there was clearly a problem. It might lead to SERC withdrawing ICL's ability to provide GKS on the PERQ unless the situation was resolved.
In November, Ken Robinson wrote to Dr Manning urging that RAL move to 16K WCS systems as the standard. Single precision floating point speeds had been increased by a factor of 3 but needed the 16K WCS to operate. GKS performance could be increased 10 times by having 16K WCS. Also, ICL were making no commitment to running PNX 2.0 on a 4K WCS system.
There was still a backlog of unordered systems. At least 20 systems needed to be ordered before the end of November if a February delivery and payment in the current financial year was to be obtained.
Central Computing Committee at its Autumn meeting had agreed in principle, that a pool of 20 PERQs be re-established. Thus in November, 132 PERQs had been ordered and delivered. We had authority from DES to purchase a further 68 systems and had a need for at least 40 systems.
The situation was complicated by ICL now having delivered 37 PERQ1s with 16K WCS although the orders were only for 4K WCS systems. ICL made a variety of proposals concerning bulk orders in order to allow us to keep the existing 16K WCS systems. The most attractive was to order 68 systems in which case all the 16K WCS boards delivered would be given to SERC free (an additional discount of £70K).
Dr Manning was unenthusiastic about making the 16K WCS system the standard or making a bulk order for 68 systems. He would be prepared to order 40 systems with a letter of intent for 75 (no real commitment) provided ICL gave us the deal on the basis of 75 systems. He was still worried about the adverse publicity particularly from the Science Board area and wanted to avoid a high profile purchase which would raise the question of whether PERQ was the right system to purchase.
The large order was never placed (only 40 were ordered) and the 16K WCS upgrades were paid for at a reduced rate agreed between SERC and RAL. Upgrading the other systems proved a long drawn out discussion lasting another year with Boards other than Engineering reluctant to pay for the upgrade initially.
It was not until the end of 1985 that all the PERQ systems were upgraded to 16K WCS. For a significant part of their life, many systems were unable to work at full speed and were criticised for poor performance. This was partly due to lack of faith by the Boards in the revised standard system.
